ABSTRACT. Spatiotemporal Traffic Data (STTD) measures the complex dynamical behaviors of the multiscale traffic system. Existing methods aim to reconstruct STTD using low-dimensional models. However, they are limited to data-specific dimensions or source-dependent patterns, restricting them from unifying representations. Here, we present a novel paradigm to address the STTD learning problem by parameterizing STTD as an implicit neural representation. To discern the underlying dynamics, we employ coordinate-based neural networks to directly map coordinates to traffic states. To unravel the entangled spatial-temporal interactions, we decompose the variability into separate processes. We further enable modeling in irregular spaces such as sensor graphs using spectral embedding. By integrating a meta-learning strategy, we decode instance-wise patterns to achieve efficient generalization across different instances. Through continuous representations, our approach enables the modeling of a variety of STTD with a unified input, thereby serving as a general learner of traffic dynamics. It also possesses properties such as implicit low-rankness, inherent smoothness, and reduced complexity, making it versatile for a wide range of practical applications. We validate its effectiveness through extensive experiments in real-world scenarios, showcasing applications from corridor to network scales. Empirical results not only indicate that our model has significant superiority over conventional low-rank models, but also highlight that the versatility of the methodology extends to different data domains, output resolutions, and network topologies. Comprehensive model analyzes provide further insight into the inductive bias of STTD. We anticipate that this pioneering modeling perspective could lay the foundation for universal representation of STTD in various real-world tasks.

ABSTRACT. Recent advancements in Connected Vehicle (CV) technology have prompted research on leveraging CV data for more effective traffic management. Despite the low penetration rate, such detailed CV data has demonstrated great potential in improving traffic signal performance. However, existing studies share a common shortcoming in that they all ignore traffic flow estimation errors in their modeling process, which is inevitable due to the sampling observation nature of CVs. This study proposes a CV data-driven robust optimization framework for traffic signal timing accounting for both traffic flow variability and estimation errors. First, we propose a novel data-driven uncertainty set of arrival rates based on the bounds information derived from CVs, which circumvents the error-prone arrival rate estimation process. Then, a generalized CV data-driven robust optimization model (CV-RO) that can be widely applied to various signalized intersection scenarios, including under-/over-saturated and fixed-/real-time, is formulated to explicitly handle arrival rate uncertainties. By means of the robust counterpart approach, this robust optimization problem can be equalized to a deterministic mixed-integer linear programming problem with an exact solution. The evaluation results highlight the superior performance of the CV-RO model compared to the deterministic model and traditional methods across various scenarios: different penetration rates, traffic demands, and control types. Notably, the CV-RO model demonstrates its excellence at lower CV penetration rates and in the presence of different traffic flow fluctuation levels, affirming its effectiveness and robustness.

ABSTRACT. This paper proposes a new spatial-temporal framework to predict the number of accidents in an urban area, by integrating multiple heterogeneous data sets such as road structure, intersection density, weather, historical collisions, traffic information and victim and party demographics data. Firstly we collect all data sets for the area of Los Angeles, which is our case study. Secondly, we propose a deep learning framework and efficient pipeline based on a three-dimensional convolutional autoencoder that determines the optimal size and sequential combinations of input for predicting the number of accidents to take place in a subarea at a given point in time. The framework is further improved based on victim and demographic party data, which indicates that the age of 23 is common among victims (mostly young men aged between 18-35 years old). Thirdly, we show that our proposed architecture outperforms other baseline models such as stacked denoising autoencoder, multilayer perceptron and convolutional long short-term memory network in terms of root mean square error and symmetrical mean absolute percentage error.

ABSTRACT. The generation of realistic samples of synthetic activity schedules is a critical component of both Activity-Based Models (ABMs) and simulation-based transport approaches such as MATSim. Conventional models decompose the scheduling process into series of discrete choices, applied sequentially. Realistic interactions between these choices are therefore limited. The combination of multiple interacting sub-models is also challenging to develop and use.

We demonstrate an alternative approach to activity sequence modeling using deep generative learning. Our model is able to rapidly synthesise activity sequences for new populations of socio-demographic attributes. This allows applications for (i) data anonymisation, (ii) diverse up-sampling, (iii) bias correction, and (iv) forecasting based on demographic shift. We comprehensively evaluate the quality of output activity schedules both individually and in aggregate.

We identify three primary benefits of our approach; (i) simplicity and speed, (ii) diversity in output sequences, and (iii) the potential for more realistic interaction of choice components. We have created the open-source Python software CAVEAT (https://github.com/fredshone/caveat) for the development and evaluation of generative activity models.

ABSTRACT. Technological advancements and interventions in the transportation sector play a crucial role in addressing climate change, given its major contribution to greenhouse gas emissions. The industry actively explores electrification, automation, and Intelligent Infrastructure to mitigate emissions. However, the successful design and implementation of these solutions require accurate and representative emission models.

The Motor Vehicle Emission Simulation (MOVES) serves as the gold standard emission software provided by the Environmental Protection Agency (EPA). Despite its prominence, using MOVES faces challenges, including a steep learning curve and technical complexities. This makes it cumbersome for macroscopic analysis and unsuitable for microscopic analyses like eco-driving, which demands emissions estimation for individual possible actions at each time step.

Efforts have attempted to address these needs by either pre-calculating emissions data or simplifying the MOVES framework. However, these MOVES variants either require extensive storage space and MOVES expertise, or sacrifice accuracy and comprehensiveness for accessibility and speed.

To address these issues, we present NeuralMOVES, a comprehensive family of high-performance and lightweight CO$_2$ emission models devised through reverse engineering MOVES and surrogate learning. Our models show a promising 6\% end-to-end error relative to MOVES, exhibit significant differences from alternative reduced-order models, and offer improved precision.

The implications of our work are twofold: our models simplify GHG emission evaluation in transportation-related analyses by providing a faster, programmatic alternative to MOVES and improve control-based approaches by offering microscopic and environment feature-rich models compared to alternative models.

ABSTRACT. In this paper, we extend the hybrid simulator from the individual to the household level by including a broader set of simulated demographic events affecting households and redefining a resampling procedure using the Gibbs Sampler. Usually, projection methods use historical demographic rates that may not account for sudden events like COVID-19, potentially hindering the accuracy of transportation models that rely on these projections. To test the resilience of projection methods to unforeseen events, we project synthetic samples from 2010 to 2021 using dynamic projection and a hybrid simulator. We test two scenarios based on pre-pandemic and post-pandemic demographic rates using Swiss Mobility and Transport Microcensus data. The results show that the hybrid simulator is more robust and less dependent on rates when it comes to unforeseen events than dynamic projection as it includes an intermediate resampling update that helps reduce the errors of dynamic projection.

ABSTRACT. Choice-based optimization problem integrates demand modeling into optimal supply decisions, which is generic for decision-making applications. Solving the problem is challenging given the nonlinear discrete choice model constraints. Existing solution methods are limited to specific problem structures, such as binary or discrete supply decisions and fixed option attributes. This paper proposes an outer-inner approximation method for the generic choice-based optimization problem without specific problem structural requirements. We validated the method using a network expansion problem on the SiouxFalls network, aiming to reduce the overall system congestion by optimally expanding road capacities considering the road expansion cost. The results show that the expansion cost is significantly lower than the total travel time savings. More experiments are expected to benchmark with existing models using more case studies, e.g., service frequency and pricing in multimodal transportation systems.

ABSTRACT. Modeling pedestrian entrance scenarios is a central focus in the field of pedestrian dynamics, yet existing models, rooted in physics, have limitations when it comes to incorporating psychological aspects of individual behavior. Despite prior efforts to integrate certain psychological concepts, this interdisciplinary perspective is relatively new, and there is room for further exploration. This study aims to initiate a discourse on the integration of motivation into pedestrian models. Motivation is believed to be one of the most apparent psychological drivers of movement behavior in pedestrian environments, capable of significantly influencing crowd dynamics. While previous approaches have often employed a simplified binary categorization of motivation, classifying agents as either highly motivated or lowly motivated, this simplification, while useful in many contexts, fails to capture the complexity of motivation, which is influenced by a multitude of intrinsic and environmental factors. We introduce two critical dimensions of motivation: heterogeneity (variations in individual motivation levels within the crowd) and dynamism (fluctuations in motivation levels during goal pursuit) to establish a foundation for modeling motivation in entrance scenarios. Incorporating these dimensions alongside related psychological motivation theories, we present a simple velocity-based model.

ABSTRACT. With increasing deployment of automated vehicles (AVs), it is crucial to understand how they will interact with human-driven vehicles (HDVs). We investigated the gap acceptance behavior of HDVs in mixed traffic at a priority T-intersection. We adopted a driving simulator, in which drivers drove four scenarios, that varied on the driving style (less defensive, more defensive) of AVs and on their recognizability (distinguishable or not from HDVs). Then, using the collected data, we estimated gap acceptance models, which we then implemented in the SUMO microscopic traffic simulation platform, in which we built a T-intersection network. We ran simulation runs differing in AV driving style, recognizability, penetration rate (0-75% in 25% increments), and whether HDV behavioral adaptation was considered. HDVs were present on the minor and major road, while AVs were only present on the major road. Vehicle delays and queue length were measured as performance indicators. We found that the delay for minor road vehicles increased with higher penetrate rate of AVs. Moreover, the minor road delays were larger when AVs were recognizable and less defensive. We also found that if behavioral adaptation was ignored, then the delay for minor road vehicles would be underestimated by up to 75%. In conclusion, we observed HDVs’ behavioral adaptation in gap acceptance behavior in mixed traffic. The recognizability of AVs, their driving style, and their penetration rate affected HDV behavior. If we want to predict and assess efficiency of mixed traffic, then considering behavioral adaptation is important.

ABSTRACT. The development of autonomous driving technology offers new opportunities for traffic management, with connected and automated vehicles (CAVs) expected to liberate drivers for autonomous cruising. However, the continued need for short-distance travel by foot or bicycle will result in traffic flows comprising both vulnerable road users (VRUs) and CAVs. Managing their interactions, particularly at accident-prone automated unsignalized intersections, poses a significant challenge: developing a conflict-free and efficient operational strategy in such multi-player, dynamic, and continuous interactions. In this work, responding to these practical needs and research challenges, we propose a dynamic interaction-based cooperative decision-making approach for heterogeneous traffic participants (DI-CDM). We introduce a differential game framework aimed at analyzing interactions between VRUs and CAVs, to ensure the safe and efficient operation of unsignalized intersections. The results demonstrate that the proposed DI-DCDM exhibits promising performance in managing interactions among heterogeneous traffic participants and generating conflict-free trajectories.

ABSTRACT. While studying the interactions between pedestrians and urban traffic remains important, there's a growing urgency to investigate pedestrian behaviour within the framework of automated vehicles (AVs). This study uses virtual reality (VR) to simulate two urban mid-block environments, one in downtown Toronto, Canada and the other in central Newcastle, UK, to investigate the crossing behaviour of 428 participants (9,092 observations). To the best of our knowledge, this is the largest study of such kind conducted on two different continents with different traffic rules and walking/driving norms. The research questions addressed in this paper, in the context of unmarked mid-block crossing, are: (a) Do various vehicle types, i.e., normal vehicles and AVs, impact pedestrian behaviour? (b) How do other pedestrians influence one's crossing behaviour? (c) How do traffic characteristics, road type, and environmental characteristics impact pedestrian behaviour? (d) What is the influence of demographics on pedestrian behaviour? and, Are there differences in pedestrian behaviour in different countries? We use advanced econometrics and data driven machine learning to answer these questions.

ABSTRACT. In the field of autonomous driving, precise evaluation of drivers' cognitive states through physiological indicators is essential for advancing safety and vehicle interaction. This investigation assesses the association between eye movement patterns and cognitive states across varied driving scenarios, including Non-Driving Related Tasks (NDRT), light intensity variations, and different driving modes. Eye movement data were collected from experienced drivers during simulations, utilizing Tobii Pro Glasses 3. The data analysis was conducted using SHapley Additive exPlanations (SHAP) to interpret the predictive power of eye movement features within a Light Gradient Boosting Machine (LightGBM) model comprising nine distinct features. The findings highlight that fixation-related features significantly influence cognitive load states. Further analysis involved clustering fixation data into seven areas, which served as inputs for a Markov chain model to simulate transition patterns among cognitive states. The study revealed that increased cognitive load correlates with more frequent checks of the dashboard, whereas lower cognitive load corresponds with increased focus on the road ahead. Additionally, the effects of NDRT and light intensity on eye movement patterns were more pronounced than those of the driving mode. These insights underline the potential of utilizing eye movement metrics to enhance safety features in autonomous vehicles, reflecting the unconscious behaviors of drivers under different cognitive loads.

ABSTRACT. Heterogeneity has been a longstanding characteristic of human-driven vehicle traffic. It is set to be compounded by the growing adoption of automated vehicles (AVs) and the emergence of mixed traffic. The behaviors exhibited by AVs differ among themselves and notably with human driven vehicles (HDVs), highlighting the complex and heterogeneous nature of mixed traffic.

Central to our work here is the challenge of data-driven learning under such heterogeneity. Specifically we recognize that some methods of aggregating data across diverse traffic agents (e.g., different human drivers) to develop predictive models (e.g., driving models) encounter significant limitations. Such methods often fail to accommodate the distinct behaviors and preferences of individual agents. Misalignment between how traffic agents desire to behave versus how we model them to can mislead both data-driven predictions and traffic-scientists alike.

We propose a framework that employs a personalized modeling\footnote{personalization refers to models that adapt their learning to agents' behavior and preferences} approach to learn models for traffic agents\footnote{by traffic agents we mainly refer to HDVs and AVs, yet we leave the notation general as the modeling framework is not restricted to particular agents. For instance, bicyclist can be seen as traffic agents.}. This approach is motivated by (i) the heterogeneous nature of driving data and challenges it poses on data-driven learning, and (ii) the distinctive requirements of mixed-traffic environments, which demand effective human-machine coexistence and behavioral alignment.

Personalization offers a valuable opportunity to identify and utilize behavioral differences among traffic agents, and leverage them for improved downstream analytics (i.e., operation management).

ABSTRACT. This study aims to illustrate the critical role of user psychological factors in travel choice and travel behaviour via social media data and natural language progressing methodologies. By integrating social media with survey data to obtain more comprehensive psychological features including dynamic user attitude and user concerns through sentiment analysis and dynamic topic modelling, we are able to demonstrate how these features help in identifying underlying causes and key determinants of both consistency and inconsistency travel behaviour within neighborhood level and different periods. We implement a case study in New York City via Twitter data to capture transit-related psychological factors on five travel modes (cycling, driving, subway, taxi and ride-hailing, walking) from 2019 to 2022. Results show the advanced ability of social media data in capturing dynamic travel-related user attitudes and concerns and further support travel choice dissonance identification and dynamic travel behaviour explanation among five travel modes. Consequently, the outcome efficiently reflects various social needs, identifies priority areas and provides valuable suggestions for policy-makers and planners to develop more targeted improvement strategies.

ABSTRACT. The causal graph is an indispensable part of causal analysis, which plays a considerable role in congestion and accident investigation, as it provides an intuitive form of causal correlations in traffic networks and the propagation process of causal impacts. Previous studies have focused on generating the causal graph using information theory and model-based approaches, which ignore region-wise mutual information and network topology, hindering the comprehensive understanding of the causal relations. To address such limitations, this work provides a new perspective to achieve region-wise causal inference of complex traffic networks, enhancing verifiability and interpretability. Specifically, the trainable causal entropy algorithm is proposed to obtain the initial causal matrix, considering the topology structure of the network. A causal-integrated Graph Convolutional Network (GCN) with an attention mechanism is then designed to capture region-wise causal effects by information aggregation. The initial causal graph is iteratively refined within an end-to-end framework. The effectiveness of the proposed method in capturing causal relations is evaluated on a real-world dataset through the improvement observed in traffic forecasting accuracy. Additionally, the distinct visualizations of the symmetric topology graph and the asymmetric causal graph reveal the complex interplay of traffic dynamics, highlighting the region-wise causal impacts that shape the underlying relations within the traffic network.

ABSTRACT. This paper aims to understand how physics-based and AI models interact and supplement each other in modeling car-following (CF) behavior. Physics-based CF models are grounded in robust theoretical frameworks, offering a level of interpretability and trust in their output. Nonetheless, achieving a simplified theoretical exploration comes at the expense of modeling complex dynamics and interactions in generalized traffic environments. This makes adopting traditional CF models to automated vehicles (AVs) control more challenging, given the crucial need to realize generalizability. Conversely, the emerging AI models (i.e., deep learning (DL)) offer unique advantages in learning and adapting from data, particularly capturing non-linear complex relationships. Yet, this requires a vast amount of high-quality training data that is often limited or unavailable. More importantly, the AI methods lead to models with few to no interpretable insights, impairing traffic-level understanding. Consequently, we see a shift in adopting AI-based models to supplement traditional CF models – where they might fail – giving rise to physics-informed artificial intelligence (AI) models, most famously physics-informed neural networks (PINN) (Mo et al., 2021, Cuomo et al., 2022). The rigorous analysis of their approximation efficiency remains unexplored. In this work, we want to understand the behavior of these physics-informed AI models. This entails analyzing how AI models interact with traditional CF models and what expectations we have on the synergy between the two. Incorporating physics-informed AI into CF models brings multiple dimensions of analysis: predictive accuracy, data needs, convergence, interpretability, etc. All these play a role in physics-AI synergy. However, in this abstract we focus on the accuracy metric. Our interest here is to investigate how the underlying physical features in different CF models across different families synergies with AI models. We do so through theoretical error bound modeling and sensitivity-based simulations.

ABSTRACT. The Kardar-Parisi-Zhang universality class (KPZ) has long been recognized for correctly describing the scaling behavior of the simplest theoretical traffic models, yet the empirical verification of this remarkable fact remains open. The only empirical indication so far has used aggregate estimates of commuter delays in US cities to show that the KPZ conjecture explains the observed super-linear scaling of delays with population size. In this paper, we use the most extensive vehicle trajectory dataset to date (I-24 MOTION) to report the strong empirical evidence that the KPZ conjecture should be the case: traffic flow near critical density indeed follows the scaling behavior predicted by KPZ. This important result is bound to have profound implications in our field since a paradigm change is needed that opens new avenues for developing effective traffic management strategies that leverage the scaling properties of real traffic flow. 

ABSTRACT. In the extended abstract we present a partial differential equation (PDE) model consisting of three parts, a crowd flow dynamics component, an epidemics spreading model, and an equation that provides the velocity field due to ventilation. Using a specific numerical implementation, we present preliminary tests, in which the crowd is moving towards an exit that is assumed to be the whole right boundary of the computational domain. The numerical results illustrate the behavior of epidemics transport under movements of pedestrians in a closed area, in the presence of ventilation.

ABSTRACT. In this paper, we question the assumption of “small” disturbances in the study of string stability of car-following models and ACC systems, which is at the basis of analytical approaches to linear analysis. This assumption is in fact unrealistic in real traffic operations, where controllers are typically nonlinear and disturbances are “large”; see sharp decelerations. Therefore, evidence of the inaccuracy of existing analytical methods for nonlinear analysis and their inapplicability to a large class of car-following models is first presented. A simulation framework is then proposed and applied to identify the limits of linear analysis and to disentangle the occurrence of rear-end collisions from that of string instability. Metastability thresholds, which define how large a perturbation must be for a linearly stable model to become unstable, are calculated at different equilibrium states for the IDM and Gipps’ model. Analytical conditions for metastability as a function of model parameters are also derived. The analysis showed that no collisions due to the amplification of a disturbance occurred for both metastable and linearly unstable parameters, even when the deceleration output of the two models was constrained not to exceed normally achievable values. However, the introduction of a perception delay made it possible to obtain vehicle collisions already in the metastable region. Finally, the study addresses a major criticism of Gipps’ model. When string stability is measured by the L_∞ norm, Gipps’ model becomes unstable for decreasing values of the equilibrium speed – in agreement with real traffic observations – both in the metastable region and in the linear instability region.

ABSTRACT. Network signal control is an effective way to deal with nonlinear congestion dynamics in urban networks. However, most network signal control methods ignore queue spillback. Few studies that considered queue dynamics assume homogeneous traffic states across different movement directions in one road segment, resulting in the underrepresentation of spillback. This study proposes a novel hierarchical model predictive control (MPC) approach that utilizes the link transmission model (LTM) with queue transmission at both road segment and turn levels to optimize signal timing for urban networks. At the network level, the controller employs a state-of-the-art LTM that can describe segment-level flow dynamics and uses quadratic programming to determine the reference effective fractions of green time for network throughput maximization. The local layer builds on a newly developed LTM with turn-level queue transmission and uses nonlinear programming to track the reference of green time to ensure that the optimization by the network layer is realized at the individual intersections while eliminating potential spillbacks. Simulation experiments on an arterial are conducted to verify the performance of the proposed approach. The results reveal that the proposed MPC leads to significant improvements in traffic performance in terms of throughput and queue length, especially in oversaturated conditions.

ABSTRACT. Electric Vehicle (EV) technology presents an effective solution to mitigate traffic-induced air pollution in urban areas, especially when paired with green energy production. However, the Market Penetration Rate (MPR) of these vehicles is not expected to surge in the near future, leading to a prolonged period of mixed traffic with both EVs and Regular Vehicles (RVs). Given the uneven distribution of congestion and its related environmental issues across the network, there is a need for innovative traffic management strategies, which prioritize EVs in areas with high levels of congestion and pollution. The concept of the Network-wide Fundamental Diagram (NFD), or Macroscopic Fundamental Diagram (MFD), has enabled transportation planners to develop regional traffic management strategies, including perimeter control. This study introduces an innovative control strategy based on dedicating a portion of the links on the periphery of the network regions, by which the vehicles can transfer between different regions of the network, to EVs, in order to minimize the total time spent by vehicles as well as the vehicular emissions over the network. Unlike existing models that do not differentiate between RVs and EVs in perimeter control, this approach specifically addresses the heterogeneous distribution of traffic and environmental concerns, particularly in Central Business District (CBD) areas, by controlling the accumulations of RVs and EVs in each region in large-scale realistic networks. A Model Predictive Control (MPC) method is used to solve the optimization problem. Also, a mesoscopic traffic simulation tool, DYNASMART-P, is incorporated to the MPC approach to simulate the movement of vehicles throughout the network. The proposed framework is applied to the city of Chicago network.

ABSTRACT. Truck platooning (TP) is defined as the linking of two or more trucks in convoy, with different levels of automation. The paper tackles the problem of optimising network-wide TP1, leveraging a graph-based formulation of the problem. The problem can be conveniently formulated on a daily basis for pre-trip/offline TP level 1, that is with all trip requests for day d+1 available at the end of day d, seeking to optimise their overlapping in platoons..The paper proposes a formulation based on the optimal partitioning of graphs. In particular, a graph G is built, with as many nodes as the cardinality of R, such that each node represents a trip request. The proposed approach has been tested on the challenging test site of the Milano-Napoli A1 motorway in Italy, for a length of about 900 km and 97 entries/exits. Real truck trip data for an entire day, together with dynamic travel times, have been provided by the motorway operator Autostrade per l’Italia.

ABSTRACT. Modular bus technologies are capturing the interest of transit practitioners due to their operational flexibility and commercial speed improvements, with prospective initiatives contributing to characterise the real performance of modular bus operation (Khan et al., 2023). However, there are still some doubts on the acceptability of the proposed strategies regarding the user experience and operational costs. This research aims to explore the potential of modular buses in an urban bus route with branches, by comparing the operating cost and temporal performance of conventional bus technologies (S0) to two scenarios with modular bus operation. In Scenario S1 regular buses are replaced by modular bus convoys with a constant length. In Scenario S2, modular buses may couple and decouple each other for adapting the convoy length to the route shape and demand variability in route segments. The modelled bus route presents a Y-shaped layout with different demand rates in each route segment. The branching scheme proposed is common in conventional transit services, when lines leave the crowded central areas in the city and branch out, providing affordable services in the city’s periphery (see Daganzo, 2010). Despite the operating cost savings, the trips whose origins or destinations in the branches are served with low frequencies, worsening their waiting times, in comparison to the central segment. Hence, modular buses may neutralise the expected increase of waiting times by just decoupling modules at branching points. The main contribution of this paper is focused on the identification of the required situations where the scenario S2 would be cost-efficient.

ABSTRACT. The interest of researchers in the problems related of ship routing is continuously renewed by the progressive technological advances and novel regulations and measures. In recent years, the environmental issues related to the preservation of seas and oceans are receiving growing attention. This has lead for instance to the designation by IMO of the Emission Control Areas (ECA), and the Particularly Sensitive Control Areas (PSSA), in which navigation is regulated by specific rules. In this work, we focus on the Mediterranean Sea, which is included among the four lighthouse areas of the EU Mission ‘Restore our ocean and waters by 2030’. Besides, the Mediterranean Sea as a whole SECA under MARPOL Annex VI will enter into force next year, and it already houses two PSSAs: the Strait of Bonifacio Area and the North-Western Mediterranean Area. We present a novel optimization problem for maritime transport under PSSA constraints, and we develop a Mixed-Integer Linear Programming mathematical model. Through simulation, we show that the developed approach enables ship owners to minimize costs while complying with PSSA regulations.

ABSTRACT. Emerging technologies offer new alternatives for providing OFD services, and a sidewalk delivery robot (SDR) system is one such option. With the development of autonomous driving technologies, box-sized robots navigating through sidewalks for food and grocery deliveries have become real-world applications. Analytical tools are needed to optimize sidewalk robot online food delivery (SROFD) systems at scale. An SROFD operator needs to determine the optimal resources required to properly serve an area with uncertain demands. Such decisions include charging infrastructure allocation to different depots and robot fleet sizing. We propose a two-stage stochastic problem that optimizes the resource allocation strategy of a multi-depot SROFD system. Each depot has its respective SDR fleet serving a dedicated area. A subset of depots also has battery swapping facilities, which has a fixed time for battery swapping and replenishes batteries to full. The first stage problem determines the optimal fleet size for each depot and which stations have battery-swapping capabilities. The chosen depots are assumed to have unlimited battery supply. The allocation of the battery swapping stations to the depots impacts the routing patterns of the robots. The second stage solves the routing problem with a late arrival penalty based on the first stage solution. The two-stage stochastic programming problem is solved using an original solution algorithm that uses continuous approximation (CA) in the first stage to quickly obtain a reduced solution space from the first stage problem. For the second stage, we propose a heuristic algorithm that provides high quality solutions under a short running time based on the first stage solution space. The second stage results can be used to further guide and adjust the CA model used in the first stage model. Finally, the sample average approximation (SAA) methods is applied and we choose the solution with the lowest average second stage objective value as the final result. Randomized instances are generated to evaluate the computational efficiency and accuracy of the solution algorithm against benchmarks. Solutions directly solved by existing solvers will be used as benchmarks. In addition, a large instance with more than a hundred orders and multiple depots will be solved and presented to showcase the model’s capability. By using the proposed approach, the stochastic model can be expanded with other factors involved such as battery swapping facility with battery capacity and adding battery capacity as part of the decision variables.

ABSTRACT. This paper proposes a privacy-preserving collaborative traffic signal control framework to fuse data from multiple data owners, such as municipal authorities (MAs) and mobility providers (MPs). Such a framework can exploit the benefits of trajectory data in traffic control before the massive deployment of connected vehicles. Unlike existing methods that assume the data of all parties to be accessible by a single trusted data owner, we explicitly address data privacy concerns of MAs and MPs via a novel vertical federated reinforcement learning algorithm, VFedLight. This algorithm leverages the trajectory data provided by MPs and loop detector data provided by MAs to learn an efficient control policy. Results show that VFedlight can yield satisfactory control performance while protecting MPs' privacy.

ABSTRACT. In future scenarios where 100% of vehicles are connected and automated, traditional traffic signals might become obsolete. Instead, so-called autonomous intersection management (AIM) schemes could coordinate traffic at intersections through communication between vehicles and the infrastructure. Although multimodality is central to urban transportation, other road users have rarely been integrated into AIM systems. This paper presents an optimization-based AIM approach that integrates multimodal and heterogeneous traffic, including cars, trucks, buses, pedestrians, and cyclists. Results obtained by simulation of a real-world intersection demonstrate the enormous potential for reducing delays and illustrate the effective prioritization of public transport vehicles. The proposed AIM scheme leads to vehicle and bus delay reductions of more than 70%. At the same time, it keeps pedestrian and cyclist delays on approximately the same level as with the currently implemented fully actuated traffic signal control. Furthermore, energy consumption and driving discomfort can be substantially decreased.

ABSTRACT. Optimal speed advisory algorithms aim to predict the passing speed of vehicles entering approaches to the intersection and then transfer this information to the target vehicles equipped with the OBU. Various algorithms have been introduced to minimize fuel consumption while considering road conditions and incorporating more realistic traffic situations. To verify the excellent performance of the proposed algorithms in reducing urban traffic congestion and fuel consumption, microscopic simulation-based validation methods have proven to be more useful in Katsaros et al. (2011) and Coppola et al. (2022). Nevertheless, several algorithms have been implemented without considering diverse traffic signal control strategies and real-time traffic conditions in the simulation environment. Therefore, this study aims to construct an actuated signal control policy with the improved GLOSA algorithm considering real-time queue lengths at isolated intersections. We present three unique contributions: 1) The GLOSA incorporates queue length estimation methods to reflect the actual conditions by applying spatial and temporal ranges, which overcomes over-calculated speed advice that can enhance the accuracy of advised optimal speed to the vehicle. We call this creative algorithm the Q-GLOSA, 2) The Q-GLOSA is deployed in the actuated signal control policy to maximize operational efficiency for a signalized corridor, and 3) The proposed mathematical framework is validated in the integrated microscopic traffic simulation, SUMO, under diverse traffic demand and V2X communication scenarios.

ABSTRACT. This research introduces a novel approach that integrates machine learning to enhance the column generation method for vaccine distribution. The objective is to optimize location and allocation decisions, thereby improving vaccine delivery's efficiency and demand coverage. Utilizing a modified Voronoi Diagram for initial data analysis, our model significantly reduces transportation costs and improves demand coverage compared to traditional methods. We employ a machine learning-based solution prediction framework, which includes training on known optimal solutions and predicting unknown optimal solutions for new problem instances. Following a two-phase approach, the proposed methodology allows for efficient adjustments in real-time operational settings. Integrating machine learning helps refine the decision-making process, significantly reducing computational time from initial projections while maintaining solution quality. This method demonstrates the potential for broader application in logistics and supply chain challenges beyond healthcare, providing a scalable model for complex optimization problems in diverse sectors.

ABSTRACT. Accurate traffic prediction is crucial for efficient transportation planning, reducing congestion, and optimizing travel time. Deep learning advancements have yielded effective traffic prediction models. However, their success depends on extensive datasets from diverse sources to capture intricate traffic relationships. Gathering such data poses challenges in security, storage, and resource consumption. Federated learning, enabling prediction with locally stored data, offers a promising solution. This paper introduces a federated learning-based heterogeneous spatial-temporal graph neural network (FLHSTGCN) for traffic prediction, addressing limitations in applying diverse data sources. Key contributions include:

1. We propose the FLHSTGCNmodel for traffic prediction, which integrates a federated learning framework with the heterogeneous spatial-temporal graph neural network (HSTGCN) model. Our method enables accurate traffic prediction by training models across different clients while maintaining the security of raw data. The performance of our model showcases its excellent capacity in balancing accurate traffic prediction and data security.

2. In the proposed federated learning framework, we employ an adaptive adjacency matrix to capture the hidden spatial correlations among subnetworks of all the clients. Additionally, we introduce a FedGSAM Aggregation Mechanism to aggregate the adaptive spatial correlations uploaded from the clients, thereby enabling the proper capture of the global spatial relationships of the network.

ABSTRACT. This study examines the relationship between mean and variance of travel times on a congested corridor using LWR theory, focusing on a freeway with a bottleneck and stochastic peak demand. We establish conditions for typical counterclockwise hysteresis loops and explain why deviations remain limited. Supported by numerical experiments, our results enhance understanding of hysteresis patterns and aid in traffic planning and control

ABSTRACT. Modeling public transit operations poses challenges because of demand fluctuations and supply variations, unpredictable external disruptions, and varying behaviors of passengers and drivers. Traditional analysis tools, such as macroscopic and mesoscopic simulations, often oversimplify these complexities by assuming deterministic patterns that may not adequately reflect real-world randomness and external effects. Microscopic simulations provide detailed insights of specific domain such as a single intersection or public transport stop but struggle with scalability and may not fully capture how micro-level variations impact the overall system performance and service reliability. In contrast, data-driven approaches utilizing machine learning and deep learning have the potential to recognize and use historical traffic patterns. Yet, they often fall short in connecting these insights across different network components. Therefore, this study introduces a novel approach that integrates microscopic simulation with data-driven machine learning and deep learning (ML-DL) to bridge the gap between detailed specific simulation and the need for its related big picture considering more input and modeling. The study categorizes transit operational components using primary and secondary activities. Primary components are involved with explicit vehicle and passenger movements while secondary components involve implicit implications of service reliability. The research framework comprises three core mechanisms or engines: the data processing and fusion engine (DPF-E), the AI engine (AI-E), and the simulation engine (SM-E) that work together in real-time. The AI-E predicts day-to-day fluctuations and patterns based on historical data, while the SM-E connects these fluctuations with the microscopic movements of the primary elements to capture their interrelationships and to find the cumulative impact of these elements on the overall system performance and service reliability. This approach effectively reproduces historical patterns, replicating passenger demand at each stop in terms of arrival rate, boarding and alighting activities, determining dwell time, headway variability and more parameters. A case study on Melbourne's Tram Route 96 using the AMATS simulator demonstrates the successful application of this AI-powered simulator. Results indicate a high level of accuracy (90-99%) in replicating complicated historical patterns and behaviors, including demand fluctuations, dwell time variability, and identification of frequent bunching events concerning time and locations. The model distinguishes between recurrent and non-recurrent reliability issues for the implementation of effective control strategies to improve operational reliability. The integration between an AI engine and the simulation enables to accurately represent the patterns and variabilities of the data and consequently to activate control strategies to solve transit reliability issues.

ABSTRACT. Demand-responsive connector (DRC) services are increasingly recognized for their convenience, comfort, and efficiency, offering seamless integrations between travelers’ origins/destinations and major transit hubs such as rail stations. Existing studies on DRC optimization often center on fixed occupancy models, where buses depart only when full, leading to potential schedule instability. In contrast, DRC services with fixed headways could substantially improve passenger experiences by incorporating the demand stochasticity and its complex impact on service design. This paper introduces analytical models to optimize the DRC design with fixed headways, examining two operational strategies: (i) the “fully-flexible routing” strategy, where a vehicle serves only the requests received before its dispatch time, and (ii) the “semi-flexible routing” strategy, where a vehicle traverses the zone by traveling longitudinally along the swath to serve the requests received on the route. We determine the optimal service headway, zoning, and vehicle capacity for each strategy and benchmark these against conventional fixed-route feeder services. Our numerical experiments reveal that our models for fully-flexible routing and semi-flexible routing are approximately 10% and 6% more accurate, respectively, than those from previous studies. Additionally, critical demand densities are identified between DRC services and fixed-route strategies.

ABSTRACT. The increasing prevalence of vehicles equipped with onboard sensors, ranging from conventional vehicles with advanced sensor capabilities to autonomous vehicles (AVs), offers significant potential for traffic state estimation. AVs provide substantial data from their surroundings, but their network penetration rate is expected to remain low in the near future. This paper introduces a lane-based queue length estimation method for city streets based on AV data, as existing works in the literature often overlooked multi-lane streets. In this study, AV data refers to the information autonomous vehicles collect as they move through the network. This includes detailed data about the AV's own status at each time step, as well as information about surrounding vehicles within a specific detection radius. Key data inputs include vehicles' position, speed, and ID. AVs can detect a specific vehicle at multiple locations and offer a continuous and real-time view of traffic conditions by capturing real-time data of surrounding vehicles. AVs close to intersections, especially those in the queues, can gather data of vehicles entering the intersection from all lanes. The data can be used to reconstruct missing data in vehicle trajectories. Aggregating data from multiple AVs across the network provides opportunities to estimate traffic states even with a low percentage of AVs in the network and enables a comprehensive understanding of queueing dynamics at intersections. The proposed approach assumes no explicit information regarding signal timings or vehicle arrival distributions. To account for the ultra-low penetration rate of AVs, we developed methods suitable for varying AV penetration rates of 1%, 2%, and 5% in the network. We propose a method for estimating traffic signal states, reconstructing vehicle trajectories and estimating queue length within urban networks. Since an existing dataset with a percentage of AVs in the network is unavailable, we simulated data using AIMSUN software, modelling a network with 16 intersections (4 actuated and 12 fixed-time signal intersections).

ABSTRACT. Understanding the relationships between network design and traffic management on the one hand and the quality of traffic flow on the other is an essential prerequisite for the development of networks and the measures that control them. The Macroscopic Fundamental Diagram (MFD) offers a way of analysing and evaluating urban transport networks on a large scale with regard to their performance and the associated external effects. However, estimating a unique MFD from empirical data given the dynamic traffic states is far from trivial. Since traffic demand’s major destinations vary in the course of the day and over the week the saturation of the network’s links is uneven even for equilibrium conditions. Consequently, the network performance at a certain period of time is characterized by the sample of links with the highest saturation. To avoid biased results in the estimation of the MFD only links on routes with the highest saturation at that time of the day should be selected. This assumes equilibrium in a partial network with directed demand and thus homogeneous conditions. To implement this consideration for MFD estimation, the saturation of the links and therefore their capacities are required. Recent research results show how the capacities and thus the saturation rates of links can be derived from the accumulated trajectories of floating car data. The saturation rates are then used to select representative samples of links for MFD estimation. First tests of the developed methodology with data from the Zurich and Braunschweig road network show the sensitivity of the MFD estimate to these factors and the plausibility of integrating them.

ABSTRACT. Recently, advances in Information and Communications Technology (ICT) have led to the development and operation of various types of Demand Responsive Transport (DRT) in rural and urban areas. These DRT services have been confined to passengers residing in low-demand areas, commuters during peak hours, and individuals utilizing specific stops or transfer hubs to reach their destinations. Additionally, because routes are planned based on information collected through in-advance reservations or subscriptions, route changes are not possible or are significantly limited during DRT service. This study aims to develop a DRT service system with fully flexible routes, free from restrictions on specific passengers and fixed stops or destinations. DRT systems employ an Insertion Heuristic (IH) approach to solving scheduling problems, utilizing an objective function to minimize travel time costs while considering constraints such as DRT capacity and the passenger's desired time for pick-up and drop-off. A case study of the Sioux Falls network environment with 80 requests and 4 DRTs confirmed that the proposed DRT system efficiently shares trips for passengers with similar origins and destinations. In addition, seven indicators from analytical, operational, and passenger perspectives were introduced to evaluate the DRT system, confirming the efficiency and scalability of the proposed system through indicator analysis. The proposed DRT system can be expected to provide services based on actual demand in a real-world network through several verifications in the future.

ABSTRACT. Current network coordination control in practice is mainly realized by control subarea partitioning, which fails to describe the intrinsic property of urbanite trip reflected by vehicle paths, while some path-based network coordination control methods depend strongly on given critical paths derived from subjective experience. In this paper, a path-based coordination control optimization model is proposed under a multi-agent control modelling framework. Aiming at minimizing the total delay, network coordination control is realized through agent self-optimization and an optimization iterator describing the interaction between neighboring agents. Evaluation was done through a simulation case with satisfactory results outperforming existing coordination signal control models by over 25%.

ABSTRACT. This paper introduces an innovative approach to model the lane-change (LC) process of vehicles by employing parallel learning, seamlessly integrating conventional physical or behavioral models with data-driven counterparts. The LC process is divided into two distinct steps: the LC decision and the LC implementation, each independently modeled. For the LC decision model, a utility-based model is embedded into a neural network. Simultaneously, the LC implementation model incorporates a conventional car-following model, replicating the behavior of the new follower of the lane-changer, within the training process of a long-short-term memory model. Empirical trajectory data collected from unmanned aerial vehicles, which provides detailed information on the vehicles' lane-changing process, serves as the basis for training and testing the proposed models. Additionally, data from a different site is employed to assess model transferability. Results demonstrate that the proposed models adeptly predict both LC decisions and implementations, outperforming baseline physical and behavioral models, as well as pure data-driven models, in terms of prediction accuracy. These findings highlight the significant potential of these models in improving the precision of microscopic traffic simulators.

ABSTRACT. Over the last decade, automation and connectivity have increasingly integrated into the transportation market. This paper builds on established problem formulations for Connected and Automated Vehicles (CAVs) to evaluate their performance at a signal-free intersection in Heraklion, Crete, Greece. Using the simulation tool PTV Vissim, we constructed a detailed model of the intersection to demonstrate our approach. Our analysis compares travel times and stop-and-go events in scenarios with only human-driven vehicles (HDVs) versus those exclusively managed by CAVs. The results indicate that CAVs significantly enhance traffic flow efficiency and reduce traffic congestion, demonstrating the potential of CAVs to improve urban mobility.

ABSTRACT. As an important part of the mobile sensing system, the vehicle sensing system has great potential to collect large quantities of spatio-temporal sensing information with low operation costs by the pre-installed mobile sensors. Sensing systems usually need to sample data with different distributions to obtain sufficient information for different usages. The crucial problem is how to predict the vehicle positions and allocate vehicles to match targeted sensing distributions. A conventional method for the problem is a 2-stage predict-then-optimize method. However, training the prediction model based on prediction error can lead to inferior decision-making compared with directly minimizing the decision error. To this end, we propose an end-to-end smart predict-then-optimize (SPO) method by integrating optimization into prediction within the deep learning architecture. We also develop an alternating differentiation method to compute the gradients of the optimization layer. This SPO framework uses the task-specific loss function of the eventual effective decision rather than the function of prediction loss. We evaluate the effectiveness and robustness of the proposed framework by conducting various experiments using real taxi datasets in Hong Kong with different scale networks.

ABSTRACT. Please refer to the attached file, which is the extended abstract itself

ABSTRACT. Walking offers significant benefits for society, public health, and sustainability. Recognizing these advantages, this research introduces the Spatially Transferable Pedestrian Demand and Network (STePNet) Model, designed to enhance urban planning for pedestrian-friendly environments. Traditional transportation models have often neglected walking, but STePNet addresses this gap by integrating advanced pedestrian demand models and examining their effectiveness across major Australian cities—Sydney, Melbourne, and Brisbane. The model evaluates how changes in local conditions can influence walking behaviors and develops a framework to support more walkable communities. The goal is to improve the adaptability of pedestrian models across different locations and integrate them into comprehensive urban development strategies to encourage walking.

ABSTRACT. Ride-sourcing platforms like Uber, Lyft, Didi, and Grab offer both ride-pooling (shared rides) and non-pooling (solo rides) services. Recently, some platforms introduced a bundled service option for passengers indifferent between ride-pooling and non-pooling if they could experience less waiting time. The bundled option assigns passengers to either service, affording the platform more flexibility in supply and demand management. When vehicle supply is low, the platform assigns more passengers to ride-pooling to improve service rate and reduce total occupied vehicle hours. Conversely, when supply is adequate, the platform assigns more passengers to non-pooling to avoid detours and waiting times, increase revenue, and improve service quality. We propose a model to optimize demand and supply management in a ride-sourcing market with multiple options, including ride-pooling, non-pooling, and bundled service. We investigate the optimal supply allocation strategy, considering the proportion of total vehicles assigned to ride-pooling or non-pooling, and the intrinsic relationship between supply allocation and the fraction of demand for bundled service assigned to ride-pooling. Our analysis shows that demand for bundled service can be determined endogenously according to the supply allocation. We identify critical thresholds over/under which the optimal supply allocation strategy is to assign all passengers opting for bundled option to ride-pooling/solo ride services, respectively. The optimal allocation depends on the level of supply, level of demand, inconvenience cost, charging price discount for pooling rides, etc. As more passengers choose bundled service, the platform should allocate a larger proportion of these passengers to ride-pooling. Our analysis offers managerial insights to help platforms design the supply allocation strategy for the emerging bundled service, improving platform profit and/or service quality.

ABSTRACT. With the advent of Modular Autonomous Vehicles (MAVs), transit agencies can adapt to fluctuating demand by flexibly coupling several MAVs into a Transit Unit (TU) per dispatch. We further envision the benefit of decoupling TU into MAVs in On-Demand Feeder Transit (ODFT), which connects a transportation hub and offers door-to-door services to patrons spread over a distant region. The benefit would come from the reduced routing distance/time to visit N points within the service region, traditionally accomplished by a single bus vehicle but now by several MAVs, each covering a subset of N points. However, whether recoupling the MAVs into a TU on return trips needs careful investigation since additional delays may be caused by waiting for the last-arrival MAV.

To demonstrate the concept’s effectiveness, we propose an optimal design model to determine the key operational features for MAV-based ODFT, such as TU sizes at dispatch, dispatch headways, and zone partitions, which can vary spatially to suit non-uniform demand distributions. The approach of continuum approximation is used to derive analytical expressions of system metrics including patrons’ routing time, waiting time, and non-stop line-haul travel time, as well as the agency’s operational costs. Closed-form relationships are obtained for the optimal conditions, leading to an efficient solution algorithm. Numerical studies show that MAV-based ODFT consistently outperforms traditional bus-based ODFT (which can be seen as a special case with no coupling/decoupling functions), with generalized system cost savings exceeding 3%. Notably, the advantage of MAV-based ODFT over its counterpart diminishes when the routing distance/time variance is large. This underscores the necessity of advanced operational algorithms to minimize MAV trips’ variance and leverage the flexibility of MAVs in practice.

ABSTRACT. Connected and Automated Vehicles (CAV) must undergo a series of rigorous tests before being authorised to operate on open roads. However, due to the high number of possible scenarios, conducting each test becomes economically unfeasible. To address this issue, scenario-based approaches [1] are being implemented to significantly reduce the number of tests required for CAV. These tests are divided between costly but realistic field trials versus easier-to-implement and cost-effective simulations. However, simulation tools provide a simplified view of reality and focus on a subset of attributes during the calibration stage. Therefore, assessing the reliability of the simulation tool per scenario is required, before admitting their outputs as acceptable. Based on expert knowledge, this paper aims to evaluate the degree of confidence that can be placed in simulation tools using Multi-Criteria Decision-Making methods (MCDM). Our paper copes with the problem of assigning labels (i.e. Confidence Levels) to any traffic scenario according to the perceived trust in the capacity of microscopic simulation tools to mimic the real world. We conducted a survey to feed a multi-criteria decision-making tool, namely the Analytic Hierarchy Process (AHP) [2], using in some studies to determine the complexity of a scenario [3].

ABSTRACT. Lane closures on highways, caused by accidents or road works, often result in lane-drop bottlenecks and consequent traffic congestion. With the development of connected vehicles (CVs) and connected automated vehicles (CAVs), the concept has been proposed that limited CVs/CAVs are used to harmonize vehicle speeds and regulate the arrival patterns at lane-drop bottlenecks on freeways under saturated traffic. However, existing studies primarily focus on formulating analytical models of controlling longitudinal speed profiles of CVs/CAVs assuming no lane changes. Further, the interactive impacts of multiple CVs/CAVs on traffic flow are overlooked due to the difficulty in analytical modeling. Inspired by the successful application of deep reinforcement learning (DRL) in vehicle trajectory planning, this study proposes a DRL-based approach to regulate the arrival patterns of saturated traffic flow at a lane-drop bottleneck on freeways by trajectory planning for limited vehicles. Considering the higher penetration rates of CVs than CAVs in the near future, this study investigates the longitudinal and lateral trajectory planning for limited CVs with discrete speeds and lane choices to improve safety, efficiency, and fuel economy of traffic flow traveling through the bottleneck. The conservative Q-learning algorithm is used for offline pretraining with the exiD dataset. Empirical policies learnt from real-world human driving data speed up the convergence of online training with the Deep Q-Network algorithm. The fully-cooperative setting for multi-CV control is employed in the centralized-training-decentralized-execution (CTDE) way to extend the single-agent DRL framework to the multi-agent one, which enables CVs to learn collaboratively. Simulation results validate the advantages of the proposed models, as well as the benefits of the pretraining and the CTDE method.

ABSTRACT. The present paper explores the management of platoons of connected automated vehicles (CAVs), focusing on their impact on multilane mixed traffic flows, where CAVs share the road with regular vehicles (RVs). Even when it is anticipated that introducing platoons could increase maximum traffic throughputs by reducing the average vehicular headway, adequate traffic management strategies are to be proposed and assessed in order to make the most of CAVs platooning. This involves analyzing if platooning lanes need to be dedicated or not (or when), how many and which lanes should allow CAV platooning, or if the platoon length must be limited, among others. With this objective, the paper adopts a simple platooning algorithm for multilane highways, and implements it in the Aimsun Next microsimulation software. Using a three-lane ring road as the simulation layout, different scenarios are run to assess the effect of platooning on mixed traffic flows, and to test the adequacy of the management strategies proposed. First results are promising showing significant increases in traffic throughput. For example, at a 50% penetration rate of CAVs and with a single platooning lane, the capacity of the infrastructure more than doubles compared to that in baseline conditions. Next steps comprise the introduction of other strategies such as the platooning dynamic management or the limitation in the platoon lengths.

ABSTRACT. Real-time crash risk estimation at urban signalized intersections is crucial for reducing accident rates and enhancing road safety. Traditional methods, such as Extreme Value Theory (EVT) models using Block Maxima (BM) or Peak Over Threshold (POT), often struggle to capture the complexities inherent in urban traffic dynamics. To address this limitation, we propose an advanced EVT model that incorporates learning-based anomaly detection to improve crash risk prediction at signalized intersections. The study introduces a novel approach where machine learning (ML) and deep learning (DL) techniques are utilized to identify extreme values. For EVT models, the Generalized Extreme Value distribution is applied to BM samples, while the Generalized Pareto distribution is used for POT samples. We tested six learning-based unsupervised anomaly detection algorithms to extract extreme values, leveraging metrics such as Time to Collision (TTC), Modified Time to Collision (MTTC), and Deceleration Rate to Avoid a Crash (DRAC).Data for the study was collected from 6 AM to 6 PM on weekdays at three intersections in Daejeon, South Korea. Our results indicate that DL methods outperform traditional and ML methods in estimating crash risk, with autoencoder and Long Short-Term Memory (LSTM) networks exhibiting the closest correlation to observed crash frequencies. The proposed integration of EVT with advanced learning-based approaches demonstrates improved precision and relevance compared to traditional methods. Real-time crash risks are estimated by correlating the refined EVT outputs with historical crash data, resulting in two key safety indices: Risk of Crash (RC) and Return Level of a Cycle (RLC). These indices offer a comprehensive temporal profile of crash risks at urban signalized intersections, providing actionable insights for traffic safety improvements.

ABSTRACT. The increasing number of accidents that occur at signal-controlled intersections in South Korea imposes a need to prioritize driver behavior approaching intersections. Especially during yellow traffic lights, many drivers' uncertain behavior may cause them slip into dilemma zone. Although connected vehicle technology can provide drivers with safety guidance in such conditions, limited research has fully explored driver behavior in real communication environments under the framework of the promising CV2-X technology. Also, it rare that difference in driving performance between different guidance information styles are investigated. To fill the gaps, a digital-twin platform combining Unity and VISSIM was built to allow a more realistic simulator experiment scenario. A simulator experiment based on this platform was conducted with 80 drivers in 5 communication environments (no communication, countdown-based perfect communication, stop or go based perfect communication, communication interruption, and communication loss). 6 yellow light signals at varying times from the stop line were set up in one trial to capturing the driver’s dilemma zone experience approaching an intersection. Several driving performance indicators during the yellow light period are extracted, including the mean and standard deviation of speed, acceleration, heading error. Specifically, the repeated measures ANOVA is utilized to evaluate the difference in driving performance between various communication environments. Furthermore, the group random parameters regression model is be applied to measure the association between the performance indicators and factors that include communication environments, time to stop line, mental workload, personal characteristic factors. The findings are expected to ensure the new CV2X technology’s evolution can meet future demand as well as better improve road safety and driving decision-making efficiency.

ABSTRACT. Bike-sharing systems are becoming popular as a standalone mode of transportation or a way to access other modes of transportation, particularly for short distances. These systems usually start as small-scale pilot projects and expand gradually depending on user demand. However, determining the optimal location of bike-share stations can be challenging. It involves considering various decision variables, such as station location across the urban landscape and their capacity, which can impact the network's efficiency and functionality. Bike-share station placement is not only about accessibility but also requires analyzing bike flow, origin-destination (OD) trip data, and mobility patterns. This study introduces a continuum approximation (CA) model aimed at optimizing the placement of bike-share stations to maximize ridership. The algorithm adjusts each station's location within a continuous service area, where demand is spatially distributed, to align with OD patterns and maximize network efficiency. Unlike previous methods focused on accessibility, this methodology prioritizes ridership by strategically positioning stations to attract users and facilitate efficient trips. Numerical experiments demonstrate the effectiveness of the CA algorithm, delivering near-optimal solutions that enhance bike-sharing ridership.

ABSTRACT. A Travel-time and Destination Distribution (TDD) Model is developed to predict passengers' TDD one-day ahead. With this TDD prediction and the real-time passengers of an originating station, the passenger flows to other stations and the associated travel duration can be estimated with high accuracy. The model is validated with real subway passenger data and outperforms other well-known models in the literature.

ABSTRACT. Obtaining street-level annual average daily traffic (AADT) by vehicle type is necessary for various applications but still challenging, especially across large road networks. Despite recent advances in adapting machine learning (ML) algorithms to estimate AADT, the specific challenges of applying ML in spatial settings are typically overlooked. Our study proposes a methodology that combines ML, spatial statistics, and extensive geospatial data to enhance AADT estimation and the assessment of spatial predictive ML models. Our approach uses a lightGBM model to estimate AADT at over 19,000 locations in England and Wales, incorporating over 900 spatial features with additional variables to account for spatial autocorrelation. The Boruta algorithm is applied to remove redundant features, proving effective in enhancing model performance. Unlike traditional methods, we evaluate model performance using a cross-validation process designated for spatial models, consequently increasing the reliability of model assessment. The AADT estimates are further split by vehicle and fuel type, thereby supporting pollution and carbon emissions estimation and offering insights for sustainable development.

ABSTRACT. Creating a strategy for effectively and safely navigating traffic in the lane-free paradigm is a challenging task. With the aim of efficiently verifying and demonstrating the TrafficFluid concept, a preliminary ad-hoc strategy was formulated for the movement of vehicles on lane-free roads. Extensive research has been conducted and documented on this particular model, encompassing its capability for safe movement in lane-free road scenarios , calibration of macroscopic models that reproduce its macroscopic effects and application of internal boundary control. These investigations revealed that the proposed moving strategy still entails the conventional traffic issue of capacity drop, caused by congestion in merging areas, as confirmed by the flow results from microscopic simulations. A commonly suggested solution to tackle this problem is the implementation of ramp metering control measures. This study explores a Variable Speed Limits (VSL) approach by implementing ramp metering in a lane-free context, providing novel aspects on traffic management.

ABSTRACT. In urban logistics, Unmanned Aerial Vehicles (UAVs) and sidewalk robots present promising alternatives for user and eco-friendly delivery solutions. Addressing the complexities of delivery in urban areas, namely, safely traversing constrained high-density areas and uncertainties such as traffic congestion, this study introduces a comprehensive multi-modal autonomous delivery and network design. This framework simulates a fleet of drones and robots through a novel cooperative bi-modal, three-dimensional centralized dispatch system to determine optimal delivery modes and routes to enhance delivery efficiency and minimize delays. The network design approach is the development of an adaptive, structured airspace network. This network extends the conventional two-dimensional road infrastructure into the vertical dimension, facilitating aerial deliveries while mitigating collision risks through urban architectural considerations. Our methodology leverages deep multi-agent reinforcement learning augmented by a dynamic transformer architecture using a heterogeneous edge-enhanced attention model. This innovation allows for real-time optimization of delivery assignments, given the stochastic nature of urban delivery challenges. By embedding and updating graph states, vehicle-customer pair selection and delivery sequences are optimized through policy gradient methods, achieving optimal dispatching based on time-sensitive demand. The performance of the proposed model is evaluated based on greedy and sampling simulation of the experiment with rollout baselines on various scenarios, validating by real-world instances from the region of Mississauga. Additionally, sensitivity analyses across varying demands and fleet sizes evaluate the model's scalability and generalize the problem through different instances. This research not only advances the operational efficiency of last-mile delivery systems but also provides significant improvements in sustainability and service quality by designing an adaptive and scalable framework for autonomous delivery.

ABSTRACT. This paper explores the disparity between traditional household travel surveys (HTS) and mobile phone data in capturing mobility patterns. While HTS offer detailed information, they are costly and infrequent. In contrast, mobile phone data are abundant but may lack completeness and raise concerns about representativeness. The paper aims to understand the characteristics of trips identified by mobile phone data compared to HTS, focusing on transport mode, duration, distance, and purpose. Additionally, it examines whether mobile phone data and HTS, under the same trip definition, yield comparable volumes and if mobile phone-derived home-work OD matrices align with those from public statistics. The case study presented in this paper deals with the Lyon region of France and compare an HTS from 2015 and mobile phone data from 2022. The paper also discusses the findings and future research directions.

ABSTRACT. This research focuses on developing and validating a Preference Questionnaire for Shared Autonomous Vehicles (SAV) in Commuting (PQSC) tailored for urban scenarios. Utilizing a comprehensive literature review, 24 questions were formulated to assess users' subjective attitudes towards SAV. The questionnaire was administered to 818 commuters in Hangzhou, resulting in the construction of a second-order measurement model that captures seven key factors influencing SAV preferences, such as functionality, safety, and shareability. The application of exploratory and confirmatory factor analyses substantiated the questionnaire's structure and reliability, offering a robust tool for gauging public sentiment towards SAV. This provides a potential research and measurement tool for understanding the attitudes of the target population towards SAV. It contributes to further improving SAV services, particularly for commuters who face challenges with traditional forms of transportation.

ABSTRACT. Travel behavior prediction is a fundamental task in transportation demand management. The conventional methods for travel behavior prediction rely on numerical data to construct mathematical models and calibrate model parameters to represent human preferences. Recent advancement in large language models (LLMs) has shown great reasoning abilities to solve complex problems. In this study, we propose a framework to use LLMs to predict travel behavior with prompt engineering without data-based parameter learning. Specifically, we carefully design our prompts that include 1) task description, 2) travel characteristics, 3) individual attributes, and 4) guides of thinking with domain knowledge, and ask the LLMs to predict an individual’s travel behavior and explain the results. We evaluated the framework through two real world case studies, travel mode choice and trip purpose predictions. Results show that, though no training samples are provided, LLM-based predictions have competitive accuracy and F1-score as canonical supervised learning methods such as multinomial logit, random forest, and neural networks. This demonstrates that the proposed framework’s adaptability to new scenarios and effectiveness in cold start situations. LLMs can also provide explanations for their results, which helps users understand them, despite occasional logical inconsistencies or hallucinations.

ABSTRACT. The advancement in traffic data collection technologies, including loop detectors and GPS devices, has significantly transformed traffic management systems. These technologies enable the generation of vast data volumes, catalyzing the development of sophisticated machine learning (ML) models tailored for real-time Traffic Incident Prediction (TIP). Traditionally, these models have focused on ’link-level’ incident prediction, assessing the risk on specific road segments by analyzing data from upstream and downstream links. Although valuable, such models often overlook broader traffic dynamics that influence incident probabilities on a larger scale. In contrast, ’network-wide’ incident prediction models provide a holistic view of risks across a city’s traffic network, supporting effective incident management and resource allocation for sub-areas (Tran et al., 2023). However, these models face challenges in handling large data volumes and computational demands, which may limit their predictive accuracy and detail in managing incidents on specific road segments or smaller regions. Recent research has increasingly leveraged ML, especially deep learning and Graph Neural Networks (GNNs), to handle the complex patterns in traffic data effectively. GNNs align with the graph structures of road networks, as demonstrated by Deep Spatio-Temporal Graph Convolutional Network for Traffic Accident Prediction (DST- GCN) for links (Yu et al., 2021). Further, ML researchers (Wang et al., 2021a,b) have employed a grid-based approach to segment traffic networks into ’cells’, each representing a specific area, attempting to predict regional risk for these cells. However, these models adopt a grid or image representation of the city network, which does not accurately reflect the real-world interconnected road structures, thereby limiting their practical applicability. Moreover, while the integration of imagery and numeric data has been demonstrated in traffic demand prediction (Wang et al., 2024), it remains largely unexplored for incident prediction. This integration could leverage diverse data sources, such as HERE or Google Congestion Maps, providing broader views on traffic network structures and conditions. Therefore, there is a critical need for models that not only utilize the complementary aspects of imagery (e.g., link congestion heatmap images) and numeric data (e.g., loop detector data) but also predict at a more granular level, such as individual links or sub-areas, rather than abstract grid cells. Such models would provide clearer, more precise predictions for traffic operators, closely aligning with the actual configuration of transportation networks. In response to existing limitations, our research introduces Multi-task Multi-view Neural Networks (M2NN), a novel approach that diverges from grid-based predictions and traditional models reliant solely on numeric data to enhance model applicability across real-world networks. M2NN achieves fine-grained predictions for links and allows for flexible application across diverse sub-areas in the large network. We innovatively integrate sub-area incident prediction as a sub-task within our multi-task learning framework. This design enables M2NN to simultaneously learn and predict incident risks at both the link and sub-area levels, effectively capturing localized patterns and providing a comprehensive risk overview by leveraging complementary link-level and sub-area-level data. Also, M2NN incorporates congestion heatmap imagery of links alongside numeric data, such as loop detector data and information vectors—a synergy not extensively explored in current research. This multi-view approach enriches the encoding process, enhancing our model’s predictive performance. Preliminary results with diverse real-world data sets demonstrate M2NN’s enhanced fine-grained predictive capability, aligning the complementary use of imagery and numeric data in improving traffic demand analysis (Wang et al., 2024) and expanding the use of map images (e.g., satellite images) for TIP models.

ABSTRACT. The emergence of shared mobility services, such as demand-responsive transportation (DRT) systems and ride-matching apps, has revolutionized urban transportation by offering flexible and efficient modes tailored to changing travel demands. DRT systems, leveraging advanced technology, optimize routes based on real-time demand, enhancing accessibility particularly in areas with limited public transportation. Concurrently, ride-matching apps facilitate convenient and sustainable ride-sharing, integrating traditional carpooling into a digital platform. This study investigates the behavioral intentions of young individuals towards DRT services and ride-matching apps integrated with carpooling. Focusing on Nicosia, Cyprus, the target group of the survey included young and highly educated people that can be deemed as the early adopters. Results indicate that socio-demographic factors like gender and household size, as well as location-based characteristics such as proximity to the university, influence DRT usage. Additionally, preferences for incentives like restaurant credits impact adoption. Similarly, the study identifies factors affecting the intention to use ride-matching apps as drivers or passengers, including gender, employment status, and perceptions of cost and reliability. Understanding these dynamics sheds light on the interplay between shared mobility services, informing strategies to promote sustainable urban transportation and mitigate reliance on private vehicles.

ABSTRACT. Population Synthesis is a crucial part of the Activity-based model (ABM). The goal of Population Synthesis is to create a realistic population and the key to it is diversity. Most of the research work till now only focuses on 2 types of diversity: spatial and households-persons connection. Another significant one that requires more attention, i.e. a methodology that exclusively includes it, is relationships within households. This paper introduces 2 new methods and a framework to combine them: Sequential Attributes Adjustment (SAA), Chained Sample Pools (CSP), and Integrated Population Synthesis Framework (IPSF). SAA aims to improve spatial diversity, while CSP will tackle the other two. Therefore, the combination of them (i.e., IPSF) will be the first to balance all 3 types of diversity, especially relationships within households. Furthermore, IPSF is designed to be modular; hence, it can be easily customised and combined with other methods. This paves the way for further research and better adaptation. We conducted a full-scale case study of Victoria, Australia across 691 Postcodes (POAs) to validate the methodologies against Iterative Proportional Updating (IPU) and Bayesian Network (BN). Specifically, as inputs, we used the Census data from Australian Bureau of Statistics (ABS) and the Victorian Integrated Survey of Travel and Activity (VISTA). The results and analysis found IPSF to be the best method. It can learn the relationships in a household well and maintain the distributions of the sample data while exceeding IPU and BN in matching the household marginal data (decreasing around 74% and 99%, respectively). Thus, IPSF is a reliable, robust, and flexible framework to generate a detailed synthetic population with the required socioeconomic attributes.

ABSTRACT. Electric Vehicles (EVs) are widely recognized as a promising solution for reducing emissions and saving energy. There has been a fast-growing adoption of EVs in recent years, mobilizing the transportation landscape towards a more environmentally friendly future (Ferrero et al., 2016). However, despite considerable technological and market progress, charging EVs is still less convenient than conventional refueling vehicles. The abovementioned issue calls for an integrated and efficient EV charging system to cater to massive and diverse charging demands in the future electric mobility system. Based on charging location flexibility, existing recharging modes for EVs can generally be classified into fixed charging and mobile charging. This study considers a mobile charging crowdsourcing online platform that utilizes crowdsourced vehicles to provide mobile charging services for EV users. The primary objective of this study is to determine the optimal operation strategies, such as the matching between mobile charging demand and supply, repositioning of the mobile chargers, so as to maximize the profit gained by the operation of crowdsourced mobile charging service.

A platform is operating the mobile charging service in this area, and the operation is a dynamic process with certain level of stochasticity. At any time, there could be a newly generated recharging order at certain potential customer location, with an associated latest allowable service completing time (indicating the expected time that the vehicle leave the parking spot) and the associated charging amount. Each order is with a recharging earning for the platform that is directly related to the charging amount. If a vehicle does not get a complete recharge before its latest allowable time, the order is canceled, and there will be a penalty associated with the platform operation. On the other hand, the platform is operating a flexible fleet of crowdsourced mobile chargers, such that there could be mobile chargers joining or leaving the fleet at any time. Each mobile charger is with a maximum battery level and a safety battery level; if the battery level is below the safety threshold after serving one order, the charger needs to go to one recharging facility. We ignore the energy consumption associated with traveling of mobile chargers. The goal of this study is to design a methodological framework for operating the mobile charger fleet, including routing, recharging, and repositioning among zones.

ABSTRACT. This paper introduces a Discrete Choice-Based Optimization Approach to address the Dynamic Large-Scale Multi-Compartment Vehicle Routing Problem (DVRP) within Reverse Logistics. The study highlights the integration of Internet of Things (IoT) technologies with advanced routing algorithms to enhance urban waste collection. By utilizing real-time data from IoT-enabled waste bins, the proposed system dynamically adjusts routes, thus improving operational efficiency and reducing costs associated with conventional waste management methods. The use of multi-compartment vehicles not only increases route efficiency but also promotes environmental sustainability by reducing the frequency of vehicle dispatches and optimizing load segmentation. Our methodology incorporates a novel application of the Discrete Choice Model (DCM) to predict the most likely next subzones for waste collection, greatly enhancing the responsiveness and adaptability of urban waste management systems to real-time conditions. Additionally, the paper discusses the computational challenges of such large-scale systems, recommending the use of Approximate Dynamic Programming (ADP) to overcome the "curse of dimensionality." Simulation results demonstrate that the proposed approach significantly outperforms traditional methods in terms of solution quality and computational efficiency. This study represents a significant advancement in the fields of smart cities and sustainable urban logistics, underscoring the practical benefits and potential of integrating advanced technology and optimization techniques in waste management.

ABSTRACT. The utilization of high-resolution satellite imagery is an increasingly prominent trend in transportation science, offering the potential to enhance existing dynamic network models and address the limitations of traditional data sources. Satellite images provide comprehensive coverage and detailed insights into road networks, traffic patterns, and infrastructure at a scale previously that was previously unachievable with data collected from local sensors. A primary challenge lies in developing a robust calibration framework (i.e., dynamic origin-destination demand estimation, DODE) for large networks that effectively integrates traffic state information (i.e., traffic density) obtained from satellite imagery with other data sources. This study proposes a computational-graph-based DODE framework leveraging multi-source data including traffic densities derived from satellite images and explores the benefits of incorporating satellite imagery into model calibration through two numerical experiments using both synthetic and real-world data.

ABSTRACT. Although the boarding processes of conventional tube-and-wing aircraft have been extensively investigated, the impact of boarding strategies and passenger space expansion on future blended-wing body aircraft remains relatively unexplored. In this study, a blended wing body aircraft boarding simulation model was used to evaluate the effectiveness of novel boarding strategies designed for the blended wing body aircraft. Multiple aisles in the BWB cabin enable novel boarding strategies and faster boarding in general compared with conventional tube-and-wing aircraft. However, unlike conventional tube-and-wing aircraft, the performances of efficient boarding strategies are limited by the boarding rate.

ABSTRACT. The study introduces a real-time method for optimizing connected and automated vehicle (CAV) trajectories at signal-free intersections, employing a greedy-based heuristic embedded within a receding horizon framework. This approach constructs multiple trajectories for each CAV, selecting the optimal one based on objective values, significantly reducing computational demands and achieving solution times in milliseconds. By utilizing a platooning logic, it effectively manages vehicle platoons, enhancing traffic flow and minimizing delays. Experiments at a simulated four-legged intersection with varying traffic demands demonstrate the method's efficiency, consistently operating under 35 milliseconds and maintaining low delay times even at high traffic volumes. This methodology offers a promising alternative to traditional centralized optimization, providing a scalable and efficient solution for autonomous intersection management.

ABSTRACT. The need for additional capacity in motorway networks during periods of high demand is unavoidable if congestion is to be prevented. Increasing capacity by building new roads is often infeasible, leaving operation-based traffic control measures as the primary approach to exploit the existing infrastructure. The lane width at motorways is much larger than the vehicles' width to consider safety during high-speed driving. This leads to the loss of lateral capacity of the existing traffic infrastructure. In this paper, the novel concept of dynamic lane configuration is introduced, which opens a new avenue in motorway traffic control. Dynamic lane configuration suggests that while current wide lanes ensure safety during high-speed driving, lower speed limits can be actively imposed during times of high traffic demand, allowing the lane width to be reduced thanks to the reduced required lateral gap between vehicles. By narrowing the lanes prior to congestion, it is possible to reclaim wasted space and add lanes to the road, leading to a dynamic capacity increase during the operation. This dynamic infrastructure layout with demand-responsive lane configuration during operation bridges the traffic management and infrastructure design. A CTM-based methodology, together with an optimal control approach, is developed to model the dynamic lane configuration and to define the time and location of changing lane configuration. The promising simulation results indicate the potential of the proposed approach in congestion mitigation and reducing travel time.

ABSTRACT. Satellite imagery has become important for the analysis of the dynamics in urban transportation. Asymmetry of the travel behavior among transportation modes are particularly important for the evaluation of urban spaces, in which multiple transportations interact each other through endogenous effects. Existing works using adversarial learning methodology estimate and simulate the route choice behavior under multi-transportation condition, but the interaction itself is not revealed, which make it difficult to understand the dynamics of the equilibrium state. In this research, a method to estimate the interaction among transportations is proposed from the view of the stability of the equilibrium introducing Lipschitz normalization into the endogenous effect. In addition, the concept of asymmetry among transportations is introduced. By combining the satellite imagery with GPS data through the route choice model, the spatial features which have strong effect to the asymmetry are analyzed.

ABSTRACT. This paper presents an exhaustive methodology aimed at modelling and calibrating Stochastic Fundamental Diagram (S-FD) for highways on three different perspectives: with respect to vehicle classes, to data source classes and to the calibration method. The former leads to the calibration of flow and density vehicle equivalence coefficients, required to consistently apply the S-FD. The second perspective allows to define a process for calibrating the speed-flow function and speed distribution based on observed trajectories and extending it to highway sections where only portal data are available. Finally, the results provided by different calibration methods are compared based on different performance indicators. The effectiveness of the proposed methodology is confirmed through the application of a real case study. The proposed methodology is suitable for project assessment and evaluation, offering the capability to consider various externalities; moreover, it is useful in transportation planning. In terms of highway traffic control, the S-FD model can be exploited to predict the impacts of speed variations, such as those arising from construction zones, as well as the effects of speed control strategies like Variable Speed Limits.

ABSTRACT. With the rise of innovative mobility options and the combination of new and traditional transportation modes, which enable door-to-door multimodal itineraries, it is important to understand how transportation users plan their itineraries. This study presents theoretical concepts and empirical evidence on the formation of an ascending-descending multimodal itinerary based on a modal hierarchy structure: A passenger begins their journey with a lower hierarchical mode of transportation, such as walking, then transfers to higher modes, such as the metro, and concludes their trip with a lower mode again. We define two types of integrative factors to define modal hierarchy: mobility efficiency, which represents the time and monetary efficiency of in-vehicle trips, and access sparsity, which reflects the accessibility of out-of-vehicle experiences. We hypothesize that a modal hierarchy exists in terms of mobility efficiency and access sparsity of each mode and that users plan their multimodal itineraries in an ascending-descending order based on this hierarchy. Using multimodal trip revealed preference data collected in Bucheon City, Korea, the results reveal a clear modal hierarchy in mobility efficiency and spatial access sparsity, except for temporal access sparsity. It also shows that users' itineraries follow this ascending-descending order with statistical confidence. We believe this approach not only explains how users plan their multimodal itineraries but also has broad applicability in urban multi-modal mobility planning and operations.

ABSTRACT. Efficient estimation of the origin-destination (OD) matrix, defined as the travel demand between network origin and destination nodes, is pivotal for effective traffic monitoring, planning, and management. Over the years, the OD matrix estimation problem has received considerable attention, leading to the development and testing of various approaches utilising traffic counts from fixed-location sensors. In this work, we introduce a novel methodology for static OD matrix estimation leveraging traffic flow dynamics and link count observations gathered from a swarm of Unmanned Aerial Vehicles (UAVs) deployed over the network under investigation. We employ a path-based cell transmission model and formulate the problem within an optimisation framework, providing a solution approach for scenarios where data is obtained from (i) fixed-location sensors and (ii) UAVs, for both free-flow and congested scenarios. Notably, the key distinction between the two data types lies in the mobility of UAVs, enabling observations across different links at varying time intervals, while fixed-location sensors measure specific links at stationary positions over time. By comparing estimation results using both data types, we show that OD matrix estimation utilising UAV-based measurements yields significantly superior performance, even with fewer measurements per time-step from the UAV swarm.

ABSTRACT. In this paper, we propose a novel theory-driven framework that is based on a Granger causality-inspired feature selection method and a multitask LSTM to jointly predict two traffic variables. At the core of our methodology, there is a novel traffic flow theory-informed multitask neural network, which is used for the joint short-term forecasting of two traffic variables. For training the model, we propose a custom-made, theory-aware loss function, which incorporates the distance of the emerging multivariate forecast (pairs of traffic variables) from the fundamental diagram of the corresponding location. To enhance the model’s performance and interpretability, network-level traffic information is selected from the most relevant locations, specific to each target location, using the Neural Granger adaptation of classic Granger causality. The proposed approach was found to lead to more accurate, as well as more trusorthy results, which are also consistent with traffic flow theory.

ABSTRACT. To assess the impact of modern mobility services on mode choice, an online survey involving 995 participants was conducted in Carinthia, Austria. Choice experiments were a key part of the survey, revealing users' preferences for transitioning from cars to other transportation modes when entering the city under varied conditions. The survey assumed availability of four mobility services: (1) automated shuttles connecting a technology park to a commuter train station in 10-minute intervals, (2) automated valet parking at Park-and-Ride facilities to save time, (3) e-bike sharing and (4) e-scooter sharing. Choice pairs featuring two routes with identical start and end points were generated via an intermodal routing service. In addition to changing the travel modes, constraints such as parking fees, limited parking availability and longer parking search times were added. Up to six choice pairs were presented to respondents, who selected their preferred route from each choice set.

A typology comprising of five social groups were used to map the general willingness of individuals to change their habitual travel behavior under specific conditions. Primary factors influencing the choice between the original and alternative routes were analyzed. Comfort and time emerged as the main drivers for adhering to the original route, while considerations of environment and costs were the main reasons for opting for alternative routes.

ABSTRACT. Traffic speed prediction is crucial in traffic management and vehicle routing, as it involves forecasting future speeds on urban roads based on historical and current traffic-related data. The trajectories followed by vehicles provide valuable insights into real-time spatial-temporal traffic patterns across a city, and thus are important in modeling traffic speed dynamics. However, existing prediction methods heavily rely on map-matching techniques, which may bring inaccuracies in trajectory generation due to drifted GPS data, particularly in densely populated urban areas. The cumulative errors from matching trajectories will degrade the accuracy of final traffic speed predictions. In this study, we propose GT-Net, an end-to-end GPS-based traffic prediction network that eliminates the need for map-matching. GT-Net initially captures spatial-temporal intercorrelations of historical traffic speed patterns through segment-level reorganization. Besides, it employs a GPS-to-Road attention network to estimate the likelihood of matching between actual coordinates and nearby road candidates, addressing the issue of drifting GPS data and eliminating absolute errors resulting from one-to-one corresponding matching. Additionally, a time-aware attention-based trajectory network is designed to model the influence of varying driving patterns generated by consecutive GPS points on traffic speed dynamics. Extensive experiments conducted on a large city-scale real-world dataset demonstrate GT-Net's outperformance over existing state-of-the-art methods, including those with trajectories involved without the necessity of map-matching.

ABSTRACT. Traffic simulation models in practice mostly utilize historical data for calibration and validation (C&V). Often, the available real data exhibit limitations, notably in terms of the number of days over which they were collected and their spatial coverage within the network. Ideally, ground truth data from different days should be employed for C&V, but collecting all the required data with the desired spatial and temporal accuracy becomes both costly and time-consuming, and in many cases impossible. Consequently, the only viable recourse, which is often encountered in practice, is to gather existing datasets from multiple sources to be parsimonious with the time and cost associated with data collection (Bartin et al., 2018). Moreover, the developed simulation models are generally used to estimate the impact of future, either non-recurrent or planned, events, such as crashes or work zones. When simulations of planned, future systems are created, validation is not possible since field data cannot be collected. It is only after the anticipated event takes place that one can look back at the model's initial assumptions and evaluate its performance retrospectively. Post-evaluation of traffic simulation models is a neglected yet valuable area of research. It offers modelers a chance to validate their predictions of the future system simulations with actual field data. To that end, the objective of this study is to evaluate the performance of a simulation model in handling disruptions and unexpected events in a traffic network with complete and partial data availability. A case study and a detailed real-world dataset pertaining to New Jersey Turnpike (NJTPK) were used to showcase the importance of ground-truth data in traffic simulation models. The results of this study offer valuable insights into the model's robustness, particularly under nonrecurrent traffic disruptions.

ABSTRACT. Rail freight transport is a potential player in the decarbonisation of freight transport. Further improving the role of rail freight in this regard is possible through the automation and digitalisation of rail freight transport to improve rail performance, multimodal services and end customer satisfaction. Various technologies are being considered and tested as part of the digitalisation of rail freight transport (RFF, 2020, CER, 2020). The introduction of digital automatic coupling (DAC) is recognised as a means of improving rail freight performance and competitiveness in the freight market (ERFA, 2022). The digital automatic coupler (DAC Type 4 and DAC Type 5) is an important building block for modern and digital European rail freight transport. Not only will it increase efficiency through automation processes, but it will also ensure a sufficient energy supply for telematics applications and secure data communication throughout the train. The introduction of DAC Type 4 will enable integrated power lines and data bus cables. DAC uses airlines, power lines and data cables to ensure simplified uncoupling, sufficient power supply and secure data lines, further automation and digitalisation of operational processes such as train integrity monitoring, automatic brake testing and wagon order recording, operational health and safety and increased productivity. In addition, electro-pneumatic brakes (simultaneous braking) can be used with DAC (Shift2Rail, 2021, Rail Cargo Group, 2021). However, there is a major dilemma between the different stakeholders regarding the introduction of DAC in rail freight, the benefits, the specifications and the requirements for DAC in Europe. To solve this problem, the benefits of DAC introduction in rail freight should be quantified. In this study, the positive impact of DAC on train processing time and capacity within the Hallsberg marshalling yard has been analysed for the first time.

ABSTRACT. Current urban transportation systems are facing considerable challenges due to the excessive reliance on private vehicles. This has resulted in a range of adverse effects, including urban congestion, air pollution, noise, deterioration of the urban landscape, global warming, dependence on fossil fuels, road safety concerns, and reduced physical activity. While some cities continue to facilitate car use by building more roads and parking facilities, others promote alternative transportation modes, mainly by improving public transportation. Research has evaluated public transportation efficiency by assessing factors like service frequency, vehicle speed, route switching, and connectivity using data from Google Maps. Policymakers should focus on reducing car dependence and enhancing public transport attractiveness and performance. GIS and spatial data analytics are pivotal in identifying areas where public transport services are lacking, facilitating targeted interventions for improvement. This paper examines the role, significance, and efficiency of public transportation in urban mobility through two research questions: the efficiency of a public transportation system and the impact of new public transport infrastructures on network efficiency. To this end, a case study of the Athens Metropolitan Area (AMA) is presented.

ABSTRACT. This paper explores the concept of an 'Idle Wage' for ride-hailing drivers, a policy initiative proposed to regulate the dynamics between drivers and ride-hailing platforms. By compensating drivers for time spent connected to the platform without actively transporting passengers, this policy aims to improve earnings stability and overall job quality. The paper presents a stylized economic model to analyze the impacts of introducing an Idle Wage, evaluating the economic equilibrium under different scenarios and policy constraints. Assuming that drivers are risk-averse, the analysis reveals that in a single-period model, it is optimal to pay the drivers only through the Idle Wage. However, during periods of fluctuating demand throughout the day, earnings through dynamic pricing become essential to match supply and demand effectively. Despite this, the model suggests that the optimal Idle Wage is still greater than zero, underscoring its potential to enhance income stability for drivers

ABSTRACT. The methodology for the URBANPARK model presented in this research offers a comprehensive approach to address urban parking challenges, considering diverse users and employing agent-based microsimulation. This model, together with its case study in Cartagena, Colombia, highlights the impact on traffic parameters, showing an increase in delay times and vehicle density. URBANPARK provides a robust methodology for exploring parking dynamics, informing policy decisions and improving the efficiency of the urban parking system. This work contributes to the development, applications and implications in the field of transport systems and emerging technologies, with a focus on the planning, design, operation and control aspects of transport systems.

ABSTRACT. Signalized intersections on arterial roads lead to prolonged idling and unnecessary accelerations, increasing CO2 emissions. Efforts to address this include cooperative eco-driving, which targets fleet-wise emission reduction at intersections by controlling a subset of vehicles in the fleet. However, devising such cooperative eco-driving strategies that generalize across different traffic scenarios is still a major challenge. We tackle this challenge using multi-task deep reinforcement learning by leveraging network decomposition as the modeling framework. By modeling and analyzing nearly 6000 intersections in three US cities alongside 33 factors influencing eco-driving benefits, our approach yields up to 14% emission reductions on average. Notably, we find interesting insights that were previously overlooked in smaller-scale analyses. We discover that 70% of total emission benefits come from 20% of intersections at every adoption level. However, the specific 20% of intersections that yield the 70% benefits change with eco-driving adoption, calling for further research on how to deploy eco-driving gradually. More broadly, this work paves the way for large-scale analysis of traffic externalities, such as time, safety, and air quality, and the potential impact of solution strategies.

ABSTRACT. Passengers and parcels have long been served by separate vehicle fleets in urban areas. This study investigates the integration of parcel delivery and passenger ride services by a single fleet of mobility-on-demand companies. While this integration holds significant promise for enhancing profitability, it poses a challenge for operation sustainability: the detours required for parcel deliveries could adversely affect the passenger experience. We explore the viability of offering compensations as incentives for detours in a dynamic and stochastic environment. In this scenario, service requests are received dynamically and the passengers' acceptance of detours remains uncertain. We propose an innovative anticipatory policy, ARCC, which integrates routing, confirmation, and compensation decisions. The proposed policy is underpinned by a novel value function approximation (VFA) with slide memory. The numerical experiments have validated the efficacy of ARCC, demonstrating that it surpasses the benchmark myopic policy by an average of 49.4% in profitability, and outperforms the best benchmark anticipatory policy by an average of 14.0%. The study serves as a valuable reference for mobility-on-demand companies regarding the combined transportation of passengers and parcels.

ABSTRACT. The existing transportation networks inherently incorporate multiple modes and are categorized or layered based on the dominant means of transport. Among these, cycling networks specifically accommodate micro-mobility vehicles like bicycles and electric scooters. A crucial concern for users of cycling networks is the perceived safety, particularly in shared spaces with motorized vehicles.

In this study, we propose a simple safety weighting function for segments of a cycling network based on their degree of separation from motor vehicles. We then take advantage of graph theory metrics to evaluate the connectivity characteristics of cycling networks, using them as safety indicators for the entire network.

The primary goal of this research is to optimize overall safety enhancements by implementing incremental road safety improvements in the cycling networks of the city of Grenoble. In addition, we aim to determine the metrics that demonstrate the greatest overall sensitivity to safety improvements at the segment level.

ABSTRACT. Connected and Autonomous Vehicles (CAVs) have the potential to navigate in Lane-Free Traffic (LFT); thereby achieving traffic flow fluidity that was previously unattainable. This research employs a Mean Field Game (MFG) theory-based framework that guides CAV interactions to simultaneously enhance traffic efficiency and safety. It introduces a novel methodological approach that regulates individual CAV actions while considering their collective impact on traffic dynamics. The framework is applied on a hypothetical 5 km freeway segment using MATLAB. The numerical results demonstrate the effectiveness of LFT in managing congested traffic conditions, suggesting that LFT can enhance traffic flow efficiency and safety with proper management strategies.

ABSTRACT. Over recent decades, a plethora of Safety Surrogate Measures (SSMs) have emerged as valuable metrics for predicting traffic conflicts. However, existing research predominantly focuses on conflict prediction at junctions or relies on a single SSM, such as time-to-collision, for detecting vehicle-related conflicts. This limitation highlights the challenge of selecting appropriate SSMs for vehicle- or segment-based conflict prediction, considering the diverse range of factors influencing conflict outcomes. To address this gap, this study leverages data collected from various infrastructure and vehicle-based sensors, including drones, lidars, radars and cameras, across different scenarios in China and the UK: urban junctions, motorway segments and vehicle-based data from instrumented vehicles. Employing machine learning approaches to handle the extensive and disaggregated data, a novel stacked ensemble learning model is proposed. This model integrates a Random Forest (RF), three-layer Deep Neural Networks (DNN), Support Vector Machine Radial (SVM-R), and a Gradient Boosting Model (GBM) meta layer to enhance prediction accuracy. The Recursive Feature Elimination (RFE) algorithm is then employed to identify the most influential SSMs for conflict prediction in each scenario. Results demonstrate the superiority of the stacked ensemble learning model, achieving accuracies of 88% for junctions, 87.5% for motorway segments and 99% for vehicle-based conflicts. Furthermore, the study highlights the necessity of employing different SSMs for conflict detection in various scenarios. These findings hold significant implications for roadway operators and vehicle manufacturers, aiding in the development of strategies to detect infrastructure and vehicle-related traffic conflicts.

ABSTRACT. Risky driving behaviours are acts or operations by drivers on roads that can lead to irregular traffic conditions, causing accidents and harm to the driver or to surrounding other drivers and pedestrians. Near misses are traditionally modelled using self-reported questionnaires, vehicle onboard video, and simulations which are very limited according to the time span and the spatial coverage. This paper utilised the latest data from vehicles that carry onboard Internet of Things devices which can collect the vehicle dynamics (driver’s behaviour in terms of speed, acceleration, deceleration, turning, braking, and g-force) and translate it into near misses at a city level. We then propose a systematic clustering and classification model to quantify how features affect the clustering results and come up with a comprehensive feature ranking with feature importance indexes. The proposed model can be generally applied in any feature engineering process. Results show that the relative location of risky driving behaviour to its nearest intersections and its nearest road curb are the most crucial features in clustering followed by geolocation, speed, road turning angle, g-force, and lane count.

ABSTRACT. Existing literature on signal optimisation mainly aims to improve the vehicle mobility by allocating green duration to typical phase structures (e.g., NEMA dual-ring, eight-phase structure). However, pedestrians with massive delays may prefer to jaywalk due to the inequitable treatment. This paper proposes a generalised signal timing scheme at an isolated intersection, which explicitly incorporates the benefits of vehicles and pedestrians in the objective function. Multi-modal pedestrian crossings is concerned, in which one-stage crossing (OSC) and two-stage crossing (TSC), together with shared and exclusive pedestrian phases appear in given platform. The general signal cycle may contain several sub-cycles. Low-demand phase could be skipped in sub-cycles and high-demand phase could have multiple right-of-way within the overall signal cycle. Over-saturated phases are allowed in sub-cycles. In this way, flexible phase sequence with sub-cycles, phase duration, and cycle length are optimised together to minimise the weighted average delay. The model is formulated as a mixed-integer non-linear programming (MINLP). To mitigate computational burden, we design an algorithm integrating Monte-Carlo tree search (MCTS) and spatial branch-and-bound algorithm with outer approximation, where the former determines the phase sequence and the latter optimises the phase duration. Our framework could achieve a nearly 51% improvement compared to classical four-phase model.

ABSTRACT. Cargo bikes are emerging as a promising and increasingly adopted alternative for last-mile logistics, as urban areas, merchants, and delivery services seek more eco-friendly solutions. Key considerations for electric cargo bikes include their energy usage and the fatigue experienced by the rider, which is influenced by various factors like the mass of the load and the rider's physical attributes. In this study, we introduce a comprehensive energy consumption model that accounts for both the rider and the electric cargo bike, along with considerations for workload balance and the impact of rider fatigue. We present a mixed-integer programming formulation to demonstrate the advantages of incorporating energy consumption considerations into last-mile delivery using electric cargo bicycles. Additionally, we explore the use of a heterogeneous fleet, providing managerial insights for delivery companies to determine the most suitable vehicle types. Our approach also includes route optimization based on the physical characteristics of each driver, ensuring a more personalized and efficient delivery process. Our numerical analysis reveals that, on average, energy consumption can be reduced by 14% and up to 27% compared to traditional vehicle routing models that only consider travel time and workload balance. Furthermore, by incorporating human energy consumption and fatigue into our model, we find that riders have, on average, 15% more energy remaining at the end of their shift compared to models that only consider traditional travel time. This not only enhances the sustainability of last-mile delivery operations but also contributes to the well-being of the riders.

ABSTRACT. Given the inherent sustainability of inland waterway transportation (IWT), it has increasingly captured societal attention. Many organizations advocate for shifting from road to IWT to boost environmental sustainability. Yet, this transition has faced more challenges than expected, with real progress falling behind the goals. Predicting IWT demand is complicated due to the complex external influences and dependency among different ports. Our paper introduces a methodology that utilizes a Spatio-temporal Dynamic Graph Attention-based Multi Attention Model (GAT-DMAN) to improve the forecasting of IWT freight demand. Accurate forecasts allow port authorities to refine strategic planning, helping society move faster towards sustainable transportation modes.

ABSTRACT. Rapid economic growth and technological advancement have fostered increased car dependence, reflected in the increased car ownership around the world. Despite the critical role of vehicles in modern life, parking-related challenges persist, leading to negative externalities such as time, fuel consumption and environmental impact. Smart Parking Systems have emerged to address parking problems, although they lack advanced search capacity, only providing available parking spaces with direct walking access from the car park to the destination. Additionally, existing parking solutions lack the flexibility to accommodate personalised parking preferences for individuals. To address these challenges, this paper proposes a personalised end-to-end parking allocation algorithm using Multi-Agent Reinforcement Learning (MARL) and Grey Relational Analysis (GRA). Experiments are used to evaluate the learning performance of Deep Q-Network (DQN) and Advantage Actor-Critic (A2C) and compares the effectiveness of GRA and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) for modelling personalised parking profiles. Results show the superiority of the A2C-GRA algorithm over others in terms of total and individual rewards, demonstrating enhanced efficiency in minimising travel time and travel distance.

ABSTRACT. Electric vans, autonomous delivery vehicles (ADVs), drones, and truck-drones have the potential to improve delivery services significantly. This study introduces a novel model known as the Parallel Scheduling Vehicle Routing Problem (PSVRP) in an endeavor to revolutionize package delivery by enhancing its efficiency, accessibility, and cost-effectiveness. The PSVRP represents a state-of-the-art approach to vehicle routing problems, incorporating a diversified fleet of innovative delivery modes. The multi-modal fleet of electric vans, ADVs, drones, and truck-drone systems works in unison to minimize operational costs in various settings. The model can also be adapted to different scenarios, such as variations in customer numbers and package weights, by strategically deploying the most suitable mode of transport. The results show that … statistically significantly reduces emissions in all scenarios. Cost is statistically significantly reduced in most scenarios, with the highest number of best results in scenarios with 100 and 250 customers or where most packages weigh less than five pounds.

ABSTRACT. Virtual coupling (VC) technique separates trains by relative braking distances and can shorten train tracking intervals. It has been regarded as one of the important methods for effectively alleviating constraints and improving railway capacity. This work mathematically describes and analyzes the scenario when coupling and decoupling occur after trains pass through a turnout at a railway station. Results show that first when the distance from merging switch to the boundary of station is small (not enough to cover the coupling process in station) the two trains can still couple with proper operation scheme. In this case, it shows that the optimal scheme for these two trains operation is when they are coupled and accelerate simultaneously. This will save more time compared to that when the front train runs out of the station and accelerates immediately. Second, when considering the whole coupling and decoupling process, the coupling operation manner can utilize the railway resources effectively compared to operating trains one by one. In all four coupling and decoupling schemes, the coupling in station (CA) pluses decoupling in station (DA) is the best mode to reduce time intervals between two adjacent trains. This work can provide useful guidelines for the applications of virtual coupling technique in trains arrival and departure.

ABSTRACT. In many recent literature works urban traffic networks are segmented into zones, each exhibiting distinct traffic characteristics, modeled using zonal Macroscopic Fundamental Diagram (MFD) models. On the other side motorways, characterized by complex traffic dynamics such as shockwaves, are often modeled using macroscopic models. Sometimes the distinction between urban networks and motorways is not straightforward, and often urban networks interact with motorway corridors. This paper proposes a hybrid approach, integrating zonal MFD-based and a continuum macroscopic model to simultaneously capture traffic dynamics across both urban zones and motorways. This integration aims to provide a comprehensive understanding of traffic dynamics and facilitate improved traffic management strategies. Furthermore, the proposed approach is computationally efficient with a few modeling complexities, and it can be efficiently applied in traffic control applications

ABSTRACT. In this work, we aim to evaluate three representative ConOps in drone delivery services. Concept A utilizes pre-planned routes with no intersections, where trajectories are confined within these routes. Concept B also employs pre-planned routes but allows intersections, with trajectories still restricted to these routes. Concept C does not pre-plan routes but generates 4D trajectories that could be spatially or temporally separated. The three concepts vary in terms of flexibility, infrastructure and control requirements, and monitoring challenges. In this study, we aim to quantify these differences by considering specific built environments and order characteristics.

ABSTRACT. While Urban Air Mobility (UAM) holds promise for improving urban travel, its integration with existing Air Traffic Management (ATM) systems poses significant challenges. As UAM technology evolves and more VTOLs enter shared airspace, ensuring interoperability with ATM becomes increasingly complex (Levitt et al., 2023). One cautious approach is to design exclusion-area airspace for UAM, simplifying management but potentially limiting operational capabilities. Vascik et al. (2020) emphasized the need for balance, where excluding UAM operations from airspace used by ATC limits available airspace, especially at higher altitudes. Similarly, simulations in the Dallas-Fort Worth area reveal challenges in achieving separation between UAM and conventional aircraft, even with well-designed corridors (Lee et al., 2022). As operational tempo increases and UAM technology progresses, there will be an increasing requirement for coordination between UAM and ATM operations. This coordination becomes crucial as automation capabilities are developed and sufficiently validated beyond UML-4 (Levitt et al., 2023). Despite its promise, relatively few studies have examined the feasibility of integrating UAM operations within airspace, especially in regions with complex flight patterns. One key aspect of safely integrating UAM with ATM is an accurate prediction of non-UAM conventional aircraft trajectories to mitigate collision risks and to ensure the necessary separation between UAM and conventional aircraft. While conventional aircraft typically follow predefined routes, uncertainties can arise especially in terminal airspace due to heavy traffic, aircraft dynamics, environmental conditions, or human factors. To this end, we propose a framework for assessing the feasibility of UAM routes near terminal airspace by leveraging probabilistic aircraft trajectory prediction. Our framework utilizes short-term predicted trajectory distribution information generated by a flow-based deep-generative model, called Normalizing Flows. This information is used to evaluate UAM routes intersecting ATM airspace at various altitudes, with UAM aircraft to decelerate by anticipating encounters with conventional aircraft in the short term.

ABSTRACT. Unmanned Aerial Vehicles (UAVs), referred to as drones, can organize themselves into swarms, fostering collaboration and efficiency in sensor data collection within Smart Cities. With their mobility, autonomy, and diverse sensors, drones have been widely used in the transportation systems. For instance, drones can be used for an accurate monitoring of traffic to detect traffic congestion at early stage. This allows traffic operators to apply mitigation actions that decrease the carbon footprint of a sector with one of the highest carbon emissions worldwide.

Utilizing multiple low-cost drones over a wide sensing area offers a flexible alternative to single high-profile drones. They complete sensing missions in parallel, benefiting from shorter recharging times. This requires coordinated actions with autonomy and computational intelligence. Recent advancements in decentralized optimization and multi-agent learning algorithms offer scalability and efficiency while maintaining privacy and autonomy. However, developing, testing, and evaluating such solutions is complex. Simulation environments simplify studying swarm intelligence and collision avoidance algorithms by reducing complexity and environmental variables. In contrast, real-world drone experiments indoors and even outdoors enhance realism and external validity.

To bridge this gap, this paper introduces a testbed to study distributed sensing problems of drones, such as such as energy consumption, charging control, navigation and collision avoidance. This testbed sets a stepping stone to emulate, within small laboratory spaces, large sensing areas of interest originated from empirical data and simulation models. As a proof-of-concept, a multi-agent collective learning approach is applied to this testbed to coordinate and optimize in a fully decentralized way the navigation and sensing of drones. Furthermore, a potential field collision avoidance method is applied to predict the fields of collisions and finds the optimal flying trajectories of drones to mitigate the risk of collisions and sensing inefficiency. Extensive experimentation using real-world data in traffic monitoring in Athens city validates the efficiency in traffic vehicle observation, demonstrating the capacity of the testbed to move complex swarm intelligence and collision avoidance algorithms for drones to real-world.

ABSTRACT. Traffic monitoring is one of the major tools used for transportation operations and planning. With the emergence of Unmanned Aerial Vehicles (UAVs), new capabilities for enhancing traffic management have emerged. Despite their potential, UAV applications in traffic management have primarily focused on sporadic surveillance of road networks and historical traffic data extraction. Path planning stands out as a critical challenge for UAVs, aiming to optimise routes from initial to target points for specific tasks. In this study, we concentrate on traffic monitoring, and more specifically on the efficient traffic density estimation. Towards this, we propose an online Bayesian optimal UAV trajectory construction methodology. The proposed method strategically selects the next sampling points to obtain traffic density measurements, while minimising the total uncertainty of the traffic density across all time-space points within the studied time-horizon. The proposed approach integrates the Gaussian Process (GP) model into a Bayesian framework to accurately estimate traffic density in multi-lane highways, considering both temporal and spatial correlations, even when data points are sparse. Employing a decision-theoretic approach, we develop a Bayesian optimal UAV trajectory construction scheme to mitigate traffic density uncertainty. Lastly, we conduct a simulation study to evaluate the proposed Bayesian optimal UAV trajectory construction methodology, showing a significant reduction of up to 70\% in the uncertainty of traffic density estimations, compared to a simplistic cyclical UAV trajectory.

ABSTRACT. The UAV (Unmanned Aerial Vehicle)-based commercial services, particularly the drone delivery, constitute a rapidly burgeoning industry. To scale up the operations, a key challenge lies in the real-time scheduling of a large volume of drone. We consider a dynamic urban drone delivery problem which determines order assignment and drone routing. The complexity of such dynamic optimization problem grows exponentially with the number of drones and the complexity of route networks. To address the computational complexity, we propose a Path Pool-based Transformer model integrated with Reinforcement Learning (abbreviated as PPTRL). Different from existing transformer-based algorithms that only use the embedding of the last visited node to predict the next visit node, we devise a dependency decay pooling strategy to capture the path history of each drone, so that past trends are incorporated in prediction decision making. The experimental results show our method's near-optimal performance in small-scale problems, with much faster runtime than Gurobi. For large-scale problems, our method surpasses both meta-heuristic and state-of-the-art learning-based algorithms. Additionally, our approach demonstrates robustness to changes in fleet size, demands, and dynamic task ratios.

ABSTRACT. This paper addresses the challenge of estimating dynamic OD matrices efficiently by drawing inspiration from video summarization techniques. Like summarizing video sequences into key shots, our proposed framework focuses on key intervals, reducing computational requirements while maintaining accuracy. The proposed framework involves Shot Boundary Detection (SBD) and Interpolation components. SBD identifies key intervals, crucial for accurate estimation. We propose a new SBD algorithm based on individual demand and interpolation feedback. Interpolation fills gaps between key intervals using techniques like linear and spline interpolation. Moreover, Split Simulations is also introduced to reduce lower-level bi-level adjustment complexity by capturing assignment matrices only for key intervals With the case study on a real network from Logan, Queensland, Australia, the paper provides justification of gradual travel demand evolution, requiring estimation for a subset of time intervals (key intervals) for accurate OD estimation. It is demonstrated the reduced runtime enables more iterations in bi-level OD adjustment, enhancing convergence. The performance of the proposed KFI method is compared against a state-of-the-art bi-level adjustment framework using a generalized least squares formulation. Results indicate a 4.7% reduction in runtime with our framework and an 80.15% reduction with split simulations compared to traditional methods. Additionally, performance evaluation based on the GEH statistic validates our approach's accuracy, surpassing industry standards.

ABSTRACT. In response to the rapidly evolving urban landscape, there is a growing demand to enhance traditional Origin-Destination Matrices Estimation (ODME) models with new data sources that offer broader perspectives. Emerging technologies and crowd-sourced data in particular, can provide promising avenues for collecting high-resolution data on destination activities, reflecting real mobility patterns. Previous research by Castiglione et al. (2024) investigates trip motivations and the varying travel flexibility associated with different activities, leveraging real-world crowd-sourced data like Floating Car Data (FCD) and Google Popular Times (GPT).

To address the need to integrate these insights into ODME models, this paper introduces the Flex-GLS approach, an extension and generalization of the GLS model. Flex-GLS accounts for multiple demand components characterized by spatio-temporal flexibility metrics derived from crowd-sourced data. It aims to offer a more precise representation of travel demand by integrating both temporal and spatial flexibility dimensions. Benchmarking against the traditional GLS model highlights the potential of Flex-GLS in enhancing ODME accuracy, providing valuable insights into urban travel dynamics.

ABSTRACT. Origin-destination (OD) demand matrices are crucial for transit agencies to design and operate transit systems. This paper presents a novel temporal Bayesian model designed to estimate transit OD matrices at the individual bus-journey level from boarding/alighting counts at bus stops. Our approach begins by modeling the number of alighting passengers at subsequent bus stops, given a boarding stop, through a multinomial distribution parameterized by alighting probabilities. Given the large scale of the problem, we generate alighting probabilities with a latent variable matrix and factorize it into a mapping matrix and a temporal matrix, thereby substantially reducing the number of parameters. To further encode a temporally-smooth structure in the parameters, we impose a Gaussian process prior on the columns of the temporal factor matrix. For model inference, we develop a two-stage algorithm with the Markov chain Monte Carlo (MCMC) method. In the first stage, latent OD matrices are sampled conditionally on model parameters using a Metropolis-Hastings sampling algorithm with a Markov model-based proposal distribution. In the second stage, we sample model parameters conditional on latent OD matrices using slice and elliptical slice sampling algorithms. We evaluate the proposed model using real-world data collected from three bus routes with varying numbers of stops, and the results demonstrate that our model achieves accurate posterior mean estimation and outperforms the widely used iterative proportional fitting (IPF) method. Additionally, our model can provide uncertainty quantification for OD demand matrices, thus benefiting many downstream planning/operational tasks that require robust decisions.

ABSTRACT. Enhancing smart mobility within a smart city is crucial for fostering sustainability and resilience, particularly in a post-pandemic context. To effectively regulate evolving travel demand, it is imperative to estimate and forecast urban mobility dynamics. Recent research such as conditional generative adversarial imitation learning (cGAIL) demonstrates successes in learning human decision-making dynamics from their behavior data. However, the effectiveness of learning directly through the cGAIL method may be hindered by the modeling data sparsity, which is attributed to limitations in data quantity, spatial-temporal coverage, and situational diversity. To address this issue, we introduce a group effect enrichment mechanism into the cGAIL model (gcGAIL) to probabilistically (rather than deterministically) assign an individual trip trajectory to a group given different feature variables. It is a generic form of cGAIL that extends the conditional function by fusing the information of heterogeneous relative importance of certain group aspects (i.e., feature variable) for different revealed expert trajectories when augmenting individual learning data with population data. That is, an individual expert trajectory in gcGAIL is augmented to include the information of several groups instead of one group in cGAIL. It is expected to improve the model performance in prediction accuracy and generalization. We take the longitudinal travel behavior response to a fare-discount promotion as an example to validate the effectiveness of the proposed gcGAIL framework, by comparing it with the state-of-the-art AI models in aspects of accuracy and generalization capabilities, including GAIL and cGAIL. Furthermore, we will compare the learned policies on predicting behavior changes compared to traditional transportation modeling approaches, i.e., the discrete choice model and discuss the advantages/disadvantages of different approaches.

ABSTRACT. Missing data is a pervasive issue in both scientific and engineering tasks, especially for the modeling of spatiotemporal traffic data. This problem attracts many studies to contribute to data-driven solutions. Existing imputation solutions mainly include low-rank models and deep learning models. The former assumes general structural priors about traffic data but has limited model capacity. The latter possesses salient features of expressivity but lacks prior knowledge of the underlying spatiotemporal structures. Leveraging the strengths of both two paradigms, we demonstrate a low rankness-induced Transformer to achieve a balance between strong inductive bias and high model expressivity. The exploitation of the inherent structures of spatiotemporal data enables our model to learn balanced signal-noise representations, making it generalizable for a variety of traffic data imputation problems. We demonstrate its superiority in terms of accuracy, efficiency, and versatility in heterogeneous real-world traffic flow datasets. Promising empirical results provide strong conviction that incorporating time series primitives, such as low-rankness, can substantially facilitate the development of a generalizable model to approach a wide range of spatiotemporal traffic data imputation problems.

ABSTRACT. Demand-responsive multimodal transit offers opportunities to complement existing public transport systems and provide an overall better service level to passengers while at the same time making better use of the resources. This study optimizes the capacity of such system by strategically sizing the required fleet and allocating it to the operating services. We formulate a two-stage stochastic optimization model that plans the transit system and the required fleet in the first stage, and optimizes the demand-responsive operations in the second stage. We develop a decomposition-based method that exploits the network-based formulation of the second stage, allowing us to solve practical instances. Preliminary results from a case study in the city of Zurich show that designing a public transport system together with demand-responsive mobility systems can benefit both transport operators and passengers. By allocating the system capacity more efficiently, operators reduce operational costs while maintaining or improving the travel experience for passengers.

ABSTRACT. Empirical evidence suggests that couriers often deviate from pre-planned delivery routes, responding to practical realities that routing algorithms may overlook, such as dynamic traffic conditions or recipient availability. This study explores the reasons behind delivery drivers’ deviations from assigned routes, particularly how their accumulated experiences inform their navigation choices. Using data from the 2021 Amazon Last-mile Research Challenge, which labels courier-performed routes based on practical considerations like productivity and customer satisfaction, we demonstrate that the quality of a delivery route is influenced by more than just travel time. Factors like turn sharpness, backtracking distance, and neighbourhood visit timing also have a statistically significant impact on route quality. This study trains an energy-based model to predict the likelihood of a route being of high quality when executed in the field. We integrate the model’s insights into an insertion heuristic to generate high-quality last-mile delivery routes.

ABSTRACT. Substantial growth was experienced in the logistics sector, driven by the rapid expansion of e-commerce and the increasing demand for e-grocery services. Collaborative strategies were adopted by logistics companies to enhance delivery efficiency and address the challenges of the evolving landscape. Transshipment, crucial for seamless goods transfer in collaborative and crowdsourced delivery services, traditionally relied on physical facilities like intermediate depots or locker stations. However, these methods faced limitations, including limited adoption, accessibility issues, and suitability constraints for certain deliveries, such as perishable goods.

A practical approach to transshipment was proposed in this paper, specifically tailored to address the needs of small businesses with limited resources and strict delivery time constraints. Introducing "sharing points" (SP) as an alternative to traditional transfer stations, this approach facilitated direct driver-to-driver exchange, eliminating the need for storage facilities and streamlining the delivery process.

Two driver types were defined in the study: "carriers," responsible for pre-assigned parcels, and "assistant drivers," tasked with supporting them. An optimal sharing plan (OSP) was formulated to assign parcels and routes to both driver types, optimizing delivery efficiency. The OSP minimized the objective function, which included total driving time, wait time for carriers and assistant drivers at SP, and maximum driving time (delivery makespan).

A sample scenario with three carrier drivers was analyzed and optimized using Google's VRP solver within a constraint programming (CP) framework. Adding two assistant drivers to help three carriers significantly decreased the maximum delivery time (approximately 42%) and average driving times (35%) while incurring only an 8% increase in total driving time.

This flexible model allows users to prioritize specific objectives based on real-world constraints, presenting a valuable solution for fast delivery services in the dynamic logistics landscape.

ABSTRACT. This study proposes a solution to address the lack of consideration for personalized needs in complex multi-modal transportation systems by formulating and solving a heterogeneous demand traffic assignment problem (HD-TAP). The HD-TAP takes into account the varying preferences of travelers when selecting travel modes and the common occurrence of multiple people traveling together. The use of modular shared autonomous vehicles (SAVs) is also considered in the model, which allows for flexibility in combining the number of modules based on the number of shared riders. The HD-TAP is formulated as a multi-modal, multi-class, multiple equilibrium principles, combined mode split traffic assignment model, incorporating a cross-nested logit model for private vehicle travelers’ route choice behavior and a multi-nomial logit user equilibrium model for non-private vehicle travelers’ mode and route choice behavior. To solve the HD-TAP, a heuristic gradient projection-based algorithm is developed. Numerical examples demonstrate that the proposed algorithm can efficiently solve large-scale multi-modal network problems. Through numerical experiments in real-world networks, the study investigates the impacts of preferred travel modes, the number of shared riders, and the modular operation of SAVs on system performance. The findings indicate that providing an excessive number of modular SAVs with a capacity of five passengers or fewer may result in a loss of public transit users. It is important to control the supply of such vehicles to ensure the preservation of public transit usage.

ABSTRACT. unknown

ABSTRACT. Wider adoption of Electric Vehicles (EV) promises major benefits in terms of reductions in CO2 and air pollution. However, access to charging infrastructure presents a major barrier to mass adoption and continued use of EVs. Charging scheduling has attracted much attention in the literature as it helps to effectively manage limited charging infrastructure and power grid capacity, thereby relieving EV users' charging anxiety. However, these studies treat the demand for vehicle charging as direct, therefore omitting the inherent link with daily scheduling behavior. This study aims to fill the gaps by incorporating EV charging into activity scheduling to capture the inherent links and tradeoffs between charging and activity scheduling. This study introduces a new activity-based framework to estimate charging behavior. Having such a model allows us to realistically understand charging and travel behaviors with a high level of temporal and spatial resolution. It also enables the evaluation of the effects of alternative management strategies (e.g., Time-of-Use tariff, infrastructure build-out) on individual charging and travel behavior.

ABSTRACT. As the growing penetration of electric vehicles (EVs), the transportation network and power grid become increasingly interdependent and coupled via charging stations. The concomitant growth in charging demand has posed a series of challenges for both networks, highlighting the importance for charging coordination. In this paper, we consider the en-route charging station recommendation problem for EVs in the dynamically coupled transportation-power systems, aiming at maximizing the system-level traffic efficiency while ensuring the safety of power grid. The problem is formalized as a constrained Markov decision process (CMDP), and we propose an online prediction-assisted safe reinforcement learning (OP-SRL) framework to find the optimal policy. To achieve this, we mainly address two challenges. First, Lagrangian method is implemented to convert the constrained optimization problem into an equivalent un-constrained optimization problem, and then PPO algorithm is extended to incorporate constraint in the sequential decision process through the inclusions of cost critic and Lagrangian multiplier. Second, we put forward a recurrent neural network (RNN)-based sequence-to-sequence (Seq2Seq) prediction model for state augmentation and online control, accounting for the uncertain long-time delay between performing charging station recommendation and commencing charging; thus offering foresightful information to guide the agent in making more forward-thinking decisions. Finally, we perform experimental studies and compare the performance of the proposed method with six RL-based baseline algorithms. Results demonstrate that the proposed method outperforms the baselines from the perspectives of road network efficiency, power grid security, and EV satisfaction; besides, the introduction of Lagrangian method and state augmentation shows effectiveness in improving the policy.

ABSTRACT. This study presents a real-time shared ride multi-depot Dial-a-Ride (DAR) service where vehicle speeds are variable with time. Accumulation based Network Macroscopic Fundamental Diagrams (NMFDs) in statically partitioned regions are used to estimate vehicle speeds at each time step. The DAR problem is formulated as a multi-objective mixed-integer linear program that considers the conflicting priorities of maximising the operator's total profit and minimising the users' total delay and number of rejections. The entire service fleet is Electric Vehicles (EVs) with homogeneous characteristics. Wait times at charging facilities due to stochastic charging demand from private EVs is predicted through an M/M/S queue model. The DAR optimisation problem includes charging facility selection and scheduling considering the predicted waiting times. Gurobi was used to solve the optimisation problem in a rolling horizon approach. The key contribution of this study is the joint consideration of the regional dynamic traffic model and stochastic charge queuing model during optimisation.

ABSTRACT. In this study, we propose a novel continuum approximation approach to optimally plan urban public charging infrastructure for electric vehicles, considering spatial heterogeneity, serviceability, and region-wide equity. We analytically estimate waiting and travel times to charge with queueing theory and evaluate serviceability over a planning region with spatial heterogeneity. The analytical model explicitly identifies three groups of factors to determine serviceability: (i) planning factors, including the spatial heterogeneity of station density and number of chargers per station; (ii) operational factors, such as station assignment rules; and (iii) exogenous factors, including charging demand, roadway network, and traffic conditions. We validate our planning framework via a numerical example within the urban network and a case study focusing on New York City. We find that solution varies with spatial heterogeneity, high-demand areas require more stations and chargers, while low-demand areas need some for equity concerns.

ABSTRACT. Electric vehicles (EVs) and the infrastructure of electric vehicle charging stations (EVCS) are emerging as essential components of sustainable energy systems. In this context, we introduce an innovative approach that utilizes aggregated EVCS to participate in the auxiliary market, thereby providing grid-balancing services. Our model continuously monitors changes in EV state-of-charge (SoC) across both time and space, taking into account various factors, including driver behavior, current SoC levels, and the associated charging/discharging costs and benefits. This approach will enable charging station operators (CSO), in collaboration with aggregators, actively engage in the frequency containment reserves (FCR) market. We introduce an optimization framework in conjunction with this EV model. For establishing pricing policies with the twin aims of maximizing profits for aggregators and charging station operators (CSOs), while also minimizing energy charging expenses for EV users. Our findings underscore the effectiveness of this pricing strategy in achieving these dual objectives, as demonstrated through realistic simulations integrating the EV mobility and the Electricity FCR market.

ABSTRACT. This research investigates the use of large language models and machine learning techniques to identify the primary triggers for air traffic flow management regulations. The study focuses on textual remarks made by flow managers who implemented these regulations. The investigation takes a concrete form by using weather-related regulations with the referenced location being an aerodrome. Specifically, a large language model is asked to assign each of these regulations to a specific group, or cluster, based on the remark made by the flow manager, where each cluster represents a particular kind of weather disruption. These clusters then act as labels for the dataset, and each regulation is combined with the weather conditions observed during its implementation. This labelled dataset is then used to train a tree-based classifier using supervised learning. This two-step methodology enables the identification of the most likely precise trigger for each regulation, such as low visibility, snow, strong winds, etc. based solely on observed weather conditions. The clusters identified by the large language model are also compared with those discovered in previous research using self-learning and supervised clustering. Nevertheless, the practical applications of this method go far beyond the classification of weather-related regulations. This approach could be used in post-operational analysis to identify the primary triggers of any type of regulation - not just weather-related. Furthermore, it enables the analysis and classification of other types of text, such as notices to airmen, further broadening its potential use cases. This paper showcases the versatility and broad application of large language models in the field of air transportation.

ABSTRACT. Air traffic management and control play distinct, critical roles in ensuring safety and efficiency within the air transportation system. Strategic air traffic management differentiates from tactical air traffic control in terms of timescales and objectives: Traffic managers (TMs) performing strategic air traffic management seek to balance projected demand (e.g., scheduled flights) and capacity (e.g., a safe rate of arriving aircraft at an airport). This is often done well in advance of flight operations. On the other hand, air traffic controllers (ATCOs) ensure aircraft are safely separated, then focus on efficiency and congestion reduction. Both TMs and ATCOs are indispensable to the safe operation of this critical infrastructure---they also face significant stress and high workloads, which is compounded by increasing aviation demand and future new users the airspace must accommodate. Our work aims to leverage generative AI (GenAI) to develop ChatATC, a decision support tool that can relieve some of the repetitive, non-safety-critical workloads of the TMs and ATCOs so that they can focus their attention on tasks that need expert judgment, and relies on specialized training and years of experience.

ABSTRACT. Accurate detecting of road markings is indispensable for autonomous vehicles (AVs) to navigate safely and efficiently. However, current autonomous driving systems still face challenges in perceiving static and dynamic road information leading to inaccurate trajectory and path planning processes, especially in complex driving scenarios. To address these challenges, recent years have seen a predominant reliance on deep learning and attention mechanisms for road marking detection (Jayasinghe et al., 2022). However, persistent challenges such as high computational demands, extensive dataset requirements for training, and manual data annotation issues necessitate exploring alternative approaches. Recent advancements in Large Language Models (LLMs) and Vision Language Models (VLMs) present the possibility of addressing these challenges (Sha et al., 2023). While LLMs have demonstrated remarkable capabilities in reasoning tasks, VLMs offer a distinct advantage by integrating visual and textual information, thereby providing a more comprehensive understanding of complex scenarios (Zhang et al., 2024). Consequently, VLMs are instrumental in interpreting complex driving scenarios and enhance autonomous vehicles’ decision-making processes in real-time (Sha et al., 2023). As a result, this paper proposes a new road marking detection model based on VLM to address the aforementioned challenges. This model leverages both image and label information of road markings, achieving optimal detection results even with relatively small datasets. Furthermore, in conventional autonomous driving, perception and decision-making are two relatively independent processes, and to fully utilise the results of perception to assist the decision-making process, a comprehensive end-to-end pipeline is established that uses GPT-4 (Generative Pre-trained Transformer), an LLM-based language model, for question answering (QA) of road scenarios based on model’s output. This pipeline enhances the autonomous driving system's ability to perceive road surfaces and aids in the decision-making process while driving. The key contributions of this paper are outlined as follows: (1) Development of a new road marking detection model based on VLM, offering reduced training costs compared to traditional methods, (2) Construction of an end-to-end pipeline to enhance the autonomous driving system’s comprehension of road information and generation of driving prompts to support decision-making processes.

ABSTRACT. This study aimed to evaluate the performance of VideoChatGPT, a large video language model, for automatic crash detection and narration from traffic camera videos. For this purpose, a dataset of 500 crash videos was processed, ranging from 2 to 40 minutes. The results showed that VideoChatGPT accurately detected crash times in 2-minute videos, with 72\% accuracy within 4 seconds. However, it struggled with longer videos until a sub-sampling strategy was applied. Additionally, the analysis of 2-minute crash narratives revealed that the model had high accuracy (73.5\%) in detecting vehicle dynamics but very low accuracy (<5\%) in identifying specific collision types such as head-on crashes and rollovers. Environmental details were described accurately 92\% of the time, while contextual details such as road type had 67\% accuracy. However, the model lacked accuracy in causal reasoning and overall description precision with only 54\% accuracy. These findings highlight the current capabilities and limitations of VideoChatGPT, which requires further training on diverse traffic data to improve its automatic video understanding for traffic incident management.

ABSTRACT. Spatiotemporal traffic data can be modeled as a multivariate time series. A crucial task in spatiotemporal analysis is to identify and detect anomalous observations and events from the signals with complex spatial and temporal dependencies. Robust Principal Component Analysis (RPCA) is widely used tool for anomaly detection. However, the default RPCA purely relies on the global low-rank assumption while ignoring the local temporal correlations. In light of this, this study proposes a Hankel tensor version of RPCA for anomaly detection in traffic data. We treat the spatiotemporal traffic data as a matrix (location by time) and decompose the corrupted raw matrix into a low-rank Hankel tensor and a sparse matrix. By leveraging the Hankel constraint on the temporal dimension, the model can simultaneously capture the global and local spatiotemporal correlations and exhibit more robust performance. We formulate the problem as an optimization problem and develop an efficient solution algorithm based on the Alternating Direction Method of Multipliers (ADMM). Despite there are three hyper-parameters in the model, they are easy to set in practice. We evaluate the proposed method by metro passenger flow collected from Guangzhou, China, and the results demonstrate the efficiency of anomaly detection.

ABSTRACT. This article introduces a method to automatically detect, track and stitch the stop-and-go wave fronts and tails in space and time for large-scale vehicle trajectory, enabling the analysis of stop-and-go wave characteristics at scale. The contributions of this paper are outlined as follows: (1) we proposed a method capable of detecting, tracking and stitching both the wave fronts and tails in stop-and-go traffic. (2) we implemented the proposed methods in Python to support reproducible research, data and tools will be provided at i24motion.org. (3) we applied our method to a large-scale dataset, demonstrating the complex phenomena of wave generation, propagation, merge, bifurcation, and dissipation.

ABSTRACT. See attached the extended abstract submitted as PDF.

ABSTRACT. While data is ubiquitous and transforming almost every aspect of transportation, the wide application of data-driven methods also reveal vulnerability of such methods when facing data poisoning attacks. This paper proposes a defense framework against data poisoning attacks using secure data from the infrastructure. We show the three key steps of this infrastructure-enabled defense (IED) framework, including the selection of the secure data, attack detection, and attack mitigation. The IED method is tested using data poisoning attacks on queue length estimation using mobile sensing data. Performance of IED is reported with comparisons with a few benchmark defense methods.

ABSTRACT. This study presents a short-term forecasting framework tailored for dock-based shared micro-mobility services, exemplified through a case study for Capital bike-sharing in Washington D.C. Addressing the complex patterns and the demand variations caused by external uncertainties, we identify four key sources of forecasting inputs: seasonality, weather conditions, passenger flows from connected metro services, and previous demand observations. XGBoost and LSTMs are utilized as the basic models of our forecasting framework to capture the nonlinear dependencies between these features and demand. Additionally, we introduce a two-stage hurdle demand forecasting framework. It first estimates expected demand from contextual inputs, and then refine these predictions by accounting for short-term temporal demand fluctuation effects. Preliminary results show substantial forecasting improvements when these sources are integrated, with the highest accuracy achieved by hurdle models tailored to individual service points.

ABSTRACT. Increasing urban density and environmental concerns have driven the adoption of micromobility (MM) solutions like bicycles and e-scooters, which help mitigate issues like traffic congestion and pollution. These MM solutions are crucial for bridging the first and last-mile connectivity with public transport (PT) systems. Integrating MM with PT enhances urban mobility by using PT data to predict MM demand, optimize bike fleet management, and improve service reliability and user satisfaction. Recent studies utilize both traditional statistical methods and advanced deep learning techniques to predict MM demand using PT data. Traditional methods such as Poisson regression and SARIMA, though interpretable, struggle with complex urban dynamics. In contrast, advanced methods like Random Forests, Gradient Boosting Machines, and deep learning models such as Graph Convolutional Networks (GCNs) are better at handling non-linear relationships and spatial dependencies. The paper proposes a novel approach using a Multi-channel Spatio-Temporal Graph Convolutional Network (MC-STGCN) that integrates PT checkout data with MM demand, also considering external factors like weather, to enhance prediction accuracy and adapt to varying demand volumes. Notably, method performance measures, e.g., Mean Absolute Error (MAE) show improvement for our proposed methodology, which includes bike pickups, subway checkouts, and weather data (MM-PT-W), compared to the other configurations.

ABSTRACT. The advent of advanced traffic data collection technologies, such as loop detectors and GPS devices, has revolutionized traffic management by enabling the generation of vast data amounts. This has facilitated the development of sophisticated machine learning (ML) models for real-time traffic incident prediction. These models, which are trained on historical traffic data, play a pivotal role in forecasting traffic incidents. However, the inherent variability and unpredictability of traffic dynamics and incident patterns necessitate that these models not only predict incidents with high accuracy but also gauge the certainty of their predictions. In critical safety applications, uncertainty estimation is vital, enabling traffic management authorities to measure the confidence levels of predictions and make informed decisions. This capability is essential for effective decision-making in road safety and for implementing these models in production environments. A particularly relevant use case for uncertainty estimation arises during adverse and rare event conditions, such as severe weather, which significantly alters road conditions and driver behavior. For example, if a model predicts a high risk of accidents on a specific road link during a storm but with notable uncertainty, traffic managers might choose to issue a general caution instead of a full closure. This informed strategy optimizes safety without unnecessary traffic disruption by considering both the predicted risks and the confidence levels of those predictions. Research in traffic incident prediction increasingly utilizes Machine Learning (ML), especially Deep Learning techniques such as Graph Neural Networks (GNNs), due to their efficacy in handling complex traffic data patterns. GNNs are well-suited for modelling the graph structures of road networks, a strength demonstrated by Wang et al. (2021a)’s GSNet for regional risk assessment and further applied in city-wide risk evaluations (Wang et al., 2021b). Our previous work advances these models through the Multi-structured Graph Neural Network (MSGNN) (Tran et al., 2023), which integrates diverse data sources for ’network-wide’ predictions. However, the dynamic nature of traffic conditions and behaviors poses a challenge to the reliability of ML models trained on static datasets, potentially leading to obsolete or overly confident predictions in real-world critical applications. Thus, ensuring the reliability and timely applicability of these models is critical, mirroring trends in safety-critical fields like healthcare (Dolezal, 2022) and autonomous driving (Dong et al., 2023), where uncertainty quantification and explainable AI are key. Motivated by the necessity for reliable Traffic Incident Prediction (TIP), we diverge from existing works that emphasize model capability, focusing instead on reliability through our novel Uncertainty Aware Traffic Incident Prediction (UATIP) framework. This enriches predictive models with an Uncertainty Estimation (UE) ability, enhancing real-time interpretability and reliability—a facet not thoroughly addressed in current research. UE aids in discerning highly confident predictions from those warranting caution, thus bolstering the practical utility of traffic prediction models. Our experiments on MSGNN with varied real-world data highlight the UE’s role in augmenting prediction reliability. This advancement portends significant contributions to Explainable AI (Dong et al., 2023) in TIP and supports the application of Trustworthy Transfer Learning (TTL) (Shen et al., 2023), promising enhanced accuracy and robustness in a diverse array of traffic scenarios. Further details and results will be presented in the full paper.

ABSTRACT. Aircraft boarding is always a critical process and mostly driven by the behavior of the passengers. The duration of boarding is influenced by their air travel experience and their willingness or ability to follow boarding procedures. In fact, there is no feedback from the cabin about the current situation or indications of future conditions. It is reasonable to describe the cabin as a black box in today's digital age. This is more surprising given that most passengers carry their own transceivers (mobile devices) and thus technological solutions for feedback from the cabin do exist. From a research perspective, we want to answer three questions. (1) What sensor information needs to be provided at what quality? (2) How to cover domain knowledge in the ML models using complexity metrics? (3) How accurately could aircraft boarding times be predicted?

ABSTRACT. As any complex system, rail networks are vulnerable to delays and disruptions. Rescheduling actions are usually carried out neglecting the presence of uncertainties with a substantial underestimation of the risks. In this paper, given a solution for the real-time train rescheduling problem computed using deterministic methods, we asses the risk caused by the presence of uncertainty on the dwell time duration. We assume that the dwell time uncertainty to be represented by fuzzy numbers and we evaluate the Conditional Value-at-Risk (CVaR) of the makespan in a temporal network with fuzzy activity durations on the arcs. Computational tests carried out on realistic problems show how the proposed approaches are able to compute the exact fuzzy CVaR values of the makespan within extremely short computation times.

ABSTRACT. Investigation of train delay evolution and prediction is one of the hot topics in railway operations research. This is required for the effective and efficient realization as well as the quality improvement of train services. A wide range of papers investigating the train delay evolution and prediction problems has been published in the literature. The Markov chains is amongst the techniques used for this purpose. The interpretability, transparency and stability support its applicability in this respect. The workability of the discrete time Markov chains approach has been shown in the previously published papers. In this paper we extended this research towards the continuous time Markov processes to investigate the stochastic evolution of train delays using the continuous probability functions of time. To the best of our knowledge, this modelling approach applied for train delay investigation is used for the first time in the literature. It is shown in the numerical tests (despite limited in this extended abstract) that the promising results have been obtained in the analysis, evolution and prediction of train delays.

ABSTRACT. This paper proposes exact numerical solution methods for eco-driving problems of a single train. To find global optimal solutions of the problem, the main difficulty lies in the non-convex time dynamics. We propose a convex relaxation for the non-convex dynamics, and a rigorous mathematical proof of the equivalence between the reformulated model and original model could be provided. Although the speed-dependent force constraint of the reformulated model is bilinear, it can be efficiently solved to global optimum by off-the-shelf solvers. Numerical experiments are conducted to investigate performances of the proposed method in terms of solution quality and computational efficiency. Computational results show that the proposed method can efficiently deliver global optimal solutions and outperforms the benchmark method.

ABSTRACT. Power peaks are an undesirable phenomenon occurring in railway networks when multiple electric trains require a large amount of power simultaneously, putting pressure on the power grids. Furthermore, with increased traffic supply in recent years, the chances of having power peaks above certain levels increase, while at the same time, upgrading or over-dimensioning energy supply systems is very costly. One solution for this is fine-tuning of timetables to minimize power peaks. In this paper, we propose a novel simulation-based local search approach for the mitigation of power peaks in railway networks and rescheduling in real-time using two types of control measures: departure time shift and traction power limitation. The framework combines a detailed event-driven train traffic simulator with a local search strategy to efficiently explore and rank combinations of control measures, balancing the mitigation of power peaks and adherence to the timetables. We demonstrate the performance of the approach developed in a real-life case in a part of the Swiss Federal Railways network. The results show its potential to mitigate peaks and reduce energy costs with slight rescheduling actions in real-time with a very fast solving time, delivering nearly optimal solutions and negligible passenger inconvenience.