ABSTRACT. This talk presents a mathematical optimization model and a case study for the Ship-Based Maritime Alkalinity Distribution Problem (SMADP). The SMADP is part of a system for Ocean Alkalinity Enhancement (OAE) where alkaline minerals (e.g., limestone) are dispersed in seawater in order to enhance natural chemical processes. The dissolved minerals increase the ocean's alkalinity, enabling it to absorb more atmospheric CO2. Despite the emergence of first OAE ideas in the late 1990s, OAE has only recently gained significant attention as a promising technology for addressing climate change. Various studies have been conducted to assess the method, including its feasibility, environmental impact, costs, and associated risks. However, little attention has been paid to the maritime logistics required for the distribution of the minerals. The SMADP aims to maximize the net uptake of CO2, considering both the uptake from the limestone distributed and the emissions associated with the ships deployed. It also seeks to find an optimal sailing speed for the ships and a discharge rate for the minerals. Formulating the model requires assumptions and decisions regarding available ships, access to minerals, port capacity, and environmentally sustainable rates of limestone discharge. The solution to the model provides further insight into the feasibility of a large-scale OAE implementation. The computational study conducted explores various scenarios testing different input parameters, to obtain reasonable numbers for these values. The test instances are used to analyze the impact of the planning horizon for distribution, the geographical area, ship capacity, and the sensitivity of the uptake factor. Ultimately, a cost per ton sequestered CO2 is estimated through the conducted computations. This cost is compared with alternative NETs, and we also discuss the potentials of a large-scale implementation.

ABSTRACT. A fundamental decision when planning a voyage for any fishing vessel is determining which locations to visit and fish at. Despite the importance of fishing in the worldwide economy, the application of modern optimization methods in tactical fishing vessel routing has been scarce. Among the numerous factors for this disparity is industry resistance, problem complexity, and the difficulty of creating a dataset that accurately reflects real-time fish populations. However, long-term environmental sustainability necessitates further optimization among Norway's fishing fleets. In accordance with international climate targets, the Norwegian government aims to cut greenhouse gas emissions in its fishing and aquaculture industries by 50% relative to 1990 levels by 2030, with more reductions to come further in the future. One potential avenue for emissions reductions in this industry is through increasing operational and tactical routing efficiency for sailing vessels. To contribute to this large-scale project, this research introduces and develops the Stochastic Fishing Vessel Routing Problem (SFVRP). In the SFVRP, we consider a single large fishing vessel, embarking from and returning to a fixed port. Each potential fishing location is treated as an individual node with stochastic size and reward, drawn from a known distribution. Incorporating fuel and cargo capacity as major knapsack constraints, the objective is to compute a profit-maximizing tour for the vessel. Each time a fishing vessel reaches a new location, it must decide how long to fish there and where to sail afterwords. We present two solution methods for making these choices: first, a deterministic method based on expected reward value, and second, a two-stage stochastic programming approach using scenario generation. In the deterministic approach, we create a simulation framework where consecutive instances of our model are solved at each location. After a fishing action is taken at the beginning of each iteration, we recompute the optimal expected future tour based on how much cargo space and fuel is available. In the two-stage method, we plan an overall route in the first stage and use a recourse function to abort the tour in the second stage when it exceeds the capacities of our vessel. We compare the performance of these methods on a realistic dataset created from historical fishing data by using a specialized heuristic algorithm to solve test instances. Furthermore, this research is presented as a piece of the FAME (curbing Fishing and Aquaculture’s Maritime air Emissions) project, funded by the Research Council of Norway to investigate pathways for reducing greenhouse gas emissions in the fishing and aquaculture industry. Thus, we discuss how this research can provide insight into the effectiveness of adopting low-emission energy propulsion systems among fishing fleets in Norway.

ABSTRACT. Short Sea Shipping (SSS) is a transportation mode that has garnered attention for its potential to alleviate the negative externalities associated with road transport. SSS offers several advantages, including increased safety, reduced infrastructure investment, and environmental benefits. This research specializes in Roll-on/Roll-off (RoRo) freight shipping, focusing on wheel-based cargo such as trucks and trailers. To encourage a shift of cargo from road to sea, we employ Game Theory to model the dynamics of the freight transport market. By analyzing the strategic interactions between transport modes, the research aims to identify conditions under which SSS can become a more competitive and viable alternative. The problem is modeled as a multi-leader-common-follower game. The game is classified as "multi-leader" because at the upper level, both operators influence the market. The sea transport operator determines both the pricing and service frequency, while the road transport operator responds by setting prices. Only the sea operator adjusts frequency, as road transport can flexibly scale capacity by deploying additional trucks. Both leaders are profit-maximizing entities, and the mathematical model incorporates a demand elasticity function, which is approximated using derivatives, to capture how changes in price and service frequency influence the modal split. At the lower level, shippers allocate demand based on perceived utility, which is a function of both price and transit time. The demand allocation is modeled using a logit-based choice model. The solution approach involves iteratively solving the model by computing best responses for each carrier, aiming to reach a stable outcome that approximates a Nash equilibrium. A key contribution of this study is understanding how carriers can optimize their decisions by prioritizing the stability and predictability of long-term contracts, while still leveraging the flexibility offered by short-term contracts. The case study is based on publicly available data for the route between Gothenburg and Ghent. A series of experiments are conducted to explore a range of market conditions by adjusting the cost structures of both sea and road operators. These adjustments simulate both cost increases and decreases, introducing diversity into the analyzed scenarios. In addition, the study examines cases where customer sensitivity to price and travel time differs, and further explores changes in service time. The objective of these experiments is to generate managerial insights with broader relevance beyond the specific case study examined. Preliminary results indicate that sea transport has strong potential to increase market share in most scenarios. However, in cases where customers exhibit high sensitivity to both price and time, the competitive dynamics modeled can result in reduced profits for both operators.

ABSTRACT. Containerized transportation involves two interdependent components: moving laden containers, which carry cargo, and repositioning empty containers. This study focuses on the latter, examining the problem of Empty Container Repositioning (ECR) from the perspective of a liner shipping company. In this context, the flow of containers follows a cyclic pattern: empty containers are provided to export customers, loaded with goods, and transported to their destinations. The containers are emptied once delivered to import customers and must be relocated to areas with new export demand. This repositioning is essential for maintaining service continuity across the transportation network. Although shipping companies generate revenue only from transporting laden containers, the availability of empty containers at export locations is crucial. Due to persistent imbalances between regions that predominantly export goods and those that primarily import them, there is often a mismatch in the geographic distribution of empty containers. As a result, repositioning activities—despite not being directly profitable—are necessary to maximize revenue potential by ensuring that laden transport demand can be met. A well-designed ECR strategy minimizes unmet export demand and the costs associated with container storage and repositioning. Additionally, it helps reduce depot congestion and the required size of the container fleet, along with their related operational costs, while lowering the environmental impact of unproductive container movements. The ECR problem is highly complex due to its scale. It involves multiple locations and time periods, uncertain parameters such as empty container demand and availability, and various operational constraints. As a result, most existing approaches in the literature address the problem at a regional level or rely on simplified assumptions. However, liner shipping companies operate global networks and require decision-support tools to manage ECR worldwide. In this study, we propose a formulation for the global Empty Container Repositioning problem and present preliminary results obtained through a mathematical programming model specifically developed for this purpose. The analysis of these results enables us to identify the main computational and modeling challenges that hinder the effective resolution of the problem. Based on these findings, we outline future research directions aimed at refining both the formulation and the solution methods.

ABSTRACT. We study the Strategic Tramp Fleet Renewal and Retrofit Problem (STFRRP) with emission reduction, in which we consider the strategic fleet management decisions for a dry tramp shipping operator between 2025 and 2050. New vessels can be ordered every five years, and the current fleet is updated by selling/scrapping some vessels, retrofitting some vessels, or keeping vessels in the fleet with increased age. For each five-year period, a fixed demand needs to be satisfied. To effectively measure the fleet emission, the profit, and the number of transported cargoes, i.e., the demand, the STFRRP is integrated with realistic routes on the deep-sea routes described by the Baltic Exchange. The resulting number of routes is exponential. Thus, we use a column generation procedure to include in the model only a small subset of all the possible routes and efficiently solve the problem. The routing problem also includes speed optimization to increase the strategic plan's flexibility.

We developed two models. The first represents individual strategic decisions as flow variables, while the second is a path-based model, where each variable represents a feasible sequence of decisions, showing that the latter can compute optimal or near-optimal solutions. We implemented three different emission reduction schemes: a tax-based scheme, where for each five-year period, the fuel price is increased proportionally to the fuel emission intensity; a hard-constraint scheme, where for each five-year period, a hard cap on the emission is fixed; finally, the most recent International Maritime Organization (IMO) Net-Zero Framework, where a two-tier time-dependent soft-constraint mechanism is used to incentivize low-emission fuels.

A realistic medium-size initial fleet and the corresponding satisfied demand are retrieved from historical data from a collaborating tramp operator. The available engines are diesel, LNG, ammonia, LPG, and methanol. Some engines can use different fuels, e.g., diesel engines can use conventional fossil fuel (e.g., VLSFO), bio-diesel, or e-diesel. We conclude by discussing the different trajectories advantages and disadvantages, mainly in terms of cost-efficiency for the cumulative emissions.

ABSTRACT. This work provides a comprehensive overview of the rail operations at a maritime container terminal characterized by limited space, with a particular focus on the management of train arrivals. Efficient handling of these trains is crucial to ensuring the seamless integration of the port’s multimodal logistics network and maintaining high levels of operational efficiency. The process comprises several interconnected activities, including the coordination of container unloading upon train arrival and their transfer to the export yards. Conversely, the import flow through rail transport requires the planning of train loading and the management of loading operations. Unloading and loading operations necessitate the synchronization of various transportation modes, such as trucks, cranes, and yard equipment. Achieving optimal coordination among these activities is essential to maximize efficiency and throughput [1]. Connected to these issues is the management of the rail yard and the storage of containers awaiting loading onto trains. In the few literature related to rail operations, [2] optimizes the import rail cycle, focusing on minimizing container transfer delays by coordinating yard storage and train dispatching, while in [3], a stochastic dynamic programming model is proposed for prestaging and discharging/loading activities in rail terminals, accounting for uncertainty and cost trade-offs. This work investigates and proposes a classification of the main optimization problems related to the rail process. Furthermore, the study highlights the potential benefits of adopting intelligent decision-support systems that enable real-time adaptation to operational variabilities, ultimately facilitating a more resilient and flexible multimodal transport system. Through extensive analysis and simulation, the research demonstrates how innovative planning strategies can effectively address the challenges of train handling, resulting in more integrated, efficient, and sustainable port-terminal operations.

1. Gharehgozli, A., Roy, D., Saini, S. & Ommeren, J. Loading and unloading trains at the landside of container terminals. Maritime Economics & Logistics, 25, 2022. 2. Caballini, C., Pasquale, C., Sacone, S. & Siri, S. A discrete-time model for optimizing the rail port cycle. (2012,1) IFAC Proceedings Volumes, 45 (24), pp 83-88, 2012. 3. Xie, Y. & Song, D. Optimal planning for container prestaging, discharging, and loading processes at seaport rail terminals with uncertainty. Transportation Research Part E: Logistics And Transportation Review, 119, pp 88-109, 2018.

ABSTRACT. The Circulation Planning Problem (CPP), also known as locomotive assignment or scheduling problem, is a well-established area of research in railway transportation. In this paper, we shift the focus from traditional CPP objectives to minimizing energy consumption. In particular, our goal is to determine the sequence of scheduled train trips for each locomotive in a homogeneous fleet, with minimizing energy consumption over empty-run trips and idling tasks as the primary objective. Empty-run trips refer to movements of locomotives without attached trains, typically occurring between two train trips. Idling refers to periods when locomotives must wait at the same station between two train trips while their engines remain on. We compare our energy-focused approach with benchmark objectives such as minimizing the number of locomotives and reducing the total distance of empty-run trips. To this end, we present a CPP model that incorporates energy consumption during empty-run trips using the widely adopted Davis equation— a physical model based on resistance forces — as well as energy usage during idling and heating phases. While substantial energy savings are achievable, focusing on energy efficiency often increases the total empty-run distance and the number of locomotives required. To address this trade-off, we propose a hierarchical optimization approach. In this framework, energy consumption is treated as a secondary objective, subordinate to benchmark goals. Pareto fronts are used to visualize and evaluate the trade-offs, offering decision support for practitioners. Using detailed gradient information from the Austrian railway network, we validate our model on geographically clustered real-world instances. This enables us to identify patterns in energy savings linked to the topography of the regions. Furthermore, we examine the impact of train delays and increased speeds on energy consumption. Our findings demonstrate that even when energy efficiency is considered as a secondary objective, notable reductions in energy consumption can be achieved through energy-efficient routing, avoiding steep gradients, and optimal locomotive utilization.

ABSTRACT. Improving reliability in a combined freight and passenger multimodal transportation system requires accurate delay prediction and timely decision-making. This research introduces a sequential, integrated framework that combines machine learning and Markov Decision Process (MDP) models to enhance Estimated Time of Arrival (ETA) prediction and operational decision-making. The approach begins with a machine learning model that predicts arrival delays at each segment of a transportation chain using contextual features like weather, passenger demand, and temporal data. At transition points, the predicted delays inform an MDP, which selects the optimal operational action (e.g., departure adjustment, rerouting). The chosen action determines the updated departure time for the next leg, which is then used in the next round of ETA prediction. This sequential process continues throughout the journey, enabling dynamic adjustments and improving end-to-end ETA accuracy. Experiments using simulated multimodal data demonstrate that the framework significantly improves delay management and transport chain reliability, outperforming traditional static models. The method offers a scalable, data-driven tool for real-time decision support in interconnected passenger and freight networks.

ABSTRACT. In maritime container terminals, the efficient coordination of inland and seaside operations is crucial for minimizing delays and ensuring smooth cargo flow. Rail-sea integration plays a pivotal role in this context [1], yet train unloading operations remain almost underexplored in the scientific literature. This study addresses the Train Unloading Planning Problem (TUPP), which involves transferring outbound containers from a stationary train to suitable export yard blocks. Each container is characterized by attributes such as size, weight class, vessel and destination. The unloading process relies on a gantry crane moving along the train and reach stackers operating in the yard. The planning problem involves assigning each container to a compatible block and sequencing the operations to optimize resource usage. In particular, the main aims are to reduce gantry crane travel distance, the number of spreader length adjustments, and reach stacker movements across the yard. We have proposed a Mixed-Integer Linear Programming (MILP) model to solve the problem and introduced some valid inequalities to enhance its computational performance. Since the model performs well only on small-scale instances, while larger cases become challenging to solve optimally, we explore the application of a local search approach where the initial solution is obtained thanks to an ε-constraint method [2]. The ε-constraint approach enables us to treat one objective as primary while bounding the others, producing a set of Pareto-efficient unloading plans [3], used as starting solution in the local search heuristic. Preliminary results are discussed.

References [1] Gharehgozli, A. H. Loading and Unloading Operations in Container Ports: Design and Practices, Journal of Maritime Research, 2023. [2] Mavrotas, G. Effective implementation of the ε-constraint method in multi-objective mathematical programming problems, Applied Mathematics and Computation, 213(2), 2009. [3] Ambrosino, D., Bernocchi, L., & Siri, S. Multi-objective optimization for the train load planning problem, Proc. 14th IFAC Symp. Control Transp. Syst., 2016.

ABSTRACT. Accurately predicting transit delays (e.g., for ferries) is important for reducing missed transfers and improving synchronization across multimodal transit systems. For example, Sarhani et al. (2025) analyzed Sydney data and found that ferries arrived three minutes late in approximately 10% of cases, while early arrivals occurred in only 1% of cases. This asymmetry highlights the value of machine learning (ML) predictions in optimizing schedules—particularly in setting appropriate buffer times between ferry and rail arrivals and connecting buses to minimize passenger waiting times. This study extends the methodology proposed in Sarhani et al. (2025), which leveraged open General Transit Feed Specification (GTFS), weather, and ridership data to identify key operational features—such as turnaround time, headway, and vessel assignment—as primary determinants of delay. The current work integrates these predictive insights into an optimization framework to support more robust and adaptive scheduling.

We consider three strategic approaches: (i) a predict-then-optimize approach, which uses ML outputs to guide schedule adjustments via Mixed-Integer Programming or metaheuristics; (ii) a robust optimization framework that embeds delay uncertainty into planning through data-driven scenarios, inspired by recent advances in transport reliability modeling (Müller-Hannemann et al., 2022; Ricard et al., 2022); and (iii) a reinforcement learning-based control strategy for real-time operational decisions, such as vessel holding and dynamic dispatching (Gkiotsalitis et al., 2022), supported by interpretable policies and feature attribution analysis.

The proposed integrated framework aims to improve schedule reliability and passenger connectivity in ferry-based transit systems by bridging predictive analytics and optimization. It demonstrates how combining explainable ML with advanced optimization supports data-driven planning and operational resilience.

References Gkiotsalitis, K., Cats, O., & Liu, T. (2022). A review of public transport transfer synchronisation at the real-time control phase. Transport Reviews, 43(1), 88–107. Müller-Hannemann, M., Rückert, R., Schiewe, A., & Schöbel, A. (2022). Estimating the robustness of public transport schedules using machine learning. Transportation Research Part C, 137, 103566. Ricard, L., Desaulniers, G., Lodi, A., & Rousseau, L.-M. (2022). Predicting the probability distribution of bus travel time to measure the reliability of public transport services. Transportation Research Part C, 138, 103619. Sarhani, M., Nourmohammadzadeh, A., Voß, S., & El Amrani, M. (2025). Predicting and Analyzing Ferry Transit Delays Using Open Data and Machine Learning. Journal of Public Transportation. (Accepted, in press).

ABSTRACT. In modern supply chains, the efficient transportation of goods is crucial for ensuring timely and effective deliveries. This work introduces a novel approach to optimize load carrier to truck assignments within supply chains. By integrating implicit truck loading constraints into an assignment formulation, we determine the number of trucks needed each day to meet resource demands and ensure on-time delivery. We develop a mixed-integer linear programming formulation that reliably provides lower and upper bounds, along with a heuristic, to minimize total supply chain costs, including item costs incurred from late shipments and truck costs. Through computational experiments in a detailed case study provided by Renault, a major car manufacturer, we demonstrate the effectiveness of our approach in improving decision-making and optimizing truck usage, considering up to 15,000 trucks and 210,000 items. The proposed approach not only reduces costs and supports more sustainable supply chain practices, but also provides decision-makers with reliable minimum and maximum cost estimates for the considered supply chains within minutes. We benchmark our results against a state-of-the-art algorithm from the literature, underscoring the value of integrating truck loading anticipation in assignment models to minimize supply chain costs and accurately determine the number of trucks required for efficient operations.

ABSTRACT. This study addresses inefficiencies in long-haul freight transportation, where trucks often operate with underutilized capacity, increasing costs and environmental impact. We propose a novel integration of Transportation Optimization (TO) with Sponsored Search Advertising (SSA), aligning SSA bidding strategies with optimized transportation routes. This approach targets regions with significant logistical inefficiencies to drive demand, improve vehicle utilization, and ensure optimal resource allocation.

We employ a multi-objective optimization framework to minimize transportation distances, reduce truck underutilization, and maximize the sales impact of SSA campaigns. At the core is our Adaptive NSGA-III, an enhanced version of the traditional NSGA-III evolutionary algorithm. Adaptive NSGA-III incorporates innovative mechanisms, including refined population initialization, dynamic diversity preservation, and advanced selection strategies, to address the complexities of integrating SSA with TO. The framework generates high-quality Pareto-optimal solutions, ensuring effective resource allocation. Mathematical formulas that underpin the optimization framework will be presented during the conference presentation. Validated through real-world data and scenario testing under varying demand conditions, the model demonstrates robustness in addressing operational uncertainties.

Results demonstrate that Adaptive NSGA-III outperforms state-of-the-art heuristics, achieving notable improvements in hypervolume, diversity spread, and computational efficiency. Additionally, adaptive SSA strategies improve truck utilization and advertising returns, effectively bridging logistics and marketing.

This research bridges logistics and marketing by showing how advanced adaptive optimization integrates transportation optimization with digital advertising. It supports scalable, data-driven solutions in dynamic retail environments and highlights broader applications of our algorithm. Future work will refine the model for added constraints, extend it to wider supply chains, and adapt it for short-haul transportation.

ABSTRACT. Air transportation is pivotal in the freight market due to its speed, reliability, and security, particularly amid the rapid growth of e-commerce. As customer expectations for expedited delivery and streamlined processes increase, express companies have begun operating their own fleets of cargo aircraft, while also heavily utilizing the belly capacity of passenger aircraft to meet these demands. To do so, we develop an automated comprehensive decision-making method tailored for next-day delivery services that require rapid responses. This approach employs an integrated optimization model that jointly optimizes dedicated cargo fleet operations, belly capacity booking from passenger airlines, and synchronized cargo allocation, thereby improving operational efficiency and resource utilization within the air cargo service network. We also incorporate through cargo connections into cargo routes to enhance network connectivity. These connections occur when the same aircraft operates consecutive flights within a cargo route. To efficiently solve the integrated optimization model, we propose an accelerated column generation-based algorithm that prices out multiple columns per iteration. We transform the pricing problem into a longest path finding problem and develop a specific longest path algorithm. Additionally, we leverage the totally unimodular property of our model to convert some integer variables into continuous variables, thereby reducing computational time without compromising solution quality.

We conduct extensive experiments using data from a major air freight company in China, comparing our algorithm with CPLEX across various scales. Our method outperforms CPLEX in large-scale instances, reducing operational costs by approximately 3–8%. We also analyze the impact of the parameter in the longest finding algorithm on the efficiency of our column generation approach and demonstrate the benefits of the proposed acceleration strategy. Managerial analysis reveals that utilizing through cargo connections, which eliminate transshipment, significantly reduces unserved demand. We also examine how unserved demand penalties affect operations. At lower penalty costs, there is a disincentive to transport all cargo demand, resulting in lower utilization of larger-capacity flights which have higher fixed costs. Conversely, higher penalty costs necessitate the operation of additional flights to increase cargo capacity, and shift unserved demand toward long-distance routes. Additionally, smaller-capacity flights show decreased average flight load factors, whereas larger-capacity flights increase.

ABSTRACT. In vendor-to-retailer systems where retailers place orders independently and vendors dispatch trucks within soft time windows, vendors must respond reactively with limited control over delivery timing. When retailers have limited inventory capacity and place orders early, trucks may arrive before sufficient space is available. Trucks are required to wait on site until unloading is possible, resulting in inventory-driven delays. Ignoring such delays may bias dispatch decisions and lead to underestimating logistics costs. While some models allow variable service times, delays caused by inventory constraints remain largely unaddressed. This study models the dispatch problem as a discrete-time Markov decision process (MDP) with inventory-dependent service times. The model is solved using Approximate Dynamic Programming (ADP) with a Partial Value Function Approximation (PVFA) that ranks outstanding orders based on the cost of delaying. The cost of delaying an order is calculated using expected rewards and an opportunity cost term to represent the future value of vehicle availability. The model is implemented at an Ethiopian brewery with a closed-loop logistics system and a limited number of crates, thereby constraining storage capacity at warehouses. The implemented model is benchmarked against a First-In First-Out (FIFO) policy across three replenishment strategies: Retailer-Managed Inventory (RMI), where orders are generated using a Poisson distribution; threshold-based replenishment, where orders are placed with a fixed lead time; and Vendor-Managed Inventory (VMI), where the model decides when to place orders. Results indicate that the PVFA outperforms FIFO across all replenishment strategies, reducing extended service times by 65% to 93%, depending on the scenario, and freeing 8% to 16% truck capacity. Unmet demand is also reduced by 53% to 98%. The results demonstrate that inventory-driven delays can be effectively incorporated into a dispatch model, providing soft handling of inventory capacity constraints. The PVFA consistently outperforms the FIFO benchmark across all scenarios, indicating its ability to anticipate in a reactive setting. Moreover, the PVFA has a lower variance on key performance indicators than the FIFO, suggesting robustness under different replenishment systems. As such, the model is well-suited for environments where order policies vary or where inventory-driven delays are common, such as in Ethiopia and beyond. Future work may address dynamic order generation, parameter uncertainty, and extensions toward the general Inventory Routing Problem (IRP).

ABSTRACT. This paper addresses a real-world construction logistics problem related to optimizing waste bin collection and transportation. The objective is to efficiently schedule a heterogeneous fleet of trucks, each with a single-bin capacity, to maximize the number of tasks realized over a single period. Each task involves making a round-trip from the platform to the construction site to collect a full bin. To do so, the truck starts with an empty bin loaded in the platform, drops it off at the construction site, and picks up a full bin, which is then unloaded at the platform. The problem is modeled as an original interval scheduling variant with multiple availability intervals for both tasks and vehicles. Each task is assigned to a compatible vehicle interval based only on its service duration, which is shorter than its time window. This partial task interval assignment and multi-window structure set the problem apart from classical formulations. We first propose a generic mixed-integer linear programming formulation, which is subsequently refined with valid inequalities to enhance its computational performance. The two models are evaluated on 99 instances, demonstrating their effectiveness in optimizing waste collection operations within practical time limits.

ABSTRACT. Improving operational efficiency has become increasingly important for the oil and gas industry on the Norwegian Continental Shelf (NCS), where cargo distribution using Platform Supply Vessels (PSVs) contribute to high operational costs and greenhouse gas emissions. This paper introduces the Collaborative Supply Vessel Planning Problem (CSVPP), in which we determine cost-effective weekly PSV routes and schedules for a number of oil and gas operators participating in a collaboration, while satisfying the cargo demands at offshore installations that are serviced from multiple supply bases. We propose a new mixed integer programming voyage-based model for the CSVPP, for which candidate voyages are generated as input using a dynamic programming label setting algorithm. The computational study is based on a real-world case with 39 offshore installations (with 101 weekly visits) from four offshore operators operating from two supply bases along the Norwegian coast. We incorporate two levels of collaboration, namely operator sharing and base sharing, and evaluate the cost reduction potential from each of these and the combined approach (i.e., full resource sharing). The results reveal a cost reduction of between 19.8% and 25.6% in the case of full resource sharing.

ABSTRACT. Global warming is significantly affecting Arctic ice conditions, making maritime routes such as the Northeast Passage (NEP) and the Northwest Passage (NWP) increasingly viable for commercial shipping. These Arctic routes offer notable distance savings compared to traditional Suez and Panama Canal routes, promising substantial reductions in fuel consumption and emissions, thus presenting both economic and environmental benefits. Despite these advantages, the Arctic routes pose inherent challenges such as limited seasonal availability, unpredictable ice conditions, and the necessity of icebreaker escort services. These operational constraints particularly affect liner shipping, which relies on fixed schedules and predictable service reliability. However, existing research on the liner shipping network design problem (LSNDP), has yet to incorporate the potential benefits and complexities introduced by these Arctic routes. A liner shipping network is composed of a set of rotations, which are cyclic sequences of ports to visit. The LSNDP seeks to develop a network of these rotations to maximise its profitability while meeting transportation demand. Solving this problem for an instance of reasonable size is challenging due to the problem complexity and the large number of possible rotations.

Given the unique operational challenges and opportunities presented by Arctic routes, it is crucial to systematically evaluate their potential integration within existing maritime networks. In this paper, we extend the WorldSmall instance of LinerLib, which is the popular benchmark instance set for the LSNDP [1]. Our approach conservatively models Arctic arcs with their lower travel distances, sailing speeds, and icebreaker escort fees. The solution process is based on [2] and [3].

Our computational experiments explore several scenarios with different time horisons and using conventional and ice-going vessels. The latter requires less icebreaker escort, capable of operating the Arctic longer, but incurs higher charter rates. Preliminary results indicate significant profitability improvements and a shift in container traffic, especially from the Suez Canal.

References: [1] Brouer, B.D., Alvarez, J.F., Plum, C.E., Pisinger, D. and Sigurd, M.M., 2014. A base integer programming model and benchmark suite for liner-shipping network design. Transportation Science, 48(2), pp.281-312. [2] Johansen, M.L., Holst, K.K. and Ropke, S., 2025. Designing the Liner Shipping Network of Tomorrow Powered by Alternative Fuels. Transportation Science. [3] Krogsgaard, A., Pisinger, D. and Thorsen, J., 2018. A flow‐first route‐next heuristic for liner shipping network design. Networks, 72(3), pp.358-381.

ABSTRACT. Decarbonizing the shipping sector is essential for advancing the maritime energy transition. Each year, international shipping contributes millions of tons of CO₂ emissions, driving a growing shift towards renewable fuel alternatives. Among these, hydrogen and its derivatives methanol and ammonia stand out as promising energy carriers, particularly due to their comparatively easier handling in maritime contexts. However, adopting these alternative fuels introduces significant challenges to the current maritime energy supply infrastructure. Since hydrogen, methanol, and ammonia possess lower energy densities than traditional heavy fuel oil, their use requires larger volumes and more frequent refueling. Additionally, it is expected that multiple fuel types will coexist within the shipping industry moving forward. Leading ship operators such as Maersk and TUI have already begun deploying vessels equipped with dual-fuel engines capable of running on methanol, while engine manufacturers like MAN Energy Solutions are developing similar engines that are compatible with ammonia. Consequently, fuel suppliers will need to diversify their offerings, comparable to the variety of fuels available at modern vehicle refueling stations. In response to these complexities, this research uses a mathematical optimization model to strategically plan the locations of renewable fuel refueling stations for maritime applications. The model integrates various factors including different shipping routes and vessel types, fuel boil-off effects, bunkering methods via pipeline or ship-to-ship transfer as well as multi-fuel considerations. The goal is to identify cost-effective sites for both shore-based supply hubs and flexible ship-to-ship refueling stations, while determining the optimal capacities for methanol and ammonia at each location. To achieve this, an existing flow-refueling location optimization model has been enhanced and adapted. The model is applied to a case study focused on the Baltic Sea region, examining multiple scenarios for future methanol and ammonia utilization in shipping. Results are expected to vary significantly depending on the scenario: if demand for one fuel dominates, supply of the other may be consolidated at fewer ports, with increased reliance on shipto-ship delivery as a cost-efficient solution. Furthermore, port-to-ship bunkering is likely to be concentrated at key shipping route intersections, whereas less trafficked ports will primarily depend on ship-to-ship refueling operations.

ABSTRACT. In this work, we study integrated production planning and vessel scheduling problem within a fertilizer supply chain. We propose integrating decisions related to production planning, incorporating constraints related to transfer capacities, storage capacities at the port and industrial entities, as well as berth allocation and vessel scheduling decisions. The problem is formulated as a Mixed-Integer Linear Programming (MILP) model aimed at maximizing total revenue while considering demurrage costs. To solve the problem, we propose a branch-and-check variant of Logic-Based Benders Decomposition. Numerical experiments based on real-life and randomly generated instances are conducted to evaluate the proposed approach. The results show that our method can solve the model with an average optimality gap of 4% within one hour.

ABSTRACT. Inland waterway transport (IWT) is increasingly recognized as a low-emission alternative to road freight, aligning with broader decarbonization objectives in the transport sector. However, the widespread deployment of battery-electric and hydrogen-powered vessels is constrained by the absence of robust and cost-effective bunkering infrastructure. This study proposes an integrated methodological framework that combines physics-based simulation and mixed-integer optimization to determine the optimal location and capacity of multi-fuel bunkering stations along a transnational corridor. Fuel consumption is estimated through a resistance-based simulation, using the Holtrop–Mennen algorithm to account for vessel properties, load conditions, and waterway characteristics. These simulation outputs inform a network-scale optimization model that minimizes infrastructure investment, operational costs, and supplier logistics, while explicitly capturing vessel autonomy, fuel-type compatibility. The framework differentiates between the distinct operational profiles of hydrogen and battery-electric vessels, enabling hybrid infrastructure layouts that balance spatial coverage with capital efficiency. Results indicate that aligning station design with vessel-specific requirements and hydrodynamic variability significantly enhances network resilience. The findings provide actionable insights for corridor-level planning of zero-emission bunkering networks in inland shipping.

ABSTRACT. The growing frequency of climate-induced disruptions place increasing pressure on Inland Waterway Transport (IWT), which is a critical component of port-hinterland connectivity in Europe. Reduced water levels constrain vessel draft and loading capacity, resulting in delays, rising costs, and weakened reliability, all of which threaten the modal share of IWT in port-hinterland logistics. This study evaluates the strategic role of intermodal transfer hubs as an adaptive solution for sustaining cargo flows along the Rhine-Alpine corridor under climate-induced disruptions. A transport network model is developed to simulate three scenarios: baseline, drought without intervention, and drought with hub intervention, focusing on port competition between Rotterdam, Antwerp, and Hamburg. Using vessel data, historical drought records, and cost functions for road, rail, and waterborne transport, the model incorporates dynamic water depth constraints and modal shifts via transfer hubs. Critical bottlenecks along the corridor emerge as key locations where transfer hubs can mitigate transport disruptions. A multi-criteria decision analysis, based on the Best-Worst Method, identifies candidate hub sites by evaluating factors such as infrastructure, environmental impact, and investment costs. Results demonstrate that hubs, particularly in Duisburg and Andernach, enable up to 80% transport operability if transshipment costs remain within feasible bounds. Our findings highlight the potential of strategically positioned intermodal hubs to buffer the effects of climate variability, secure port competitiveness, and reinforce sustainable freight transport in Europe. The proposed methodology offers a replicable framework for planning resilient port-hinterland corridors in other climate-sensitive regions.

ABSTRACT. Modern ports are increasingly integrating advanced technologies to monitor, analyze, and manage operations within cyber-physical smart port systems. These environments generate vast amounts of heterogeneous data with the potential to support system-wide decision making and optimize performance across diverse stakeholder interests. However, the current research landscape in the domain remains fragmented, characterized by divergent terminologies, disconnected frameworks, and varying disciplinary perspectives. This conceptual ambiguity hinders the ability for organized system-wide integration of big data and artificial intelligence-driven solutions in port performance optimization. The current paper addresses the gap by clarifying key concepts and synthesizing existing theoretical strands into an integrated conceptual model that connects the smart port system components with the decision areas across four strategic performance dimensions – efficiency, capacity, safety, and sustainability – while identifying relevant artificial intelligence applications and techniques for addressing these challenges. Drawing on the five-layer IoT architecture and the Industrial Artificial Intelligence pipeline, the model provides a layered lens through which to understand how data flows, is processed, and informs performance-oriented decision making in smart ports. By linking technological layers with performance objectives and moving beyond isolated artificial intelligence methods, the proposed model promotes a more holistic understanding of intelligent decision making in smart port infrastructure. The model aims to offer both theoretical clarity and practical guidance, supporting future empirical research and informing structured implementation efforts in the evolving landscape of smart port development.

ABSTRACT. This paper introduces a voyage optimization model designed for containerized battery-powered inland vessels. The objective of the model is to minimize the total cost associated with the voyage. It optimizes the number of batteries on board and the average speed at each leg of the journey. It incorporates a computational framework for energy consumption and hydrodynamic resistance, accounting for vessel characteristics and environmental conditions. Applied to a round-trip mission between Rotterdam and Ghent, the model illustrates the operational and economic benefits of adjusting battery configurations per leg. A low-water scenario is also analyzed to provide insights into the adaptability of battery-swapping technology under challenging conditions. This study supports the development of low-emission voyage strategies and strengthens the case for broader adoption of battery-swapping technology in the inland transportation sector.

ABSTRACT. As the global energy transition accelerates, importing green hydrogen emerges as a crucial strategy for countries with limited domestic production potential. However, selecting the most suitable import route involves complex trade-offs across economic, environmental, technological, and social dimensions. The literature shows that there is a need for multi-objective and multi-decision maker frameworks in sustainable supply chain planning.

In this presentation, we propose a mathematical optimization model designed to support decision makers in identifying the optimal green hydrogen import option based on individual preferences. The model evaluates direct import options from single exporters as well as more complex import networks involving multiple source regions. It integrates a multi-objective utility function composed of weighted and normalized values, including cost, emissions, social acceptance, technology efficiency, technology readiness, and resilience. Resilience is expressed through diversification of both the production and the last stage of import. Normalization and thus comparability is achieved through benchmark values derived from the optimal import options of each criterion in isolation. By adjusting the weights, the model adapts to the priorities of different stakeholders, offering a flexible and transparent decision-support tool.

Results include the identification of suitable import options under varying stakeholder preferences and scenarios, as well as insights into trade-offs between criteria such as cost and resilience. The model is expected to reveal how diversified import networks can enhance resilience at the expense of other performance measures, and under what preference profiles direct imports might remain the preferred option. These findings are intended to guide strategic decision-making and the formulation of effective policies for the development of green hydrogen import infrastructure.

ABSTRACT. Climate change and environmental degradation have intensified the need for effective reforestation strategies to restore ecosystems and mitigate biodiversity loss. This study presents a mathematical and computational model designed to optimize plant distribution in reforestation projects, ensuring species coexistence and minimizing competition. The research employs stochastic optimization techniques, incorporating elements such as Monte Carlo simulations and probability distributions to enhance decision-making in plant allocation.

The proposed model considers multiple constraints, including species compatibility, site-specific characteristics, and cost minimization. A vertex-coloring approach from graph theory is utilized to prevent monoculture, ensuring biodiversity conservation. The model’s effectiveness is validated through computational simulations, where plant distribution is optimized within a structured triangular grid system. Preliminary results indicate that the optimization model significantly reduces species competition while maximizing plant survival rates.

Additionally, the study explores the potential of heuristic algorithms, such as genetic algorithms, to further refine plant distribution strategies. Future implementations will incorporate adaptive learning methods to enhance the model’s flexibility and applicability to diverse environmental conditions. The findings underscore the importance of optimization in ecological restoration, demonstrating how mathematical modeling can contribute to sustainable reforestation efforts and climate change mitigation.

ABSTRACT. Deforestation in Mexico has reached alarming levels in the last decade, impacting ecological balance and exacerbating issues such as the water crisis and climate change. In response to this problem, this study develops a mathematical optimization model to improve reforestation logistics in the Mexican Altiplano. The research focuses on minimizing the distance traveled by transport vehicles by optimizing plant distribution routes from a central depot to various planting sites.

The problem is modeled as a Vehicle Routing Problem (VRP), considering constraints such as vehicle load capacity, the demand of each planting site, the available working hours per day, and the homogeneous distribution of native species. A mathematical model is proposed with an objective function and constraints, complemented by a heuristic method based on the nearest neighbor algorithm.

The results show that the heuristic approach provides efficient solutions with reduced computational time. Experiments were conducted by varying the number of planting sites and evaluating the impact on solution quality and processing time. The analysis indicates that the model successfully minimizes operational costs and optimizes resource use, ensuring the proper distribution of plants in the reforestation area.

This study highlights the importance of optimization in environmental projects, demonstrating that optimization techniques can significantly contribute to sustainability and climate change mitigation. The implementation of mathematical models and heuristic algorithms in reforestation logistics represents a crucial step in improving the efficiency of these initiatives and ensuring the conservation of affected ecosystems.

ABSTRACT. In the mining industry, the production of marketable ores involves multiple processes, such as extraction, treatment, transportation, blending, and drying. Blending decisions depend on previous processing options. Ignoring these dependencies can lead to suboptimal solutions, while integrating blending decisions into the supply chain increases planning complexity.

Uncertain demand, short lead times, and prioritized orders lead managers to use substitution strategies, providing products with deviated qualities or delayed delivery. This approach often results in blending rare superior qualities to fulfill less profitable demands, leading to significant production cost losses and valuable resource destruction. These substitution strategies often lead to inefficient asset utilization that results in over-consumption of water and energy due to missed optimization opportunities.

To address these issues, we propose a spatio-temporal decomposition methodology dividing the mining supply chain into Tactical, Operational, and Emergency Levels. These planning levels communicate via an integrated rolling horizon approach, ensuring a continuous information exchange, synchronized mid-, short, and urgent-term purposes, and adapt dynamically to unforeseen events affecting the production system. In real-world phosphate mining applications, this methodology has reduced total costs by 30\% and significantly decreased water and energy usage compared to traditional management practices.

ABSTRACT. This study addresses the integrated design and optimization of a water supply system based on reverse osmosis desalination to meet the needs of users with differentiated water quality requirements. The main objective is to minimize the Levelized Cost of Water (LCOW) while jointly considering production, multi-quality transportation, and storage decisions. In contrast to previous studies that address production and transportation separately or do not consider water quality differentiation, we propose a comprehensive mathematical model that integrate treatment processes, transport flows, storage dynamics, and quality constraints. To deal with the non-linearities of water transport modeling, we introduce an approximation procedure that preserves the essential system characteristic while ensuring computational tractability. Through numerical experiments under varying demand and cost scenarios , we identify the conditions under which the supply of multiple water qualities becomes economically viable. The analysis yields also some managerial insights to assist decision-makers in configuring cost-effective treatment and transportation infrastructures, prioritizing investments under budget limitations, and assessing tradeoffs between internal storage capacity and external storage contracting. These results offer valuable guidance for the strategic planning of large-scale water supply systems in contexts marked by resource scarcity and heterogeneous demand.

ABSTRACT. In door-to-door intermodal freight transportation, container drayage plays a crucial role in ensuring efficient and timely service (Yu and He, 2025). Container port drayage, also known as port-hinterland transportation, is crucial for facilitating intermodal transport services, encompassing both short- and long-distance movements (Chen et al., 2024). This study examines a logistics operator conducting drayage operations for shippers utilizing Chile’s two major ports: San Antonio and Valparaíso. The company’s facilities serve as empty container depots and bonded warehouses, offering a range of services, including customs inspections, cargo stuffing and destuffing, storage, and transportation. The company operates three depots, located in Valparaíso, San Antonio, and Santiago, each managed by a logistics coordinator responsible for assigning transportation orders to a heterogeneous fleet of external trucks. The company aims to optimize assignment plans that comply with operational constraints while ensuring a high service level, measured by on-time freight delivery and cost minimization. Additionally, depots function as staging facilities, requiring coordination between multiple trucks, for example, one to retrieve containers from the port terminal and transport them to the bonded warehouse for temporary storage, and another to deliver them to customer facilities, oversee destuffing, and return empty containers to a designated depot, either company-owned or operated by competitors. Since external truck operators’ earnings depend on the number of assigned jobs, workload balancing is a key social consideration. However, the company currently assigns cargo daily, limiting its ability to incorporate workload balancing measures. To address this, we propose a decision-support framework to assist depot managers in planning operations over a one-week horizon. The mathematical model developed seeks to minimize operational costs while achieving balanced workload distribution among truck operators, commonly referred to in the literature as route balancing. This aspect has not been widely considered in the literature, except for Wang et al. (2024), that model this criterion based on the route length. In our case, we define it in terms of the number of assigned orders. Furthermore, the proposed framework aims to reduce empty truck trips by coordinating movements across the three depots, thereby lowering emissions.

References Chen, R., Meng, Q., & Jia, P. (2022). Container port drayage operations and management: Past and future. Transportation Research Part E: Logistics and Transportation Review, 159, 102633. Wang, D., Zhang, R., Dong, M., & Xie, X. (2024). Drop-and-pull container drayage with route balancing and its matheuristic algorithm. Expert Systems with Applications, 255, 124625. Yu, X., & He, C. (2025). An Improved Large Neighborhood Search Algorithm for the Comprehensive Container Drayage Problem with Diverse Transport Requests. Applied Sciences, 15(11), 5937.

ABSTRACT. The rapid growth of international maritime trade has intensified the operational challenges faced by marine terminals. These challenges stem from the increased interaction between multiple transport modes, including vessels, trucks, and trains. Key issues include ensuring efficient berth allocation to manage vessel traffic, reducing truck congestion through effective appointment scheduling, and optimizing terminal resource utilization. Seamless coordination among these modes and collaboration between stakeholders such as vessel operators, logistics companies, and terminal managers, is essential to mitigate inefficiencies. To address the complex coordination challenges in terminal operations, this paper proposes a multi-agent, multi-objective model that synchronizes vessel berth allocation with truck appointment scheduling. By generating Pareto-optimal solutions, the model explores a transparent and fair approach to resolving competing objectives, such as minimizing vessel waiting times, reducing truck congestion, and optimizing terminal resource utilization. A solution methodology based on prioritized planning and neighborhood search is proposed and tested against multi-objective evolutionary algorithms (MOEAs), such as NSGA-II and SPEA2, commonly used for similar problems. Numerical experiments demonstrate that the proposed algorithm not only generates a higher number of Pareto optimal solutions but also achieves a superior hypervolume compared to MOEAs. This would enable decision-makers to evaluate trade-offs related to port terminal operations in a more informed way.

ABSTRACT. This study addresses a problem of city logistics planning faced by a central orchestrator responsible for coordinating delivery operations between retailers and freight transporters. The orchestrator has specific contractual agreements with the couriers. Specifically, each courier performs a single tour per day, with the contract specifying a time limit on tour duration. As long as this limit is not exceeded, no additional costs are incurred. Conversely, if a courier’s tour exceeds the time limit, an extra fee proportional to the extent of the violation must be paid. In this paper, we focus on operational planning and assume that contracts have already been established. As such, the problem we introduce is a variant of the Capacitated Vehicle Routing Problem with Time Windows (CVRPTW), where the objective is to minimize the total penalty costs resulting from violations of the duration of the courier tours. This variant incorporates a nonlinear objective function, which cannot be accommodated by standard VRP solvers, thus requiring a specialized solution approach. To address this, we develop a mathematical formulation of the problem and we design a branch-and-price algorithm. Computational experiments are conducted to evaluate the performance of the algorithm and provide practical insights into the structure of optimal solutions.

ABSTRACT. As urban logistics systems shift toward more sustainable and adaptable delivery models, a growing number of alternative vehicle types have been introduced or tested in recent years. Cargo bikes and delivery robots, in particular, are increasingly used either as supplements to or replacements for conventional delivery vans. However, their operational efficiency varies considerably depending on the spatial context in which they are deployed. Urban areas differ widely in their structure and composition, ranging from dense inner-city neighborhoods to suburban zones, and often include natural barriers or fragmented street networks. These differences raise the question of how spatial characteristics affect the performance of various delivery modes, and whether such characteristics can serve as reliable indicators for vehicle allocation in the planning process. This presentation explores the potential of spatial data to support such decisions by estimating which delivery modes are best suited to specific urban settings. Using Berlin as a case study, the city is segmented into a uniform grid of 1 x 1 km cells. For each cell, delivery routes for vans, cargo bikes, and delivery robots are generated using an Adaptive Large Neighborhood Search (ALNS) heuristic. The resulting routing outputs are analyzed in relation to spatial variables, including features of the built environment (e.g., building density, land use diversity, proportion of green and open space) as well as network-based indicators (e.g., node count, connectivity, and centrality). Preliminary findings show that delivery performance varies systematically across different spatial contexts. Robot-assisted deliveries are particularly sensitive to characteristics of the built environment and land use, while vans and cargo bikes exhibit stronger correlations with the structure of the street network and customer density. These results suggest that spatial indicators can offer valuable guidance for early-stage planning, even in the absence of detailed routing data.

ABSTRACT. Electric vehicles are proposed as the sustainable transport technology for the future. Last-mile parcel delivery is an area of transport that has a hunger for new technologies offering efficiency and reputational benefits. E-vans have high range, capacity and maximum speed compared to e-cargo bikes. However e-cargo bikes can be effective for addressing highly localised demands, while also offering social and environmental benefits when compared to e-vans.

This work explores an emerging courier service business model prioritising the use of e-cargo bikes over e-vans where possible and performs a range of tasks types in addition to parcel collection and delivery. Namely the maintenance of shared micro-mobility (e-scooter) assets. The related tasks include reverse logistics tasks such as battery swaps, repairs, maintenance checks, asset recovery and redistribution. A main depot serves as a charging centre, repair centre and base for a fleet of e-cargo bikes and e-vans. The overall service can be modelled as a pickup and delivery problem with time windows and multiple task types. The e-cargo bike prioritisation aspect of the business is addressed within the optimisation by ignoring e-cargo bike rider wages.

This work introduces a heuristic routing optimisation algorithm that can easily model electric vehicle battery usage in detail in addition to time window and pickup and delivery type constraints. The heuristic is validated by finding the same solutions as an exact mixed integer programming formulation for small problem instances. The proposed e-cargo bike prioritising business model is contrasted with a range of variants to assess the social, economic and environmental trade-offs of runs such a courier service. The environmental benefits of using only e-cargo bikes are clear but at a financial cost. It is shown that a multi-modal fleet of e-vans and e-cargo bikes leads to the lowest cost routing solutions as it is possible to leverage the contrasted benefits of both types simultaneously. We also identify physical environmental scenarios where a simple e-cargo bike range constraint is insufficient for guaranteeing the feasibility of planned routes, such as when terrain becomes too rough or steep, or when parcels become too heavy. We also show scenarios where e-cargo bikes can travel further than their specified range.

ABSTRACT. Traditional conflict analysis methods, relying on the assumption of constant velocity, often fall short in capturing the dynamic nature of driver behavior randomness during the interaction process. Accurate trajectory forecasting is critical for many transport and logistics decision problems, such as risk-aware vehicle control, traffic flow management, and safety planning. To address the challenge of accounting for trajectory randomness in car-following scenarios, this study introduces a noise-enhanced generative adversarial network, named Car-Following GAN, designed for predicting collision trajectories based on data from the Shanghai Naturalistic Driving Study (SH-NDS). The model employs an encoder-decoder framework, integrating a noise enhancement module to capture the intrinsic randomness of driving patterns. Demonstrating notable robustness across varying environmental conditions, our model showcases adaptability for trajectory prediction in diverse driving scenarios. A conflict measure, termed the Rear-end Collision Risk Index based on Car-Following GAN (RCRIC), is proposed to quantify the risk of a rear-end collision. Our approach conducts a comprehensive case analysis to assess the impact of various traffic risk factors on RCRIC. The results underscore that our noise-enhanced approach significantly improves the trajectory prediction accuracy of the model when compared to other noise addition methods. This enhancement is observed across various prediction time windows and under different weather conditions. Moreover, RCRIC, derived from the model employing our noise-enhanced approach, effectively mirrors the dynamics of rear-end collision risk by explicitly incorporating trajectory randomness into its assessment. Furthermore, the findings underscore the significant influence of light conditions, traffic density, and weather conditions on driving risk.

ABSTRACT. Recent technological advancements have facilitated the coordination of automated vehicles into platoons—train-like formations that offer notable benefits, including reduced energy consumption. However, forming such platoons efficiently requires careful planning of both travel routes and charging schedules, particularly for long-haul electric commercial vehicles. In this study, we propose a polynomial-time heuristics to address the platoon formation planning problem with charging. The performance of the proposed approach is evaluated against a Mixed-Integer Linear Programming (MILP) model. Through numerical experiments, we show that the heuristic obtains close to optimal solutions for small to medium instances and is able to solve instances of up to 2,500 vehicles in just 12 seconds, whereas the MILP requires over an hour to solve instances with more than 100 vehicles. These findings underscore the potential of the proposed method for scalable and real-time deployment in large scale electric vehicle coordination.

ABSTRACT. Truck platooning uses an automated driving technique that allows multiple heavy-duty vehicles to travel safely in a line with close vehicle spacing and constant speed, so that the non-leading vehicles benefit from reduced aerodynamic drag, resulting in lower fuel consumption, better traffic flow, and reduced environmental impact. This paper focuses on the path-finding and coordination problem associated with heavy-duty vehicle platooning, also known as coordinated vehicle platooning problem. We propose a heuristic based on Dijkstra to solve the problem by adjusting edge weights according to temporal utilization patterns and generating path deviations stochastically according to vehicles directional alignment. For comparison, we present an exact solver solution based on the mathematical model formulation. The numerical experiments demonstrate the performance of our directional time-dependent Dijkstra on large instances. The results show that environmentally sustainable freight transportation with maximal fuel savings is possible.

ABSTRACT. The new generation of warehouses is gradually becoming robotized. Managers can select from many different competitive robotic techniques to store and retrieve loads and to fulfil the customer orders. In my talk, I discuss some very popular order picking systems involving robots, including cobot order picking systems, where people work closely together with robots to pick the orders in the warehouse. Using a mix of deterministic and stochastic quantitative modelling and empirical modelling, I show how questions can be answered like 1) How to measure the performance of such systems? 2) how to use these insights to create good-quality designs and select good policies? 3) How do human factors influence performance and how can we increase the well-being of workers in these systems?