ABSTRACT. This extended abstract shows a new algorithm for the construction of the root locus. The new algorithm is proven to be much more efficient in terms of execution time than the dedicated MATLAB function, while providing the same output results.

ABSTRACT. Mechatronic systems often present integral actions and time delay. Time delays may arise in motion systems due to several reasons, e.g., (i) the process to acquire the data from the sensors, (ii) the time needed to filter and analyse the data, or (iii) the actuator’s time to produce its control action. The Smith Predictor (SP) scheme is a common way to deal with time delay in control systems. The SP is a type of predictive control for stable processes that allows designing the controller of the system by neglecting the presence of the time delay. Nevertheless, the SP exhibits poor disturbance rejection capabilities, in particular when the system features an integral action (Sonzogni et al., 2024). Systems with an integral action are quite common in motion control, as the two-mass-spring-damping system (Wie and Bernstein, 1991), or when the system has to be controlled in velocity and the measured output is the system position. As a result, to cope with the negative effects of integral actions, different Modified Smith Predictors (MSP) have been developed. ˚Astr¨om et al. (1994) proposed a MSP to control a system modelled by an integrator and a time delay. The method is extended in (Zhang and Sun, 1996) to systems with an integrator times a constant. A drawback of both methods is that they do not consider the presence of poles or zeros with a value different from zero. In (Watanabe and Ito, 1981), a MSP for unstable systems is presented based on the knowledge of the statespace model. However, in motion systems the state space description is not always completely available as often the transfer functions are identified from data, focusing on the frequency bands of interest. In (Matausek and Micic, 1996, 1999), MSPs for controlling time-delayed systems with an integral action are presented, but they only considering systems without resonant poles. In (Normey- Rico and Camacho, 2007), a MSP approach based on the introduction of a new transfer function in the classical SP scheme is proposed. Nevertheless, this solution may lead to instability when the delay value is high. This work introduces a Modified Smith Predictor, based on the MSP in (Normey-Rico and Camacho, 2007), to effectively control systems with both delays and an integral action while ensuring closed-loop stability. The effectiveness of the approach is verified on an experimental motion setup.

ABSTRACT. The primary aim of this work is the fault detection of Petri Nets using observer design, when only partial observations are available. In the Petri Net framework, we have two types of sensors: place sensors (which measure the tokens contained in the places, nodes) and transition sensors (which measure the firing of transitions). We use the descriptor framework formalism to show the evolution of marking of Petri Net, and develop observer based fault detection methodology, to detect input faults.

ABSTRACT. In this abstract, we provide an overview of monviso (monotone variational inequalities solver), a novel Python package for solving monotone VIs. It acts as a wrapper around cvxpy [1] in order to simplify the process of implementing montone VIs. Its design is modular and simple to extend: the algorithms implemented in the current version, in fact, do not include all the ones existing in literature, allowing for simply adding (possibly new) methods. monviso, together with all the code snippets discussed in the following, is publicly available at github.com/nicomignoni/monviso.

ABSTRACT. In this paper, we address the problem of designing an aperiodic sampled-data controller stabilizing the zero-input equilibrium of an uncertain affine plant. The closed-loop system is modeled as a hybrid dynamical system incorporating a timer triggering the occurrence of the sampling events and two memory states storing the value of the controller state and controller output at each sampling time. Necessary and sufficient conditions on the controller parameters are given to establish the sought property. A constructive controller design algorithm based on sum-of-squares programming is given. A numerical example illustrates the effectiveness of the approach.

ABSTRACT. This paper addresses the problem of robust output regulation for exogenous signals modelled as Bézier curves. A regulation framework for compact Bézier curves is first developed, introducing a Bézier-based steady-state notion and a regulator built around an internal model composed of integrators. The framework is then extended to scenarios where the Bézier trajectory is reconfigured at runtime. We show that regulation performance degrades gracefully, with error depending on a generalised distance from the original trajectory. Crucially, the internal model retains useful information after such changes, unlike spline-based methods. This makes the approach well-suited for regulation problems in changing environments.

ABSTRACT. This work addresses the problem of stabilization of switched affine systems under dwell-time constraint, giving guarantees on the bound of the quadratic cost associated with the proposed state switching control law. Specifically, two switching rules are presented relying on the solution of differential Lyapunov equalities and Lyapunov-Metzler inequalities, from which the stability conditions are expressed. The first one allows to regulate the state of linear switched systems to zero, whereas the second one is designed for switched affine systems proving practical stability of the origin. In both cases, the determination of a guaranteed cost associated with each control strategy is shown. In the cases of linear and affine systems, the existence of the solution for the Lyapunov-Metzler condition is discussed, and guidelines for the selection of a solution ensuring suitable performance of the system evolution are provided. The theoretical results are finally assessed by means of three examples.

ABSTRACT. This contribution presents a recent theoretical result concerning Positive Real (PR) systems and their inverses. It is shown that, under standard assumptions, the inverse of a PR system is itself PR and shares the same characteristic values, derived from PR Riccati equations. These values enable structure-preserving model order reduction with guaranteed H-infinity error bounds. The invariance result implies that both direct and inverse systems can be reduced using a unified procedure, preserving symmetry and energy consistency. Theoretical findings are supported by numerical simulations on benchmark RLC networks, with further applications to electrochemical models. The result broadens the applicability of PR-based reduction techniques in control and energy-aware system design.

ABSTRACT. This abstract presents a distributed comprehensive framework enabling a networked multi-robot system to serve time-varying requests arising from different locations within the environment in a distributed and safe manner, i.e., by guaranteeing no collisions with possible obstacles and preserving connectivity among the robots. To this aim, a two-layer architecture is proposed where the top layer is in charge of assigning new service requests to the robots in a distributed manner, while the bottom layer is in charge of safely navigating the environment to serve the assigned requests. Special attention is given to addressing the limitations imposed by connectivity constraints, which might lead to deadlock situations where robots cannot reach the designated locations due to loss of network connectivity. Hence, a distributed strategy to detect and solve deadlocks in a distributed fashion is proposed. The completeness of the approach is proved. Simulation results in an agricultural setting and real-world laboratory experiments validate the effectiveness of the proposed approach.

ABSTRACT. The robotics research community is placing a major focus on the automation of difficult operations like the production of wiring harnesses, which require a high degree of dexterity and precision. In current industrial production lines, this operation is still manually done since it requires manipulating deformable linear objects (DLOs) like cables and wires. Thanks to advanced sensory systems, it is possible to develop robotic systems capable of autonomously executing such complex tasks or assisting human operators. This work introduces a robotized system for autonomous wiring harness assembly composed of a robotic arm equipped with tactile and proximity sensors for wire positioning and routing, and another one with a taping gun to secure wires. To control the robotic arms' movements, a modular control architecture has been developed, providing adaptability of the system to different product designs or operational conditions. Experimental evaluation of each component of the system is also provided, showing the effectiveness of the proposed solution in a real industrial scenario.

ABSTRACT. To successfully integrate circular economy principles into current value chains, it is essential to ensure the economic sustainability of disassembly processes. Disassembly is the first step in remanufacturing, repair and recycling of end-of-life (EoL) products. Recent scientific and technological advancements, including Large Language Models (LLMs) in artificial intelligence, are accelerating the development of robotic capabilities. A fully automated approach for the removal of clothing zips from garments is proposed. The approach detects the pixels corresponding to the zip using LLM-Based AI techniques to plan the path of the robotic arm. The solution reduces the effort of AI training in industrial applications.

ABSTRACT. A disturbance observer-based control strategy following a geometric approach is proposed. It is applied to the class of multirotor aerial vehicles with co-planar and collinear propellers, following the well-known inner–outer loop structure. This extended abstract builds upon a published work, to which the reader is referred for further details and proofs.

ABSTRACT. Artificial Intelligence (AI) has established itself as a pervasive driver of robotic autonomy, enabling machines to perceive, adapt, and make decisions in complex and evolving environments. However, pinpointing how and where AI integrates within the functional layers of robotic architectures remains a critical and unresolved question. This work addresses the challenge by combining large-scale bibliometric analysis of 26 years of scientific literature (2000-2025) with supervised multi-label classification, applying a custom control-oriented architectural framework tailored to Robotics and Autonomous Systems (RASs). The results reveal dominant research themes and structurally central functions, clarifying how AI supports core robotic behaviours, from perception to execution. This system-level mapping bridges thematic trends and control architectures, fostering a deeper understanding of AI’s role and promoting cross-disciplinary dialogue on intelligent robotics.

ABSTRACT. This study presents a novel approach based on the Hand Blink Reflex (HBR) to quantitatively estimate the perceived sense of control during human-robot interaction. In twenty participants, we measured HBR while a robotic arm entered their defensive peripersonal space under three control modes: autonomous, discrete, and continuous control. As expected, HBR amplitude increases as the robot neared the user, but crucially, increased human control confidence corresponded to reduced HBR amplitude. This method offers a quantitative index of human confidence in robot control, paving the way for safer, more effective collaborative interfaces and algorithms

ABSTRACT. This brief contribution aims to present some recent advancements in the study and development of max-plus algebra-related solutions to be applied to novel contexts, especially marine robotics. In particular, we focus on managing, planning, and modelling the coordination of a shoal of fish robots, also through a Graphical User Interface (GUI) developed to analyse the estimated mission times.

ABSTRACT. The Koopman operator framework allows to transform nonlinear dynamical systems into equivalent linear ones within a higher-dimensional state space. Its application can be extended to nonlinear optimal control problems, enabling their efficient solution in the Koopman lifted space. Here, we present a comprehensive analytical framework to lift general Nonlinear Model Predictive Control (NMPC) problems in the Koopman space, converting them into equivalent quadratic programs (QPs) - referred to as Koopman NMPC (K-NMPC) - that can be solved with superior computational performance. Moreover, we advance analytical Koopman operator methods by proposing an algorithmic procedure to generate an invariant basis of Koopman observables to lift both the nonlinear prediction model and the nonlinear state constraints of NMPC; additionally, we present a general method to arbitrarily reduce the dimensionality of the Koopman lifted space, lowering the K-NMPC complexity and handling the infinite-dimensional case. Our K-NMPC approach is validated through hardware-in-the-loop experiments on the case study of mobile robot navigation in cluttered environments, showcasing its solid performance and a ten-fold reduction in computation times.

ABSTRACT. Despite recent achievements, state-of-the-art autonomous agents can misbehave when training and environmental conditions differ – even minor mismatches can lead to undesirable behaviors or catas- trophic failures. Robustness towards these training/environment inconsistencies is a key challenge for the design of agents that can robustly interact with nonlinear, non-stationary and stochastic environments. In this context, we present DR-FREE – a Distributionally Robust Free Energy framework that instills robustness in autonomous agents by design. DR-FREE embeds robustness within its decision-making core through a robust reformulation of the variational free energy prin- ciple. The approach accounts for epistemic uncertainty by minimizing the worst-case free energy across an ambiguity set around the trained model. Experiments involving real robots are used to complement our theoretical findings. This is an extended abstract reporting key findings from the recent work [4].

ABSTRACT. Proximal gradient methods, such as the iterative shrinkage-thresholding algorithm (ISTA), are commonly employed for sparse optimization; however, they inherently yield biased solutions and require many iterations to converge. To address these issues, we propose designing an integral control of the algorithm's hyperparameter. We analyze the convergence of the proposed approach for strongly convex problems and extend the numerical results to the non-strongly convex framework.

ABSTRACT. One of the challenges in Model Predictive Control (MPC) is how to tune the parameters of the objective function so as to obtain the desired closed-loop behavior. In many practical cases, there may be no explicit function for characterizing system performance, further complicating the MPC tuning problem. This work tackles the problem of using machine learning to train the parameters of an MPC objective function from a dataset of human preferences over trajectory pairs. We learn a surrogate model that estimates the likelihood of the human preferring one trajectory over another. The surrogate model can be directly used as the objective function of an MPC controller, providing closed-loop trajectories that align with the expressed human preference. We show preliminary numerical results where a quadratic function is used to determine the preference.

ABSTRACT. In this work, we address the problem of designing stochastic model predictive control (SMPC) schemes for linear systems affected by unbounded disturbances. The contribution of the paper is rooted in a measured-state initialization strategy. First, due to the nonzero probability of violating chance-constraints in the case of unbounded noise, we introduce ellipsoidal- based probabilistic reachable sets and we include constraint relaxations to recover recursive feasibility conditioned to the measured state. Second, we prove that the solution of this novel SMPC scheme guarantees closed-loop chance constraints satisfaction under minimum relaxation. Last, we demonstrate that, in expectation, the need of relaxing the constraints vanishes over time, which leads the closed-loop trajectories steered towards the unconstrained LQR invariant region. This novel SMPC scheme is proven to satisfy the recursive feasibility conditioned to the state realization, and its superiority with respect to open-loop initialization schemes is shown through numerical examples.

ABSTRACT. Global optimization problems over a multi-agent network is addressed in this work. The objective function, possibly subject to global constraints, is not analytically known, but can only be evaluated at any query point. It is assumed that the cost function to be minimized is the sum of local cost functions, each of which can be evaluated by the associated agent only. The proposed algorithm asks the agents at each iteration first to fit a surrogate function to local samples, and subsequently to minimize, in a cooperative fashion, an acquisition function, in order to generate new samples to query. Experimental results on different problems demonstrate the effectiveness of the proposed method.

ABSTRACT. This work investigates the design of output-feedback schemes for systems described by a class of recurrent neural networks. We propose a procedure based on linear matrix inequalities for designing an observer and a static state-feedback controller. The algorithm leverages global and regional incremental input-to-state stability (incremental ISS) and enables the tracking of constant setpoints, ensuring robustness to disturbances and state estimation uncertainty. To address the potential limitations of regional incremental ISS, we introduce an alternative scheme in which the static law is replaced with a tube-based nonlinear model predictive controller (NMPC) that exploits regional incremental ISS properties. We show that these conditions enable the formulation of a robust NMPC law with guarantees of convergence and recursive feasibility, leading to an enlarged region of attraction. Theoretical results are validated through numerical simulations on the pH-neutralisation process benchmark, demonstrating the effectiveness of the proposed schemes.

ABSTRACT. This work contributes to ongoing research in sustainable mobility, with a focus on Battery Electric Vehicles (BEVs) and the challenge of extending driving range. Building on prior studies in Thermal Management System (TMS) optimization, it explores the use of neural network-based modeling to improve system representation and control performance. The proposed solution integrates data-driven techniques within a hierarchical Model Predictive Control (MPC) framework, enabling more efficient energy management and predictive thermal regulation. This combined approach supports the advancement of electric vehicle technology through improved operational efficiency.

ABSTRACT. This extended abstract shows an algorithm to address the capacitor voltages balancing problem in Multilevel Flying-Capacitor Converters (MFCC). The good performance of the algorithm is tested against the optimal solution given by dynamic programming.

ABSTRACT. We propose a novel approach to system identification using constrained optimization and feedback control of Lagrange multipliers. The method effectively circumvents vanishing and exploding gradient issues, improving accuracy and training time compared to standard methods. Moreover, we perform a theoretical analysis, study the algorithm’s complexity, and illustrate numerical examples.

ABSTRACT. Bacterial cellulose (BC) is an emerging biopolymer synthesized via microbial fermentation of eco-friendly substrates. When functionalized with ionic liquids (ILs) and coated with conductive polymers, BC becomes electroactive, enabling its use in wearable, biomedical, and environmental sensors. Despite its potential, predictive models for BC-IL devices remain underdeveloped. We propose a first-principles modeling framework coupling 2D finite element methods (FEM) for mechanical deformation with 1D FEM sub-models for ionic transport and voltage generation. This hybrid approach balances accuracy and computational efficiency. The model, calibrated with experiments, captures the spatio-temporal dynamics of dual-ion transport and electromechanical coupling. It reveals the effect of curvature and ion–mechanical phase shift on the sensor output. By including both cation and anion dynamics, it sets the foundation for the scalable design of sustainable smart sensors.

ABSTRACT. This contribution, presented at the European Control Conference 2025, proposes a discrete-time framework for parameter identification and state reconstruction of stochastic nonlinear systems with additive noise. Specifically, we build on the Physics-Informed Neural Network (PINN) framework and integrate a discrete-time Extended Kalman Filter (EKF) to manage stochastic noise. The EKF serves as the information model for the neural network, complemented by the likelihood function of the output innovation sequence to enhance accuracy of parameters estimation. Numerical simulations on the Chialvo oscillator demonstrate the effectiveness of the proposed approach in handling noisy and nonlinear system dynamics.

ABSTRACT. This extended abstract shows the Fast Modeling Power-Oriented Graphs, which is a step-by-step methodical procedure that guides the users through the automatic derivation of the power-oriented block schemes and, ultimately, the state-space models of physical systems starting from their schematics.

ABSTRACT. The transition to electric mobility is essential to meet the climate targets of the European Union, but poses critical challenges for the stability of the power grid due to the high and concentrated demand from electric vehicles (EV). This paper proposes a novel discrete event (DE) optimization model for the allocation and scheduling of EV charging, focusing on electric buses (EBs), which are characterized by large battery capacities and predictable schedules. The mixed-integer linear model introduces a detailed battery representation that captures nonlinear charging behavior and incorporates key constraints such as power limits, time windows, and charging station availability, assuming a flat electricity tariff. Compared to traditional discrete-time approaches, the DE formulation significantly reduces problem size while maintaining modeling accuracy and exact optimization. The proposed framework is validated using a real case study provided by IVECO S.p.A..

ABSTRACT. An approach for automated driving in highway scenarios based on Super-Twisting (STW) Sliding Mode Control (SMC) methodologies supported by the use of Artificial Potential Fields (APF) is presented. The use of APF allows us to propose an effective SMC solution based on the gradient tracking (GT) principle. In this regard, a novel formulation of the APF functions is introduced that exploits a sequence of attractive quadratic functions. This solution simplifies the computation of the fields and allows for trajectory generation with improved regularity properties. Extensive simulation tests, as well as comparisons with baseline and state-of-the-art solutions, show the effectiveness of the proposed approach.

ABSTRACT. The anti-lock braking system (ABS) prevents wheel lock during emergency braking by modulating the brake pressure at each wheel, helping maintain steering control. While mandatory in modern vehicles, ABS systems lack real-time road friction awareness, reducing their effectiveness on low-friction surfaces like ice or snow. Building on our recent work, we propose a friction-aware ABS that adapts online using road friction estimates. The system combines a new model-structured neural network (MS-NN) that learns the vehicle longitudinal dynamics across varying road surfaces, a robust friction estimation method based solely on standard vehicle sensors, and an online adaptation of the ABS parameters given the estimated road friction. Simulations demonstrate significant performance gains over a standard ABS, on diverse road conditions. Our implementation is open-source and available in a GitHub repository.

ABSTRACT. The rapid spread of networked multi-vehicle systems, particularly Autonomous Vehicles (AUs), in applications ranging from logistics to urban traffic management has exposed their significant vulnerability to cyber-attacks. The integrity and availability of communication channels are paramount for coordination, yet they present prime targets for malicious adversaries aiming to inject false data, disrupt operations, and compromise mission safety. This reality necessitates the development of advanced control architectures that are not only robust to disturbances but are inherently resilient to sophisticated cyber threats. This abstract summarizes the principal results of a research initiative designed to meet this challenge by creating a multi-faceted, layered defense strategy for networked autonomous systems.

ABSTRACT. This work deals with the real-time estimation of the lean angle of high-performance motorbikes. The estimate is obtained through measurements provided by an onboard inertial sensor and a GNSS receiver. A two-stage state observer, implementing a kinematic model developed under the novel assumption of coordinated manoeuvre, processes these measurements. A theoretical analysis demonstrates the observer's stability, the estimate's accuracy, and the estimation error bounds. Finally, experimental results obtained on race-track tests show the superior performance of the proposed estimator compared with other competitive approaches.

ABSTRACT. System identification exploits statistics and optimization to derive mathematical models of dynamical systems using input-output data. Machine Learning (ML) and Deep Learning (DL) have enabled data-driven identification of complex nonlinear systems without requiring knowledge of the structure or class of the nonlinear model to be identified. Among DL methods, autoencoders have shown promising performances for the identification of nonlinear state-space models by extracting latent representations. However, despite their modeling capabilities, no formal guarantees have been established regarding their robustness and reliability when employed in control applications. Adversarial learning consists of adding small input perturbations to degrade model performance and has been extensively studied in classification tasks, but remains largely unexplored for regression and system identification. In cyber-physical systems, adversaries may have access to and manipulate output and input sensor measurements, committing the so-called false data injection attacks, compromising predictions based on these models. The main contribution of this work is an algorithm to design an adversarial attack, based on adding a small perturbation to the input, that directly targets autoencoder-based models of nonlinear systems. This work highlights the sensitivity and vulnerability of these system identification methods to false data injection attacks, corroborated by numerical simulations to validate the performance.

ABSTRACT. Environmental safety and operational efficiency are critical challenges in modern port management, particularly in response to oil spills and hazardous sea conditions. This study presents two data-driven methodologies to support decision-making in maritime contexts. The first approach leverages a Conditional Deep Convolutional Generative Adversarial Network to simulate the spatiotemporal evolution of oil spills, using initial spill maps along with environmental data such as wind conditions and coastline features. The second methodology focuses on forecasting significant wave height using a hybrid deep learning model that integrates Convolutional Neural Networks and Long Short-Term Memory layers to process atmospheric inputs. Case studies conducted in the Mediterranean Sea demonstrate the effectiveness of both models in enhancing solutions.

ABSTRACT. In footwear manufacturing, robotic deburring presents a unique challenge due to high variability in product geometry and flexibility of materials. We propose a novel, fully self-supervised framework that enables robots to autonomously learn deburring trajectories directly from vision, without CAD models or human demonstration. This approach is inspired by human cognitive strategies, where vision and motor planning co-evolve. Our method comprises three key stages: (1) Learn to See, where visual networks extract a latent representation of relevant information of burrs from sole images; (2) Learn to Act, where, through reinforcement learning, an agent learns to map the latent features to 3D trajectories; and (3) Learn to Transfer, which uses Sim-to-Real methods including domain randomization and few-shot fine-tuning to ensure real-world applicability. We validate the framework in both simulation and physical experiments, achieving robust trajectory generation for diverse shoe sole geometries.

ABSTRACT. A key challenge in bridge monitoring systems lies in the accurate detection of sensor faults, which can severely compromise the quality of collected data and, consequently, the reliability of the monitoring outcomes. This study presents a novel data-driven approach for sensor fault detection in bridge monitoring applications, based on a combination of an encoder-decoder architecture and a Siamese network. The former aims to reconstruct the spectrograms computed from the accelerometer signals, whereas the latter aims to learn a similarity metric to better discriminate faulty readings from healthy ones. The experimental results of this study demonstrate the potential of the proposed approach in enhancing sensor fault detection performance compared to several baselines.

ABSTRACT. Reinforcement Learning (RL) has emerged as a promising approach for controlling complex dynamical systems without requiring an explicit model of the system. This research investigates the application of the Deep Deterministic Policy Gradient algorithm to control a Quadruple Tank Plant (QTP), a benchmark system characterized by strong subsystem interactions. The proposed DDPG-based controllers, both with a centralized RL agent and with a distributed multi-agent setup, are trained and evaluated in a simulated QTP environment, demonstrating their ability to achieve precise setpoint tracking, also in the case of partial knowledge of the system. Future works aim at testing the proposed architecture on a real QTP and tackle generalizability of the RL agents.

ABSTRACT. Microfluidic platforms have emerged as powerful tools for handling and analyzing small-volume biological samples in highly controlled environments. In the context of cell viability assessment, traditional methods often rely on chemical dyes or labels, such as trypan blue, that are invasive, time-consuming, and typically require end-point analysis, preventing sample reuse [1]. Moreover, these approaches often depend on bulky, high-cost laboratory setups like flow cytometers or fluorescence microscopes. This highlights the growing need for automatic, non-invasive and portable systems capable of evaluating cell viability on-chip, preserving sample integrity while reducing operational time and complexity. This research activity addresses the development and automation of a label-free system for on-a-chip cell viability assessment, integrating hydrodynamic actuation with optical monitoring in a microfluidic platform. The approach leverages the dynamics of cell motion under an oscillatory flow stimulus as a viability signature, specifically estimating the spatially averaged velocity profile of suspended cells [2]. The system design block scheme is reported in Figure 1. It uses a hydrodynamic oscillating signal with fixed frequency (f_i) and amplitude (A) as input and measures the average longitudinal velocity of cells in a microchannel as output. In this context, an automatic, label-free image-based system for cell viability monitoring on-a-chip has been developed [3]. A cyclically structured algorithm enables automated acquisition of video data, analysis, result visualization, and reinitialization for subsequent acquisition within predefined Time-Slot, thus allowing for continuous and automatic monitoring (see Figure 2(a)). For this purpose, high-speed scientific video acquisition was employed to capture cell motion and a customized Digital Particle Image Velocimetry-based algorithm (DPIV-based) was developed by the authors. This algorithm integrates the standard DPIV technique with an ad hoc-post-processing procedure, enabling the estimation of the average velocity profile and the extraction of process parameters in both time and frequency domains (see Figure 1, blue box). The identification of these parameters has proven to be reliable and effective for cell viability classification (see Figure 3(c)) using inert silica beads, viable and apoptotically induced yeast cells via hydrogen peroxide (H2O2), enabling time and concentration dependent cell death induction, prior to observable morphological changes. This approach has been further validated from a biological perspective using more complex cell types, including HL60 and HeLa cells, and benchmarked against standard viability assays, such as MTT, confirming its effectiveness and generalizability across different cell populations [4]. However, despite its accuracy, the DPIV-based’s computational intensity, requiring approximately 3 min to process 10 s of video, limits real-time applicability and constrains the range of feasible hydrodynamic oscillating input frequencies. To overcome this limitation, a deep learning-based model was developed to perform the video analysis and estimate the average longitudinal velocity of cells in a microchannel in real-time [5]. The ad-hoc designed Average Particles Velocity network (APVnet), comprising convolutional and recurrent layers, takes a short window of video frames as input and outputs an estimate of the cells’ average velocity profile. Its architecture is represented in Figure 2(b). The model was trained using supervised learning on a comprehensive dataset, consisting of a total of 21 acquired 60 s videos, experimentally collected by varying the biological condition in terms of micro-particles employed, the hydrodynamic stimulation parameters (f_i and A) set for the oscillatory flow and the optical setup. The model was labelled with DPIV-based derived ground truth, achieving high accuracy level (6% RMSE compared to DPIV-based, see Figure 3(a)-(b)), with real-time performance (3.4 ms latency) and a memory footprint of only 4.21 MB, enabling edge deployment. The performance of APVnet were also compared with MobileNetV2, a widely used architecture well-known for balancing accuracy and efficiency [6]. The outcome showed that APVnet outperforms MobileNetV2 in both predictive accuracy and computational speed, achieving deployment times four times faster than the benchmark. The proposed system represents a significant step toward automatic, real-time cell viability assessment on-a-chip. By integrating hydrodynamic stimulation, optical monitoring, and deep learning-based methodologies, it addresses the limitations of traditional viability assays and computational load inherent to image-based processing, eliminating the need for chemical labelling and complex instrumentation. To further enhance portability and automation, two alternative implementations strategies have been explored. Notably, first a reflective smartphone-based microscope was developed as a lens-less solution compatible with on-a-chip imaging. This design eliminates traditional optical lenses by leveraging the smartphone’s built-in camera sensor in an ad-hoc optical configuration, combined with an optical reflective surface (see Figure 4). Secondly, moving from on-chip imaging towards optical signal detection on customized and integrated chip, the Dual-Slit Particle Signal Velocimetry (DPSV) technique, which rely on optical signal acquisition rather than full-field imaging, was validated against the DPIV-based algorithm, yielding consistent, comparable results [7]. This approach is particularly promising for integration with micro-optical components, with significant efforts made in microfabrication [8,9]. These two alternative implementation strategies significantly contribute to reducing system size, cost, and complexity. Together, they demonstrate how consumer electronics and custom-integrated chips can support the development of compact and portable solutions suitable for autonomous, automatic liquid diagnostics and next-generation point-of-care devices.

ABSTRACT. This contribution summarizes the activities focused on within the framework of the collaboration between the University of Catania and ENI S.p.A. on the topic of control of magnetic confinement fusion plants. The activities cover a wide range of control-related topics, including the design and implementation of real-time reliable data acquisition systems and the development of novel paradigms for data analysis and identification for several plasma phenomena.

ABSTRACT. Airborne Wind Energy (AWE) is an emerging technology that aims to harvest wind energy at high altitude, using an aircraft connected via a tether to a ground station. While this technology holds great potential to support the clean energy transition, it also comes with important challenges, also on the modeling and control domain. The systems are inherently nonlinear, operate in uncertain and rapidly varying environments, and require tight coordination between the aircraft and the ground station. This is particularly true during takeoff and landing phases, which are the most critical phases of the systems' operation. This abstract presents recent results in the development of a Model Predictive Control (MPC) for the AWE system during these phases. Additionally we describe the development of an experimental setup aimed at testing such algorithms. This work contribute to closing the gap between simulation and field deployment, and underline the crucial role of advanced control in unlocking the full potential of AWE.

ABSTRACT. This work presents preliminary experimental results on a scaled floating offshore Airborne Wind Energy (AWE) system tested in a wave basin. The study aims to validate a numerical model of the floating platform and to assess the effectiveness of a control strategy designed to mitigate resonance effects. Results show that while the model captures the system’s dominant dynamics, it tends to underestimate motion amplitudes, likely due to hydrodynamic modeling uncertainties. The resonance-avoidance controller performs effectively, reducing platform oscillations as intended. These findings confirm the potential of the proposed approach and underscore the importance of further model refinement through system identification.

ABSTRACT. The problem of optimizing the power output of Fly-gen Airborne Wind Energy Systems (AWES), also called windplanes in this work, is considered. Windplanes harvest wind power with small wind turbines mounted onboard an autonomous tethered aircraft that carries out periodic trajectories (loops or figure-eight) roughly perpendicular to the wind flow (crosswind conditions). Electric power, generated onboard, is transferred to the ground via the tether. The most notable prototypes of Fly-gen systems were developed by Makani Power, which stopped operations in 2020. Their prototypes experienced large power fluctuations over the crosswind loop, often using the turbines as propellers in the ascending part. These fluctuations also imply oversizing the onboard electrical machines, which strongly impacts the overall system design. Recently, a new design methodology for windplanes has been proposed, resulting in a low aspect ratio wing and conventional, efficient airfoils. However, the resulting flight path presents a rather large power variability. In this abstract, which summarizes the main findings of a submitted paper, an optimal control approach and a multi-objective analysis are proposed to search for the Pareto front of flight trajectories with various trade-offs in terms of average power and its variability. This task is particularly challenging due to the system nonlinearities and non-convexity of the resulting optimization problems.

ABSTRACT. Predictive maintenance in marine propulsion systems plays a crucial role in ensuring safety, operational efficiency, cost saving and reduced emissions. This paper summarizes two innovative data-driven strategies applied to marine diesel engines. The first leverages Long Short-Term Memory (LSTM) networks for real-time, online prediction of component degradation using sensor data. The second proposes a novel catastrophic failure detection method that evaluates the second derivative of the error between measured and predicted variables, where predictions are obtained through a Decision Tree (DT). Both approaches were validated through realistic simulations in GT-Suite®, achieving early and reliable fault detection without requiring historical data. Results demonstrate the feasibility of online, flexible diagnostics for both slow degradation and sudden failures in maritime engines. we discuss about the possibility to integrate such necessary prediction methods into a Decision Support System.

ABSTRACT. In this paper we address the speed planning problem for a vehicle along a predefined path. A weighted sum of two conflicting objectives, energy consumption and travel time, is minimized. After deriving a non-convex mathematical model of the problem, we introduce a convex relaxation of the model and show that, after the application of a suitable feasibility-based bound tightening procedure, the convex relaxation shares the same optimal value and solution as the non-convex problem. We also establish that the feasible region of the non-convex problem is a lattice and use this property to derive a necessary and sufficient condition for the non-emptiness of the feasible region. Finally, we develop a dynamic programming approach which returns an approximate solution to the convex relaxation which converges to the optimal one as the discretization step decreases. To conclude, computational experiments highlight the performance of the proposed dynamic programming approach.

ABSTRACT. An important problem in motion planning is the computation of the speed profile along a predefined path, with the objective of minimizing travel time and energy consumption. This is a multi-objective optimization problem of clear industrial relevance. For control engineering students, addressing this problem requires the development of key skills. These include writing the dynamic equations of a road vehicles, formulating a multi-objective optimization problem, and using both convex and non-convex solvers for solving optimization problems. This paper presents a teaching unit focused on the speed planning problem, consisting of an initial lecture-based component delivered by the lecturer, followed by an interactive session supported by the use of a dedicated MATLAB-based application with a graphical user interface. This tool allows users to visualize instances of the speed planning problem and observe how solutions change as the problem parameters vary. Unlike traditional analytical approaches, this tool provides an interactive learning environment, where students can experiment with different scenarios in real time, gaining a more intuitive understanding of the trade-offs in multi-objective optimization.

ABSTRACT. A predictive and decentralized coordination framework is proposed for multi-AGV systems operating in realistic automated warehouses with bidirectional roads and complex layouts. The method combines a hierarchical environment model with an Event-Driven Predictive Control strategy over a Finite Time-Expanded Network. Conflict and deadlock resolution mechanisms ensure safe, congestion-aware navigation. Simulation results in real-world layouts show improved scalability, responsiveness, and mission efficiency compared to traditional solutions.

ABSTRACT. We present ARD, an artificial race driver capable of online time-optimal trajectory planning with adaptive driving style. ARD uses a nonlinear model predictive control (MPC) framework that includes an exit-speed maximization term and a novel terminal cost extrapolated from the previous solution. This design allows ARD to shift its behavior from early to late apexes in real time, without relying on precomputed minimum lap time solutions. Simulations on the Catalunya circuit show that ARD achieves lap times close to offline optimal control and faster than pure minimum-time MPC, while preserving real-time feasibility and enabling human-like maneuvers.

ABSTRACT. Despite the wide body of literature on motion planning for autonomous robots targeting structured agricultural environments, there remains a need for more efficient and reliable approaches for unstructured terrains. This work is motivated by a cooperation with the startup Paltech, which develops a weed-removal robot for grasslands. In this work, we employ a typical layered approach for robot navigation. However, in contrast to standard navigation tasks, such as point-to-point navigation, the global planner must solve the more challenging problem of visiting multiple targets in an optimal way while considering the kinematic constraints of the vehicle. Specifically, we propose as global planner a graph-based Dubins touring algorithm to find a globally near optimal solution. The goal is to define the shortest path that, given a set of target locations, ensures the robot visits each location once, while enforcing the kinematic constraints. We determine the order in which the waypoints should be visited by solving the travelling salesman problem with an Euclidean metric. Then, we design a graph-based solution for the Dubins touring problem to determine the optimal headings in the ordered sequence of targets. The local planner is based on a novel nonlinear model predictive control (NMPC) scheme designed for non-holonomic systems to guarantee convergence to a desired target without getting stuck in local minima. Instead of a classical path following, the NMPC algorithm takes the path generated in the first layer only as a reference, to then compute a locally optimal solution in real-time based on the current state of the vehicle, a more detailed dynamic model, and the perceived environment. The overall outcome of this work is a near-optimal motion planning algorithm that enables the UGV to visit points of interest in unstructured grassland and, in our specific application, perform autonomous weed treatment.

ABSTRACT. This paper presents a hierarchical planning architecture for multi-vehicle autonomous systems in dynamic industrial environments, built on a novel hybrid roadmap that integrates static and dynamic areas. The approach addresses a multi-objective optimization problem, balancing high-level criteria and generating feasible low-level paths. Experimental results demonstrate improved mission efficiency and reduced travel time variability.

ABSTRACT. The rising spread of Unmanned Aerial Vehicles (UAV) for applications in smart cities requires robust and scalable solutions for fleet management. This work presents a gerarchic approach for the optimization of multi-drone logistic operations in complex urban scenarios. The approach faces the problem of finding flyable paths and optimal target assignment for a fleet of UAVs operating in an urban environment. The considered architecture is distributed on two levels: (1) a high level mission planner based on Ant Colony Optimization (ACO), used to solve a Vehicle Routing Problem (VRP), assigning targets to the available drones, in order to minimize the overall energy consumption; (2) a geometric planner to generate continuous piecewise trajectories, composed of straight lines, clothoids and circular arcs, guaranteeing the feasibility for drones dynamics. The effectiveness of the proposed approach is proved by means of numerical simulations in realistic scenarios, considering the presence of obstacles and no-fly zones.

ABSTRACT. This abstract presents a comprehensive stochastic model for miniaturized interferometric fiber-optic gyroscopes (IFOGs) targeting next-generation Attitude Determination and Control Systems (ADCS). The model incorporates key noise sources—optical power fluctuations, thermal noise, and Kerr-effect phase shifts—enabling high-fidelity simulation of IFOG behavior. Experimental validation on a 500-meter coil prototype demonstrates strong agreement (<1% error in-phase, <10% quadrature) across the ±10°/s operational range. Allan deviation analysis confirms accurate prediction of bias instability (2.1×10⁻²°/h model vs. 2.2×10⁻²°/h measured) and angular random walk (3.2×10⁻³°/√h vs. 3.0×10⁻³°/√h), critical for spacecraft maneuver precision. The results validate the model's utility for optimizing miniaturized IFOG designs in small satellite applications.

ABSTRACT. We propose FATS-V, a lightweight RGB–IR fusion pipeline for robust obstacle and path detection in low-visibility agricultural environments. It combines agreement-based fusion, trust-aware interpolation, and a patch verifier to suppress false positives while maintaining real-time efficiency.

ABSTRACT. Extracellular DNA (exDNA) consists of fragments of genetic material released into the environment due to necrosis, cell lysis, apoptosis, natural realising, or decomposition, and is found in various ecosystems, including soils, marine sediments, aquatic environments, and culture systems. In addition to being a structural component of complex organic matrices, exDNA may play relevant functional roles by influencing organism interactions and local regulation mechanisms. Recent studies have shown that self-DNA, i.e., DNA originating from individuals of the same species, can exert a specie specific and dose-dependent inhibitory effect on the growth of a wide range of organisms, including bacteria, fungi, microalgae, protozoa, insects, and higher plants [1, 2, 3]. This effect has been observed under controlled conditions and is not reproduced by heterologous DNA, suggesting a mechanism of autoinhibition. In closed-cycle cultivation systems, such as hydroponic systems, microalgae photobioreactors, and microorganisms bioreactors, the progressive accumulation of exDNA may represent a critical factor: it could interfere with the vitality and productivity of the cultures and alter the associated microbiome composition. In such systems, due to decrease yield cycle or issues in cultures, nutrient solutions are periodically completely or partially replaced or diluted, generating loss of nutrients, wastewater that contains residual nutrients (e.g., nitrogen, phosphorus, and organic compounds) and genetic material. If not adequately treated, this effluent poses a potential sustainability and environmental attention by promoting eutrophication phenomena, as documented by recent studies reporting high levels of ammonium, nitrates, and phosphates in hydroponic wastewater [4], or potential issues on releasing genetic material in the effluent from wastewater treatment plants [5]. This condition justifies the development and analysis of selective removal strategies and engineering systems for the removal of extracellular DNA in biological closed systems cultures, evaluating their impact on the improvement of yield and the sustainable management of water resources. This PhD project, partially funded by the company No Self s.r.l. (https://www.noself.it), aims to design, build, and test a system for the selective removal of exDNA from culture fluids in closed systems. The objective is to mitigate the autoinhibitory effects associated with extracellular self-DNA, optimize crop productivity, efficiency and efficacy on using nutrients resources, reduce waste water production and improve sustainability cultivation processes.

References [1] Fabrizio Carten`ı et al. “Self-DNA inhibitory effects: Underlying mechanisms and ecological implications”. In: Plant Signaling amp; Behavior 11.4 (Mar. 2016), e1158381. ISSN: 1559-2324. [2] Stefano Mazzoleni et al. “Inhibitory and toxic effects of extracellular self-DNA in litter: a mechanism for negative plant–soil feedbacks?” In: New Phytologist 205.3 (Oct. 2014), pages 1195–1210. ISSN: 1469-8137. [3] Stefano Mazzoleni et al. “Inhibitory effects of extracellular self-DNA: a general biological process?” In: New Phytologist 206.1 (Jan. 2015), pages 127–132. ISSN: 1469-8137. [4] Jinkwan Son. “Evaluation of the Characteristics of Pollutant Discharge in Tomato Hydroponic Wastewater (HWW) for Sustainable Water Management in Korea”. In: Water 16.5 (Feb. 2024), page 720. ISSN: 2073-4441. [5] Markus Woegerbauer, Xavier Bellanger, and Christophe Merlin. “Cell-Free DNA: An Underestimated Source of Antibiotic Resistance Gene Dissemination at the Interface Between Human Activities and Downstream Environments in the Context of Wastewater Reuse”. In: Frontiers in Microbiology 11 (Apr. 2020). ISSN: 1664-302X.

ABSTRACT. Microgrids and interconnected power systems represent complex and nonlinear systems, whose operation is increasingly challenged by the high penetration of renewable energy sources (RES). The stochastic and uncertain nature of RES introduces significant difficulties in the control of key parameters such as voltage and frequency. Reinforcement learning (RL) has emerged as a promising approach for addressing these challenges due to its robust performance in stochastic nonlinear environments. However, ensuring the safety and stability of RL controllers in power systems remains a critical concern. This study presents a distributed control approach for voltage regulation using inverter-based distributed energy resources. The proposed distributed control strategy is based on the definition of a neural network with parameters optimized by an RL-based routine. By imposing independent Lipschitz-like conditions on each agent controller, local exponential stability of the microgrid voltage mismatch variable is guaranteed. Such Lipschitz conditions are constructed as the solution to a constrained optimization problem that aims to maximize the suitable space for the neural network parameters. Simulation results confirm the stability features and the effectiveness of the proposed approach.

ABSTRACT. Pin-point lunar landing with high precision and minimal fuel consumption is a critical challenge for future space exploration missions. This research focuses on the development and assessment of guidance and trajectory planning algorithms for Entry, Descent, and Landing (EDL) maneuvers. In particular, the landing phase of a lunar spacecraft. The problem is formulated as a multiphase optimal control problem, where each phase is characterized by different dynamics and constraints, and where the transition times between phases are themselves optimization variables. The current phase of the work consists of a comparative study of multiple offline optimization approaches, including a preliminary step using an NMPC algorithm, and both gradient-based methods (NLP) and derivative-free strategies such as Genetic Algorithms (GA) and Particle Swarm Optimization (PSO). Although quantitative results are still in progress, the goal is to evaluate the performance of each solver in terms of trajectory accuracy, fuel efficiency, and robustness. This comparative analysis serves as the foundation for the development of a novel algorithm aimed at real-time on-board implementation, where computational efficiency and scalability will be critical for success.

ABSTRACT. The widespread deployment of Automated Guided Vehicles (AGVs) in modern warehouses demands planning algo- rithms capable of coordinating large fleets in complex environments. Classical multi-agent path finding (MAPF) approaches typically rely on spatial and temporal discretization, leading to plans incompatible with continuous dynamics, while prioritized planners lack completeness guarantees in densely constrained scenarios. This extended abstract presents continuous-time priority-based search (CPBS) framework, a novel extension of the priority-based search (PBS) approach to continuous-time domains. CPBS incrementally constructs a binary search tree of partial agent priorities, dynamically refining the ordering only where conflicts arise. To handle continuous-time domains, each agent plans its trajectory using a modified safe interval path planning (SIPP) algorithm. Preliminary simulation results on topological maps derived from real-world case studies demonstrate the potential of the proposed methodology to provide efficient and reliable coordination of large AGV fleets in complex environments.

ABSTRACT. Attack detection is crucial in Internet of Things (IoT) networks to ensure system integrity and optimal performance. Existing methods struggle to handle large-scale data, missing values, and high variance. Furthermore, the lack of a unified infrastructure hampers coordination across detection processes. To address these challenges, we propose a three-phase attack detection framework. The first phase focuses on handling missing data, standardizing features, and reducing complexity by grouping similar attack classes. The second phase implements data balancing techniques to reduce bias due to class imbalances. Finally, the third phase utilizes machine learning (ML) models to improve detection accuracy and efficiency. This framework enhances classification performance and effectively manages complex attack scenarios. Experimental results on real-world IoT datasets demonstrate that the Histogram-Based Gradient Boosting (HGB) model outperforms existing ML and boosting approaches, making it promising for industrial IoT environments.

ABSTRACT. Trajectory optimization for robotic manipulators must balance smoothness, feasibility, and effective collision avoidance in cluttered environments. While classical methods such as Covariant Hamiltonian Optimization for Motion Planning (CHOMP) and Stochastic Trajectory Optimization for Motion Planning (STOMP) have proven effective, they typically rely on Signed Distance Fields (SDFs) to encode obstacle proximity, introducing geometric approximations, discontinuities, and computational overhead. This paper proposes a geometrically-informed extension of the STOMP framework, where both the robot and obstacles are modeled using differentiable primitives—cylindrical links and spherical obstacles. This formulation enables direct and accurate computation of collision costs, eliminating the need for precomputed SDFs or explicit gradient-based updates. By preserving the stochastic optimization backbone of STOMP, our method avoids the limitations of gradient descent-based approaches such as CHOMP, while maintaining robustness and scalability. The proposed approach results in smoother, collision-free trajectories with lower computational cost, making it suitable for high-dimensional manipulators operating in complex environments.

ABSTRACT. Agricultural robotics has seen a growing interest in lightweight, map-free navigation strategies capable of operating reliably in semi-structured environments. This paper presents and compares two LiDAR-based autonomous navigation strategies designed for ground rovers operating within orchard rows. Both approaches rely primarily on real-time 2D LiDAR data, avoiding the need for prior maps or external localization systems. We describe the core principles behind each method, including row-following and turning maneuvers. A comparative evaluation is carried out highlighting the strengths and limitations of each strategy in terms of accuracy and robustness.

ABSTRACT. This work introduces a novel decentralized framework for Multi-Robot Task Allocation (MRTA), designed to be application-agnostic and capable of addressing heterogeneous task requirements and agent capabilities. The proposed approach models the task allocation process as a Partially Observable Markov Decision Process (POMDP) and leverages Multi-Agent Reinforcement Learning (MARL) with Proximal Policy Optimization (PPO). Agents make decisions based on local observations of task cost, execution efficiency, and task resource requirements, while inter-agent communication enables dynamic strategy adjustment for improved collaboration. Two MARL training paradigms, Independent PPO (IPPO) and Multi-Agent PPO (MAPPO), are evaluated, with MAPPO showing superior performance in terms of learning efficiency, scalability, and robustness in stochastic environments. The results demonstrate the framework’s potential for real-time, distributed MRTA in complex, dynamic settings.

ABSTRACT. Optimal control theory is widely used to develop control strategies that minimize a given cost function while ensuring both system stability and performance. Within this framework, Riccati Equation-based optimal control has proved its remarkable effectiveness. Commonly non-linear optimal control involves the pointwise evaluation of system matrices and the repeated solution of the Riccati equation at each iteration step, computationally intensive and prone to numerical inaccuracies. A closed-form solution of the State-Dependent Riccati Equation (SDRE) for the class of second-order non-linear systems is proposed for significantly reducing computational efforts while offering analytical insights into the control strategy, providing a comprehensive exploration of the relevant conditions and properties.

ABSTRACT. Microalgae offer a dual solution to water pollution and CO₂ emissions by removing pollutants like nitrogen and phosphorus from wastewater while capturing CO₂ through photosynthesis. Raceway bioreactors are especially effective, yielding high biomass and enhancing both pollutant removal and carbon capture. However, their operation is sensitive to environmental fluctuations, requiring advanced control strategies. Our work employs model-based control to maintain optimal growth conditions, improving water treatment, CO₂ absorption, and biomass output with economic applications. Initially, we focused on pH control using Model Predictive Control, integrating feedforward actions based on disturbance forecasts. This required carefully designed experiments for nonlinear modeling to support accurate control. Now, we are developing a dynamic model for microalgae growth. Unlike static or overly complex models, our approach introduces time-dependent behavior via a linear filter, forming a Wiener model. This builds on our prior control efforts and enables dynamic real time optimization while having a simpler model easier to handle and calibrate.

ABSTRACT. This paper presents a multi-objective optimization framework for a thermally controlled room integrating an HVAC system in a school building. The problem is formulated as a constrained quadratic programming (QP) problem, aiming to minimize the total energy cost while ensuring the occupant's comfort. The results suggest that integrating thermal storage strategies with optimal HVAC control can reduce energy costs while maintaining comfort in which the effect of increasing thermal resistance shows a saturation in both energy consumption and cost. Future work includes real-time optimization, considering the model uncertainty data-driven methods will be integrated, using MPC. Moreover expanding the approach to multi-rooms and multi-buildings configurations.

ABSTRACT. Faults are the primary cause of economic losses, equipment damage, and blackouts in distribution networks. These faults are categorized into various types and induce rapid fluctuations in voltage and current signals. In this work, a machine learning-based fault detection method is considered. The proposed methodology effectively addresses the challenges of identifying fault types and locations in distribution power systems. By applying the Wavelet Packet Transform feature extraction method to superimposed three-phase voltage signals, the approach achieves high accuracy and robustness, even under noisy conditions and varying disturbances. The uncertainties associated with Renewable Energy Sources are considered, and the optimal locations of Monitoring Units are determined using a Voltage Stability Index-based optimization framework. Simulation results on a detailed IEEE 33-bus network validate the method's reliability, demonstrating its potential to enhance the efficiency and resilience of modern distribution networks.

ABSTRACT. Estimating Origin-Destination (OD) matrices is a fundamental problem in transportation planning, as they provide critical insights into travel demand and traffic flow distribution. Traditional methods rely on traffic surveys, vehicle tracking, and network tomography techniques, but these approaches often suffer from high costs, limited data availability, and significant estimation uncertainties. In this paper, we present a novel approach to OD estimation that leverages joint cumulants and bootstrapping techniques to improve the robustness of OD demand predictions. Unlike previous methodologies that rely on extensive prior information or require full statistical knowledge of network flows, our method operates under realistic constraints where only a subset of flow measurements is available. By estimating the covariance matrix of joint cumulants and applying a generalized least squares (GLS) approach, we systematically reduce estimation errors while ensuring computational efficiency. Simulation results on both synthetic and real-world datasets indicate that our method performs well in terms of accuracy, suggesting its potential usefulness for traffic management applications.

ABSTRACT. This work presents the design and development of an autonomous wheelchair equipped with a multi-layered control architecture. The system aims to achieve safe and efficient mobility for users in both structured and dynamic environments. The control architecture is divided into low-level and high-level components, enabling fine-grained motor control and robust motion planning in the presence of obstacles.

ABSTRACT. Synchronizing the joint operation of a group of Unmanned Aerial Vehicle (UAV) indoors is a challenging task due to the cluttered environment and dynamic obstacles, and the resulting instability of control signals and localization sensors. In this paper, the navigation accuracy of two UAV is investigated using the DJI Tello as an example during synchronized flight. In addition, the tracking problem is studied taking into account the delays that may occur in the system, since this represents a more consistent approximation to the real control scenario. A series of experiments are conducted in which different options for synchronous flight control are tested and compared while performing various aerial maneuvers. The results show that the proposed architecture options affect the navigation accuracy of UAV indoors by taking into account the shared use of vision, localization using the internal Inertial Measuring Unit (IMU) sensors, and Ultra Wide Band (UWB).

ABSTRACT. The cutting-edge contact-aware ability of aerial platforms has opened new frontiers in aerial robotics, enabling applications beyond traditional contact-free operations. Performing in-contact tasks and smoothly transitioning between navigation and interaction phases introduce significant challenges, especially in complex scenarios. This paper focuses on multi-rotors designed for physical interaction tasks, reviewing the most used aerial platforms, interaction tools, and control methodologies tailored for contact-aware applications. Special attention is given to the contact detection phase, which bridges the gap between contact-free and in-contact operational phases, ensuring precise and safe engagement with the target.

ABSTRACT. For control-affine nonlinear systems, we augment a predefined Lyapunov-based global stabilizer with a hybrid obstacle avoidance design preserving the Lyapunov decrease. While the method can be applied to the general class of control-affine systems, the size of the avoidance neighborhood is not a design parameter. Our design shows that a system can achieve global asymptotic stability with simultaneous unsafe set avoidance via hybrid feedback, which overcomes well-known issues of topological obstructions

ABSTRACT. High-precision reference tracking for Unmanned Aerial Vehicles (UAVs) is one of the main challenges in the aerospace control community, primarily due to aerodynamics effects. These effects are difficult to model accurately and, therefore, are often neglected. As a result, model-based control methods frequently fail to achieve the desired performance. With the growing popularity of machine learning techniques, a promising research direction is the integration of model-based control with data-driven approaches. The goal of this paper is to design a learning-based controller for a simplified quadrotor system and validate it in a real-world platform. By exploiting the property of differential flatness, a Feedback Linearization controller is used to linearize the system, while a Bayesian inference method combined with a recent sparsification technique, called Recursive Gaussian Process (RGP), is employed to compensate for residual nonlinear dynamics. Simulation results show that the proposed approach is effective in compensating for the non-linear mismatch, improving the tracking performance. Additionally, the controller is validated on a real 1D quadrotor platform, showing promising results.

ABSTRACT. A new framework, called Distributed Unified Architecture (DUA), has been introduced to tackle a key challenge in multi-agent robotic systems: achieving smooth coordination among heterogeneous platforms. As collaborative robotic construction becomes more prevalent, often involving agents with differing functionalities and capacities, the supporting software infrastructure must ensure effective interoperability without sacrificing performance. DUA addresses this need through a containerization-based approach that delivers uniform execution environments across various hardware systems, while still retaining each platform’s unique features and optimizations.

ABSTRACT. We introduce an ongoing research on hybrid controller design exploiting switchings and state resets to maximize a quadratic performance index that relates to the $\mathcal{L}_2$ norm of the tracking error and its derivatives. The proposed strategy hinges upon switches and resets of different element of the feedback control loop elements such as bumpless transfer, additive filters for signal regularization, reference re-definition and switches among of a set of controllers in order to increase the aforementioned performance index.

ABSTRACT. Autonomous UAVs face significant challenges when attempting to land on sloped or uneven terrain, especially in demanding scenarios like disaster response or hazardous environments. In this work, we propose a lightweight, inertially balanced landing mechanism that enables safe landings on steep inclines through robust terrain characterization. Our design ensures stability and maneuverability while maintaining a minimal structural footprint. Future developments aim to further enhance the platform versatility and performance in complex real-world missions.

ABSTRACT. A cooperative Range-Only SLAM algorithm based on a Multi-Hypotheses Extended Kalman filter is presented. An efficient solution is obtained by stacking all hypotheses in the same filter. A Federated Information Sharing mechanism is introduced to distribute the information coming from range measurements among the different hypotheses. This approach, already considered in the literature for single robot applications, is here extended to the case of a team of robots, obtaining a significant improvement with respect to the single robot solution.

ABSTRACT. This work introduces an Digital Twin system for simulating the spatiotemporal dynamics of surface wildfire spread across vegetated landscapes. At its core lies the Fire Engine, which is implemented through an innovative modeling approach that combines the continuous Rothermel’s model with the discrete Cellular Automata (CA) framework. Unlike most existing simulators in the literature, this Digital Twin requires only a limited set of inputs for a user defined area of interest: four satellite-derived products (NDVI, NDWI, LCT and DTM), combined with local wind and temperature and humidity measurements. Based on these minimal inputs, the simulator produces a dynamic map that visualizes the real-time evolution of surface wildfire, starting from one or multiple ignition points. The simulator is computationally lightweight, easy to deploy on standard hardware, and suitable for both real-time forecasting and offline scenario analysis. This makes it a valuable tool to support decision-making in wildfire risk assessment, early warning systems, and emergency response planning.

ABSTRACT. This work introduces the development of fully integrated, soft, and fiber-shaped artificial muscles capable of both actuation and sensing. By combining thin McKibben actuators with ElectroHydroDynamic (EHD) fiber pumps, we realize ElectroFluidic Fiber Muscles that generate mechanical motion through electrical input alone, eliminating the need for bulky external pumps. To enable control without compromising compliance, we integrate stretchable, fiber-shaped sensors in parallel with the actuator bundle, inspired by the layout of biological muscle spindles. This configuration maintains the slender, modular form factor of the bundle while enabling position estimation. Demonstrated on a robotic arm, the system showcases the potential for closed-loop control in soft and wearable robotic applications.

ABSTRACT. Electroadhesive grippers are well suited for handling fragile or irregular objects, but their effectiveness strongly depends on the local curvature of the contact surface. This work introduces a novel grasp planning pipeline that incorporates a curvature-based contact model to estimate the voltage required for stable electroadhesion. The model enables direct evaluation of grasp feasibility based on local geometric data and is integrated into a real-time perception system that processes partial point clouds of unknown objects. Without requiring prior object knowledge or training, the system accurately identifies physically valid grasp regions, making it compatible with both analytical and learning-based planning frameworks.

ABSTRACT. This paper addresses the problem of current sharing in multi-input converters (MICs) for DC microgrids, where multiple sources must be coordinated while ensuring stable bus voltage. In particular, MIC includes several voltage sources connected to a common DC bus with a bulk capacitor, through several parallel synchronous boost converters, and an aggregated load modelled as an ideal current source connected to the DC bus. The dynamics of current distribution can be controlled without impacting the voltage regulation. Conventional methods often rely on time-scale separation between voltage and current dynamics, which limits performance and slows down current distribution.

We propose a dynamic allocation control strategy that decouples bus voltage regulation from current sharing without requiring time-scale separation. The method combines a min-type control law (to guarantee bus voltage regulation and system stability) with a dynamic allocation scheme (to optimally distribute currents according to source limits and efficiency without impacting the voltage). 

Dynamic allocation exploits the input redundancy of MICs by solving an online optimization problem that enforces fair or optimal current sharing, while keeping the bus voltage unaffected. This decoupling enables faster and more reliable operation compared to conventional methods.

The proposed approach provides: 

• fast voltage response;
• independent and optimal current distribution;
• robustness to load variations and parameter uncertainties.

The proposed solution has been experimentally tested on a suitably devised test setup to validate its effectiveness under different operating conditions. 

ABSTRACT. Critical infrastructures, such as power and transportation, are frequently modeled as interconnected Cyber-Physical Systems (CPS). This interconnectivity makes CPS vulnerable to cyber-attacks, threatening system stability and operational continuity. Notably, many of these \textcolor{black}{malicious attacks involve manipulating} sensor measurements. In this context, accurately reconstructing the attack signal is crucial for recovering true measurements, distinguishing cyber-attacks from faults, and enabling effective countermeasures. This work presents a distributed scheme that leverages sliding mode observers to reconstruct attack signals targeting sensors in heterogeneous linear cyber-physical networks. The proposed approach achieves attack reconstruction with suitable bounds on the reconstruction error that depend on the system disturbance characteristics. Additionally, we present an isolation mechanism that utilizes the developed reconstruction scheme and analytically show that it accurately determines the attacked nodes

ABSTRACT. Economic Nonlinear Model Predictive Control (E-NMPC) is a peculiar variant of classic NMPC, which directly includes an economic criterion within its stage cost function, allowing to steer the system under control towards an economically optimal equilibrium and ensuring a profitable economic performance during the transient. While its formulation provides significant advantages for many practical applications, E-NMPC suffers from two main drawbacks: first, it only accounts for the economic objective, disregarding additional conflicting tasks, such as tracking; second, it forfeits classical Lyapunov-based stability guarantees, since, in general, its stage cost is non-minimal at the optimal economic equilibrium. To address these limitations, we present a novel E-NMPC formulation, accounting for both economic and tracking tasks together. Additionally, we propose a general constructive procedure to design suitable stabilizing terms for E-NMPC, ensuring its closed-loop stability with minimal impact on the economic performance. Our E-NMPC approach is validated on two different case studies: energy-efficient Adaptive Cruise Control (ACC) in electric vehicles, and mitigation of epidemic spreading on networks via vaccination and non-pharmaceutical interventions (NPIs).

ABSTRACT. Model Predictive Control relies on the availability of an accurate model of the process to be controlled. In practical scenarios involving biological systems, deriving a model based on physical equations can be challenging. This work proposes a neural MPC controller that learns an approximation of system dynamics from data, enabling constraint-aware control without prior model knowledge. The proposed methodology is applied to a model for cancer control through chemotherapy and immunotherapy in presence of clinical constraints. In-silico numerical simulations are performed, exhibiting performances comparable to the case of standard MPC with complete model knowledge and to the case of Optimal Control using Pontryagin's Maximum Principle.

ABSTRACT. The growing diffusion of Autonomous Vehicles (AVs) is reshaping urban mobility by enabling new forms of intelligent transportation, such as shared and on-demand services. Among these, carpooling with AVs represents a promising solution to reduce traffic congestion, emissions, and operational costs. How- ever, optimizing shared rides in real-time urban environments remains a complex challenge due to dynamic traffic conditions and the need for efficient routing. This paper presents a hierarchical framework for carpooling management based on two strategies: i) the Deep Reinforcement Learning (DRL) to optimize the sequence of pick-ups and drop- off operations performed by AVs, ii) a rerouting algorithm for selecting the best route considering traffic conditions. To this aim, the carpooling problem is formalized using the Markov Decision Process framework. The approach is implemented using the Proximal Policy Optimization algorithm and validated through simulations conducted by the Simulation of Urban Mobility tool. As a case study, the cities of Bari (Italy) and Tampere (Finland) are analyzed to assess the model performance in real urban scenarios. The results demonstrate the potential of the proposed method to improve route efficiency, reduce travel time and emissions, and support the development of intelligent and sustainable mobility systems.