Autonomous search and rescue (SaR) requires an agent to explore an unknown, partially observable and hazardous environment quickly without harming itself or the victims it is looking for. Reinforcement learning (RL) produces adaptive policies but offers no safety guarantee, while model-predictive control (MPC) enforces model-based safety constraints but is conservative and slow. We combine the two, shielding a learned policy under an MPC backup and ask when the result is preferable to either on a partially observable grid-world SaR task. We first build a pure-RL explorer and show its coverage hits a ceiling near 70% success, a limit that is due to the method rather than to tuning. We then wrap it in a slip-aware safety shield that certifies every action's reachable "slip tube" against the agent's belief and defers to a model-predictive backup when the policy proposes an unsafe move. Across 250 held-out hazardous maps, the shield makes pure RL deployable by eliminating its collisions in evaluation, reducing it from 117.7 to 0 and cutting damage around 15x while still retaining 93% of its success (from 0.43 to 0.40), at 53 ms per decision and beating the slower tube-based MPC on coverage, victims and speed. Only the exhaustive MPC is more effective, but it runs at approximately 5x the latency and does not scale. We thus position the shield not as a universal winner, but as the best choice when collisions cannot be tolerated and decisions must stay fast. ... Read more

Autonomous robots increasingly support Search and Rescue (SaR), yet victim trust in these robots remains largely unaddressed in simulation, where robots typically neither estimate nor respond to it. This paper implements a transparent, bidirectional trust model in a SaR simulation with cognitively realistic victims based on the GT-BDI framework. Victims maintain a true internal trust state driven by observable signals–proximity and perceived danger–while the robot estimates this trust from proximity and switches between approaching the victim to build trust and leading it to the exit. An ablation study over 100 matched seeds compared trust-aware against trust-blind robots. Trust-aware behavior saved significantly more victims (+0.74 on average, p < 0.001), reducing victims stranded at timeout rather than lowering mortality. This gain stemmed not from higher trust, but from better robot positioning. These results suggest that how a robot responds to trust can meaningfully improve evacuation outcomes, offering a practical and interpretable path to better robot-guided rescue. ... Read more

Robots are increasingly used in search-and-rescue (SaR) to perform dangerous tasks in place of humans, and to free human rescue workers to focus on other tasks. These robots require effective and reliable control methods to ensure that they perform as well as the humans they replace. Model predictive control (MPC) is an optimisation-based method which innately handles hard constraints such as not colliding with victims, and has been well-studied in SaR contexts. Tube-based MPC (TMPC) is an extension of MPC which is robust to the bounded disturbances inevitable in real-life SaR, but due to its fixed tube is often overly conservative. Surma and Jamshidnejad propose state-dependant dynamic TMPC (SDD-TMPC), which adjusts tube size online based on an estimate of the current maximum disturbance, allowing for reduced conservativeness when possible disturbances are small. We develop a grid-based SaR simulation to evaluate how SDD-TMPC's adaptive tube affects conservativeness and the safety-performance trade-off in a dynamic, partially observable environment. SDD-TMPC, MPC, and TMPC are compared over many simulated search missions, where the agent is tasked with finding all victims in an environment. The results show that while retaining the empirical safety of TMPC, SDD-TMPC behaves less conservatively and outperforms TMPC in mission feasibility and duration. ... Read more

Effective autonomous search and rescue (SaR) requires safe and efficient traversal of hazardous regions. While model predictive control (MPC) offers a solid framework for controlling an autonomous agent, the robust tube MPC (TMPC) approach sacrifices too much performance, especially where disturbance probabilities vary spatially. Current approaches to address this problem explore varying tube size in scenarios with perfect information, but none test whether local perception of the environment can be used to dynamically scale tube size to reduce conservativeness adaptively. A state-dependent dynamic TMPC (SDD-TMPC) formulation is proposed, scaling tube dynamically on perceived disturbance risk and confidence. Three controllers: MPC, TMPC, SDD-TMPC, are tested over 250 seeded SaR simulations, with additional runs introducing noise into perceived disturbance risk to assess the formulation's robustness. SDD-TMPC is shown to substantially outperform TMPC (odds-ratio (OR)=40.0) while maintaining 0 wall/victim collisions, operating in TMPC mode only 32.3\% of the time. Robustness and performance advantage are retained under perceived noise of magnitudes up to 0.2, however, 0.4 is identified as the boundary where this no longer holds. Overall, this suggests that perception-based tube scaling is a promising direction for autonomous agents. ... Read more

Task allocation in multi-robot Search and Rescue (SaR) is challenging: the environment is unknown, time-critical, and too dynamic for continuous centralized planning. This paper asks how a lightweight supervisory framework can dynamically allocate sub-tasks within a multi-robot team to balance exploration and victim retrieval without incurring prohibitive computational costs. We propose an event-driven, auction-based supervisory layer that intervenes only on discrete triggers; spatial conflicts, agent stagnation, and victim discovery, leaving robots to operate autonomously otherwise. Compared to a purely decentralized baseline across 216 simulated missions, the supervisor raises the victim discovery success rate from 50.5% to 91.7%, locates all victims roughly 24% faster without meaningfully increasing computation time. When extended to a heterogeneous fleet including a dedicated rescue agent, these exploration gains persist, but the increased rate of victim discovery outpaces single-agent rescue capacity, revealing a trade-off between exploration efficiency and victim retrieval throughput that the current allocation mechanism does not yet resolve. ... Read more

Trust plays a central role in human–robot interaction, as it shapes how humans rely on and collaborate with autonomous systems under uncertainty. Existing approaches often treat trust as a static value or rely on highly data-driven models that lack interpretability and require extensive amounts of context-specific datasets. This limits their applicability in simulation and control. Human trust is modeled as a latent, time-varying inter nal state and is formulated as an end-to-end identification problem grounded in empirical interaction data. A modular, deterministic dynamic model integrates dispositional, situational, and learned components within a discrete-time framework, capturing both stable predispositions and interaction-driven adaptation. A human subject experimental paradigm including 35 participants elicits informative trust dynamics and enables pa rameter identification and evaluation in terms of fit, generalization, and robustness. The resulting model remains competitive against validated baselines in reproducing observed trust trajectories while maintaining interpretability, modularity, and practical identifiability. Closed-loop coupling with probabilistic decision models examines the relationship between inferred trust and observable behavior, showing that trust is behaviorally relevant but insufficient on its own to explain human actions. The resulting framework enables executable, empirically grounded modeling of trust dynamics in human–robot interaction and supports the development of trust-aware simulation environments and human-centered robotic systems. ... Read more

The growth and aging of the world population is causing an increase in the number of vulnerable individuals who require continuous support. However, health and social care systems do not possess enough human and material resources to address this growing demand. Social robots are a potential solution to help relieve this pressure by providing affordable, personalized, and continuous support to vulnerable individuals that can complement existing health and social care systems.

Nevertheless, state-of-the-art social robots struggle to sustain user engagement, especially in the long term, which ultimately limits their usability. Importantly, it is precisely through long-term interactions that social robots have the most potential to systematically tackle the overload in care systems. Therefore, enabling social robots to sustain user engagement is of utmost importance to facilitate their real-world deployment and societal impact.

The ability to sustain user engagement is strongly linked with the behaviors displayed by social robots and, consequently, with the approaches used to control said behavior. In state-of-the-art social robotics, most behavioral control approaches rely on either rule-based systems or machine-learning approaches, where both paradigms struggle to generate adaptive and personalized robot behavior. To sustain user engagement, social robots must replicate the responsiveness and adaptability that characterize human-to-human interactions, including being sensitive to others’ mental states. However, behavioral controllers of social robots do not typically model the dynamic mental states of users, nor do they adapt robot behavior to those states.

Therefore, this thesis addresses this crucial limitation by presenting a behavioral control framework for social robots that models the dynamics of user mental states and systematically generates robot behavior according to these dynamics. To do so, the proposed framework is grounded in systems and control-theoretic methods, which enable transparent and systematic decision-making.

The first step toward such a behavioral control framework is presented in Chapter 2, where a dynamic mathematical model of human perception, cognition, and decision-making is developed based on Theory of Mind. This model represents the dynamic evolution of beliefs, goals, and emotions as a state-space formulation, enabling systematic integration of the model with systems and control-theoretic approaches that can be used to steer the behavior of social robots. The model’s capability to estimate the mental states and predict the actions of users was evaluated in an experiment with human volunteers.

Building on the dynamic user model introduced in Chapter 2, Chapter 3 presents a predictive, model-based behavioral control framework for social robots. The model is incorporated into a closed-loop, constrained multi-objective optimization problem that systematically selects robot actions to steer the interaction toward desired user mental states. This chapter constitutes the first integration of systems and control-theoretic methods for behavioral control of social robots, enabling transparent, adaptive, and cognition-aware decision-making. The proposed controller was validated in an experiment with human participants interacting with a humanoid robot, demonstrating improved engagement and sustained interaction quality.

Since the performance of model-based controllers depends on accurate estimation of the user’s mental states, Chapter 4 presents three model-based state estimators for social robots that embed the dynamic model presented in Chapter 2. These estimators explicitly address the sparse and subjective nature of measurements in human-social robot interaction by personalizing process and measurement noise covariance matrices and proposing three tailored Extended Kalman Filter variants. The methods were validated using data from an experiment involving human volunteers interacting with a humanoid robot, demonstrating reduced cumulative prediction errors and improved robustness of mental state tracking over time.

To position the proposed behavioral control framework within the field of social robotics, Chapter 5 presents a systematic review of control approaches for behavior generation in social robots. The survey identifies dominant algorithmic paradigms and analyzes their underlying assumptions, strengths and limitations, and capacity to support adaptive and socially appropriate behavior in long-term human-social robot interactions. Building on this review, the chapter then synthesizes cross-category insights, evaluates control approaches by interaction length, system role, and application, and outlines emerging research directions toward more adaptive, interpretable, and cognitively grounded behavioral control frameworks. By clarifying existing strengths and gaps, this chapter motivates the need for transparent, model-based, systems-theoretic behavioral control architectures.

Together, these contributions establish a coherent, theoretically grounded behavioral control framework that bridges cognitive modeling, state estimation, and behavioral control for social robots. This thesis demonstrates that treating human-social robot interaction as a dynamic control problem, where mental states are estimated and steered over time, is both feasible and effective. The proposed framework, therefore, offers a concrete path to deploying social robots that can meaningfully support individuals across sustained, real-world interactions.
... Read more

Model Predictive Control (MPC) is an advanced control method that utilizes a model to make predictions about the state evolution for the controlled system, as well as an optimization solver to compute an optimal future control trajectory that satisfies state and control constraints. Inaccuracies in model predictions, caused by modelling errors or external disturbances, significantly hinder the performance of MPC and potentially lead to constraint violations. To address these challenges, Robust MPC (RMPC) explicitly accounts for uncertainties in the model, enhancing the control reliability. Tube-based MPC (TMPC) is a well-known RMPC method, where a nominal MPC generates a reference state trajectory, a tube (i.e., a trajectory of model uncertainties centered around the reference trajectory), and an ancillary control law that ensures the actual states of the system remain within this tube —following the reference trajectory, despite uncertainties.

This thesis presents a critical analysis of current state-of-the-art MPC methods, and proposes novel theoretical developments, architectures, and extensions for effective and computationally efficient handling of uncertainties via MPC . These contributions are rigorously supported by formal proofs. Furthermore, the proposed control frameworks are systematically evaluated through dedicated computer-based simulations, with comparisons drawn against existing methods in terms of optimality, robustness, and computational complexity.

The thesis begins with introducing State-Dependent Dynamic Tube-based MPC (SDD-TMPC), an extension of TMPC designed to more effectively handle the variability of model uncertainties and environmental disturbances. By leveraging available information about state-dependent uncertainties, SDD-TMPC enhances optimality and reduces risks of infeasibility, while maintaining the same level of robustness as TMPC. Although SDD-TMPC demonstrates applicability to systems with varying uncertainties across the state space, its practical implementation is limited by high computational demands.

To mitigate this limitation, Approximate State-Dependent Dynamic Tube-based MPC (ASDDTMPC) is developed. This approach employs Spiking Neural Network (SNN) to approximate the behavior of SDD-TMPC. SNNs were selected for their event-driven processing and biologically inspired efficiency, offering significant advantages for low-power, real-time control. Recognizing that this approximation introduces additional uncertainties, and thus the risk of insufficient robustness to uncertainties, the SDD-TMPC framework is extended to incorporate these approximation errors as additional state-dependent disturbances, thereby preserving robustness. The reduced computational requirements of spiking neural networks enable implementation on resource-constrained platforms, such as small-scale robotic platforms.

Next, the Parent-Child MPC (PC-MPC) architecture is proposed to further reduce computational complexity across a wide range of MPC frameworks. Compatible with both tube-based and deterministic MPC, the PC-MPC architecture decomposes the optimization problem into two linked problems: The Parent MPC (P-MPC) addresses long-term stability and constraint satisfaction, while the Child MPC (C-MPC) focuses on short-term stability and disturbance rejection. P-MPC communicates additional constraints to C-MPC, which determines and executes control strategies. This hierarchical approach, which guarantees robustness and stability, is extendable to systems with complex dynamics and large scales. This is done by incorporating additional Parent layers to further manage computational complexity in such systems.

While robustness is critical, there are environments where maintaining strict constraint satisfaction is infeasible due to the nature and extent of uncertainties. To address this, a new theoretical framework called Fuzzy-Logic-based MPC (FLMPC) is developed, particularly suited for controlling multi-agent systems with imperfect environmental perception operating in unknown environments. FLMPC uses fuzzy vectors to model uncertainties, where each element—being a fuzzy variable—represents the degree to which a region exhibits properties such as “dangerous” or “certain”. Fuzzy maps are constructed by grouping fuzzy vectors. FLMPC performs fuzzy optimization to compute optimal trajectories for all agents, demonstrating superior performance and reduced computational complexity compared to state-of-the-art control methods. This efficiency is achieved by enabling the execution of computationally intensive tasks of fuzzy map generation outside the real-time optimization loop. This is made possible by inheriting the fundamental strength of fuzzy logic, particularly its ability to handle uncertainties over a continuum of values, rather than discrete thresholds. This allows for reliable decision-making based on fuzzy maps within a flexible computational window, rather than being restricted to a specific time step.

The inherent limitation of finite prediction horizons in MPC poses a challenge for exploring tasks in large-scale environments. To improve scalability, a bi-level FLMPC framework—leveraging the PC-MPC architecture in the context of FLMPC — is introduced, with potential for extension to multi-level hierarchies. The Parent Fuzzy-Logic-based MPC (P-FLMPC) formulates a global plan using comprehensive environmental knowledge, while the Child Fuzzy-Logic-based MPC (C-FLMPC) focuses on local enhancement and execution of the plan, retaining flexibility for real-time adaptation.
... Read more

This PhD thesis addresses the optimal control and AI-based control of Search-and-Rescue (SaR) robots. The work is motivated by the need to improve the efficiency of SaR operations after disasters, using robots. In fact, the main benefits of using SaR robots are reduced cost, improved speed of response, increased search performance, extended reachability to otherwise inaccessible places, and fewer risks for the SaR crew. Robots can optimize the mission plans and safely explore the environment through systematic mathematical approaches. Therefore, novel control approaches are needed to enable
robots to perform SaR operations autonomously and time-efficiently, and this is the main objective of this thesis.
The main contributions of this PhD thesis are the following:

1. We propose novel mission planning frameworks and architectures for ground or flying SaR robots based on Model Predictive Control (MPC), Fuzzy Logic Control (FLC), and other control approaches, in some cases combined to exploit the advantages of multiple methods.

2. We integrate our architectures with models for moving targets and dynamic obstacles, and we leverage robust control formulations to deal with uncertainties and perception approaches to map the SaR environment and track targets.

3. We validate our approaches by comparing them to other state-of-the-art approaches in case studies with simulations and in some cases with real-life experiments in the lab. ... Read more

This dissertation develops a unified framework for defining, computing, and certifying safe operating regions for autonomous dynamical systems operating under deterministic, robust, and stochastic dynamics. As these systems are increasingly deployed in safety-critical environments—ranging from autonomous vehicles and air traffic management to robotics and automated decision-making—ensuring that their behaviour remains within safe bounds is of vital importance. Classical reachability methods offer strong theoretical guarantees, but they often rely on idealised models, struggle with high-dimensional systems, and cannot formally certify learned approximations. This work addresses these limitations through new safety definitions, robust generalisations of reachability theory, and a learning-verification pipeline for scalable certification.

The first contribution focuses on the notion of safe sets themselves. \Cref{ch:maneuverable} introduces the concept of the maneuverable control-invariant set—a novel refinement of traditional invariant sets. Unlike classical safe sets, which guarantee that the system can remain within a given region, maneuverable sets also ensure the system's ability to move between any two states within the set in a given time horizon. Although such maneuverability and invariance cannot always be achieved simultaneously, this definition adds a valuable dimension to safety characterisation. The set is computed using a combination of forward and backward reachable sets derived from Hamilton-Jacobi (HJ) reachability analysis. Building on this, \Cref{ch:safeset} presents an improved definition of a safe set that does guarantee both viability and full internal maneuverability simultaneously. It is obtained through a different combination of forward and backward reachable sets. While the first two chapters establish refined definitions of safe sets under idealised dynamics, real-world systems rarely operate with perfect models. To address this, the next contribution turns to the challenge of ensuring safety under model uncertainty and disturbances.

\Cref{ch:RobustBRT} addresses robustness under model uncertainty. It extends the HJ reachability framework by formulating a three-player differential game that includes unmodelled dynamics as an additional adversary alongside external disturbances. The resulting Robust Avoid Backward Reachable Tube (RABRT) generalises previous formulations by ensuring safety in the presence of both disturbances and uncertain system dynamics. The associated value function is shown to be the viscosity solution to a well-defined Hamilton-Jacobi-Isaacs (HJI) partial differential equation (PDE), and the method integrates with existing level-set tools, preserving compatibility with standard computational frameworks. While this formulation provides robustness to unmodelled dynamics, its computational demands grow quickly with system dimensionality. As classical level-set methods struggle in high-dimensional settings, recent work has turned to neural network–based approximations. The next chapter explores how such approximations can be made verifiable and trustworthy through formal certification.

To overcome the curse of dimensionality inherent in classical reachability methods, \Cref{ch:Care} introduces the Certified Approximate Reachability (CARe) framework. Instead of solving PDEs directly on grids, CARe leverages neural networks to approximate the value function and then uses formal verification tools—specifically Satisfiability Modulo Theory (SMT) solvers and Counterexample-Guided Inductive Synthesis (CEGIS)—to certify the learned approximation within a bounded error. This framework bridges the gap between deep learning and formal safety verification, offering a scalable method for high-dimensional systems while preserving correctness guarantees. While CARe provides a scalable and certifiable solution for deterministic systems, many real-world applications involve stochastic dynamics and probabilistic safety requirements. To address this, the final part of the thesis extends the reachability framework to systems governed by uncertainty and noise.

Chapters \ref{ch:StochasticABRT} and \ref{ch:stochasticsafeBRT} extend the thesis into the stochastic domain. \Cref{ch:StochasticABRT} introduces the Probabilistic Forced Backward Reachable Set, which identifies states from which the system enters an undesirable set with high probability, regardless of the control policy. The chapter develops convergent algorithms for computing this set in both discrete and continuous settings and provides error bounds under discretisation. \Cref{ch:stochasticsafeBRT} then presents the Safe Probable Backward Reachable Set (SPBRS), which guarantees not only that the system will reach a target with a given probability but also that it will remain inside a certified safe set throughout the trajectory. This is especially important in stochastic systems where intermediate safety violations can occur despite a safe terminal state. Two algorithms are developed—one computationally efficient and the other optimal—and their convergence and tractability are formally analysed.

Each chapter is supported by theoretical proofs, algorithmic implementation, and numerical validation on benchmark systems such as the Dubins car and double integrator. This dissertation offers a cohesive and extensible framework for analysing and certifying safety in both deterministic and stochastic systems. By introducing new definitions of safety, developing scalable algorithms, and integrating formal verification with deep learning, the work makes contributions that are both foundational and practical. It addresses longstanding limitations in the scalability, conservatism, and verifiability of reachability-based safety analysis. The results have direct implications for real-world autonomous systems, including autonomous vehicles, robotics, air traffic management, and safety-critical embedded systems. More broadly, the methods contribute to the growing effort to make machine learning and autonomy formally trustworthy, especially in domains where safety cannot be left to statistical performance alone. Looking ahead, this work lays the groundwork for future research in adaptive safety verification, stochastic control synthesis, and trustworthy decision-making in uncertain environments—pushing the boundaries of what is possible at the intersection of formal methods, control theory, and autonomy. ... Read more

Urban search and rescue (USaR) missions operate in volatile environments where time, information, and energy are scarce. This thesis presents a hierarchical decision-making and coordination architecture that elevates heterogeneous robotic autonomy from local actions to mission-level reasoning. Robots are modeled by capabilities of action and perception rather than fixed roles, allowing robot-specific suitability to be evaluated for any tasks that evolve during operation considering mission metrics. Sequential mission planning is formulated as a nonlinear model predictive control (MPC) problem with safe-return constraints over a finite horizon. Task suitability is first estimated by a fuzzy inference system that fuses robot capabilities with task-specific priorities. A multi-criteria decision-making (MCDM) framework assigns mission utilities to candidate actions by weighing urgency, information gain, and inter-task preferences, providing flexibility when task definitions are ambiguous and priorities shift.

The decision process is implemented in a two-tier hierarchial control scheme. Locally, each robot solves a tractable linearization of the MPC by filtering independent actions using utilities and observation overlaps, followed by a mixed-integer linear programming (MILP) scheduler subject to energy-aware safe-return constraints. Globally, a multi-robot extension reconciles local plans to produce a consistent allocation that balances distributed adaptability with centralized coordination.

The architecture is validated in realistic simulations with heterogeneous robots and evolving tasks. Relative to heuristic baselines, it accelerates victim identification and area mapping while achieving higher utility in most of 17 randomized single-robot scenarios. In multi-robot experiments, it allocates high-priority tasks to the most suitable platforms and completes missions faster than non-cooperative approaches, while using fewer computational resources than comparable optimization-based methods. ... Read more

This thesis presents a hybrid control framework that combines Reinforcement Learning (RL) and Model Predictive Control (MPC) to achieve constraint-satisfying flutter suppression and load alleviation in flexible aircraft subject to turbulent gusts. During training, MPC computes safe input bounds using high-fidelity Linear Parameter Varying (LPV) models and long prediction horizons, exploiting known disturbances to accurately capture aeroelastic behavior. A Q-learning agent is trained to select control actions within these bounds, adapting to nonlinear dynamics and actuator delay. At deployment, the learned policy operates from a lightweight Q-table, with certified interpolation ensuring constraint satisfaction even for unseen states. By integrating the anticipatory capabilities of MPC with the adaptability of RL, the framework enables effective control under turbulence and structural uncertainty. While demonstrated on an aeroelastic aircraft, the approach can be generalized to other systems with similar dynamics, where constraint handling and adaptive control under uncertainty are critical. ... Read more

In time-critical disaster scenarios, heterogeneous robotic teams combining unmanned aerial vehicles (UAVs) and unmanned ground vehicles(UGVs) have the potential to enhance situational awareness and improve navigation capabilities in complex indoor environments. This research presents a hierarchical mission planning approach designed for multi-agent search-and-rescue systems (SaRs), explicitly leveraging the distinct sensory and physical capabilities of UAVs and UGVs to maximise victim detection, area coverage, and energy efficiency. At the local control level, each SaR agent utilizes a heuristic controller based on fuzzy logic control (FLC) and the A* pathfinding algorithm, enabling efficient determination of individual targets and paths. At a higher level, a centralised supervisory controller employing model predictive control (MPC) coordinates global mission activities, intervening specifically when an agent’s capabilities become suboptimal within certain regions of the environment. Under randomised indoor SaR simulations featuring varied victim distributions, obstacle placements, and environmental conditions—including terrain complexity and visibility, the proposed cooperative approach consistently outperformed baseline methods. Although an exploration-focused heuristic approach (ACS) achieved slightly higher area coverage, the cooperative framework provided comparable coverage while achieving significantly higher victim detection performance, superior consistency, and better overall efficiency. Specialised scenario evaluations further illustrated that the cooperative approach can allocate and execute SaR-related search tasks more effectively, achieving superior operational performance compared to purely local-control-based (selfish) methods. These findings demonstrate that the hierarchical integration of heuristic local control and centralised MPC-based supervision, combined with the complementary capabilities of heterogeneous UAV-UGV teams, results in
robust, scalable, and efficient mission planning solutions suitable for complex, uncertain indoor SaR scenarios. ... Read more

his thesis introduces a Model Predictive Fuzzy Control (MPFC) framework for mission-planning in multi-agent systems operating in dynamic, uncertain environments. MPFC integrates predictive decision-making from Model Predictive Control (MPC) with the adaptability of Fuzzy Logic Control (FLC) to handle complex, time-sensitive tasks. A generic mathematical formulation is developed and applied to a multi-agent Search and Rescue case study with fire and wind dynamics. Comparative performance analysis against MPC and Pre-Tuned FLC controllers, along with sensitivity analyses and design exploration, demonstrate that MPFC can outperform existing approaches and maintain robust performance under varying conditions. Although this technique is promising, several limitations and challenges are identified, suggesting avenues for future research to refine and enhance its applicability. ... Read more

Marine pollution is a critical issue impacting the global community, with underwater waste a particularly daunting challenge. While autonomous detection and collection of underwater waste is highly desirable, these are extremely difficult tasks. This difficulty arises from the intrinsic complexities of the aquatic environment, including variable lighting conditions, reduced visibility, and the complex nature of water currents. This paper focuses on novel approaches for autonomous detection of underwater waste and proposes to incorporate domain knowledge to refine deep-learning-based underwater object detection techniques. More specifically, the domain knowledge is represented with models in the state space form that describe the motion of the target objects to assist the classification of objects. Moreover, optical flow is combined with a k-means clustering algorithm to extract the trajectory of the target objects from videos. These trajectories are subsequently fed into a neural network that is trained using knowledge distillation, enhanced with domain knowledge. For our experiments, a simulator is devised to facilitate the creation of a dataset for developing and testing the proposed architecture. The results of the experiments demonstrate that including domain knowledge within the object detection approach with neural networks provides numerous substantial advantages, including enhanced robustness against noisy and poorly labelled data, facilitation of semi-supervised learning, and consistent superiority in accuracy over the baseline scenario. Additionally, combining the domain knowledge with a neural network significantly increases the computational speed of object detection compared to using the standalone domain knowledge module. ... Read more

Search and Rescue (SaR) missions present challenges due to the complexity of the disaster scenarios. Most life losses and injuries occur in developing countries. Robotics has become indispensable for rapidly locating disaster victims. Combining flying and ground robots more effectively serves this purpose due to their complementary features. To this end, a cost-effective framework to perform conventional SaR tasks is presented. The method leverages You Only Look Once and video streams by an Unmanned Ground Vehicle (UGV) and an Unmanned Aerial Vehicle (UAV). In exploiting pose estimation to perform human depth estimation, the susceptibility of the algorithm to variations in poses was unveiled. In tracking object trajectories, the collaboration is advantageous in wide-area cluttered trajectories as opposed to narrow-area unobstructed trajectories. In mapping terrain elevation, errors drop significantly with the assistance of the UAV. Moving forward, devising adaptable strategies tailored to diverse SaR scenarios will be pivotal. ... Read more

Tiny object detection (TOD) in satellite imagery is critical for applications including pipeline monitoring, where the detection of tiny objects, such as excavators near the pipeline networks, can prevent potential incidents. However, TOD faces challenges due to the limited pixel representation of objects, complications with Intersection over Union-based label assignment, and semantic confusion between visually similar classes. This paper introduces OE-Net, a modular extension of Oriented R-CNN,
specifically designed for TOD in satellite images. OE-Net integrates three novel components: a Dilated Fusion module to enhance the feature extraction, a Balanced Ranking Assigner to improve the anchor matching, and Vector Mapping for more precise classification of visually similar objects. The modular design of OE-Net allows its novel components to be easily integrated into other detection models. Tested on the TinyDOTA dataset, OE-Net sets a new benchmark in TOD performance, achieving a 6.3
percentage points improvement in mean Average Precision (mAP) over the Oriented R-CNN baseline. On a novel excavator detection dataset, named ExcaSat and developed for pipeline monitoring, OE-Net outperforms the Oriented R-CNN baseline by 11.1 percentage points in mAP. Furthermore, OE-Net surpasses state-of-the-art methods on both TinyDOTA and ExcaSat, establishing a new standard in computational efficiency and detection precision for satellite-based monitoring tasks. The code is available at: https://github.com/OE-jimvanoosten/mmrotate-OE.git. ... Read more

The impact of disasters on the affected population is catastrophic. Proper disaster management practices are needed to reduce their societal damage. This includes Search and Rescue (SaR) missions, which pertain measures to find potentially trapped victims. This task can be alleviated by the support of a fleet of SaR robots. This fleet is deployed to locate trapped victims and to report back their position to the emergency responders and potentially to follow them to track of their potential movements. Locating trapped victims is challenging, as the number of these people, their initial position and potential displacements may be unknown. A behavioral model of the victims can give insight on how they take decisions and act during an evacuation situation. Several techniques are used in the state of the art. This study proposes an evacuation model that integrates game theory into the belief-desire-intention framework. The model is validated with a benchmark from the state of the art. It is found that the model is able to produce realistic evacuation times. This depends on the distribution of the game theoretic strategies in the population. SaR robots are then added to the validated model. It is found that their presence reduces the evacuation time, depending on several parameters influencing trust of the victims in the robots. Thus, this research contributes to the field of research SaR operations by providing insight into the behaviour of the trapped victims in the presence of rescue robots. ... Read more

A key challenge for SaR robotics is to avoid dynamic obstacles in cluttered environments, with limited and noisy information. In this research, a controller for SaR robots is developed by coupling a local heuristic motion planner with a model predictive control (MPC) based trajectory tracker. Constraint tightening and tube-based control are used to make the MPC robust to model mismatch and additive measurement noise, while the motion planner is integrated with the MPC. The motion planner periodically supplies a reference trajectory to the trajectory tracker, but the MPC can request additional updates in case of a noticeable mismatch between the predicted and measured environment, based on a user-defined threshold. A case study is designed in MATLAB where a single robot needs to reach a goal through a cluttered environment with dynamic obstacles. Results from the case study show that the MPC method outperforms two state-of-the-art control approaches that are based on the rapidly-exploring random tree (RRT) and artificial potential function (APF) methods. In particular, the heuristic and MPC coupled controller showed a higher success rate in reaching the goal without collisions, and displayed a lower path length in cases with both low and high computational budget. ... Read more

An aging population puts a pressure on health-care workers working with dementia patients globally. A potential solution is to provide care with Socially Assistive Robots (SARs), i.e. robots who help people through social interaction. However, for effective care these SARs must be able to personalize their behavior to individual patients and adapt this behavior to changes in this patient’s references. This paper presents the decision making process of a SAR that enables the SAR to personalize its behavior to the personality of the patients and adapt this behavior to their current state-of-mind. The system consists of a Fuzzy Logic Control personalization module, which personalizes the SAR’s behavior and a Reinforcement Learning based decision making module together with a Fuzzy Logic Control reward module for the adaption of the personalized behavior. The personalization of the SAR’s behavior is assessed by comparing the output of the system with answers from a survey. The average scatter index over all different behavioral parameters of the SAR is 20.4% The adaptation of the behavior is assessed with computer-based simulations, where an overall accuracy of 81.8% is achieved. A third experiment is carried out to assess the effect of adding the Fuzzy Logic Control personalization module to the system. This experiment shows that adding the personalization module to the decision making system of the SAR decreases the time for the learning process to converge with 13.3%. Although the first assessments of the system look promising, more extensive experiments should be held in later stages of the research. A crucial experiment that must be held in future research is performing real-life interactions between dementia patients and the SAR, in such an experiment the functionality of the system really can be assessed. ... Read more