We present an online underwater localization method that enables passive vehicles to localize themselves by overhearing two-way ranging (TWR) communication between an active vehicle and multiple acoustic anchors. Swarm simulations confirm the scalability of the system and show faster position updates, as well as improved accuracy compared to the active benchmark. A field experiment in a shallow water canal using the low-cost BlueROV2 underwater vehicle and ahoi modems has demonstrated the real-world applicability of the system, achieving an RMSE of approximately 0.82m.
An efficient transportation system leads to increased safety, better financial security, as well as a cleaner environment. We propose an approach for significantly improving efficiency in heterogeneous unregulated traffic environments. Existing intelligent transportation solutions for improving efficiency, however, apply only to traffic intersections with sparse traffic or traffic where drivers obey the regulations, or both. Our approach combines an auction mechanism with an optimization-based trajectory planner. The auction mechanism generates a fair priority entrance sequence for heterogeneous agents while the trajectory planner computes velocity controls in real-time, taking less than 10 milliseconds even in high-density traffic with over 10, 000 vehicles per hour. Compared to fully optimization-based methods, it operates 100x faster while ensuring fairness, safety, and efficiency. Tested on the SUMO simulator, our algorithm improves throughput by at least 25%, reduces the time to reach the goal by at least 70%, and decreases fuel consumption by 50% compared to approaches using traffic lights and stop signs. Our approach is also unaffected by unbalanced traffic inflows, whereas some of the other baselines encountered a decrease in performance in unbalanced traffic inflow environments.
This paper presents an Auction-Consensus Algorithm with a Loss Mechanism (ACALM), a decentralized task allocation method for multi-robot systems that enhances the existing Consensus-Based Auction Algorithm (CBAA) by incorporating a novel loss propagation mechanism. In contrast to purely greedy bidding strategies, it enables agents to dynamically update task priorities based on the accumulated loss from previously unsuccessful bids. This extended work reduces globally inefficient allocations caused by early suboptimal decisions. The proposed approach is evaluated through large-scale simulations in thousands of randomized scenarios and swarm sizes ranging from 5 to 120 robots. Compared to existing CBAA and GCAA algorithms, ACALM yields task assignments with higher global efficiency on average. The results also show that ACALM maintains effectiveness with larger swarms, suggesting strong robustness in large decentralized contexts. However, this improvement comes with an increase in communication overhead due to additional consensus rounds. Potential extensions include methods to reduce communication costs, support for multiple assignments per robot, and verification of functionality under asynchronous communication.
Multi-Agent Teamwise Cooperative Path Finding (TC-MAPF) seeks collision-free paths for the agents from their start to goal locations. In addition, agents are grouped into multiple teams, and each team has its own objective function to optimize. TC-MAPF arises in scenarios such as the coordination of multiple autonomous vehicles at a signal-free traffic intersection, where, for example, the vehicles from each direction naturally forms a team. TC-MAPF was recently studied and optimal planners such as TC-CBS has been developed. While being able to find all Pareto-optimal solutions for TC-MAPF, these optimal planners usually suffer from limited scalability as the number of agents grows. This paper develops a bounded sub-optimal planner TC-CBS-TF for TC-MAPF, trading off solution quality for scalability, by leveraging and extending several bounded sub-optimal search techniques for MAPF to handle multiple teams as in TC-MAPF. We test TC-CBS-TF and baselines on various maps with up to 50 agents, and TC-CBS-TF achieves up to 68 % higher success rates.
We consider the problem of combining potential field and ergodic search on multi-robot systems. Traditional ergodic search algorithms use metrics for ergodicity that account for the desired distribution at different scales. Recently, a heat equation-driven ergodic approach was proposed, which adds flexibility to the smoothing of the ergodic metric. However, such an approach, as it is an isotropic diffusion, propagates the error uniformly in all directions, regardless of changes in the desired distribution. We introduce a general class of anisotropic diffusion formulation of the ergodicity problem, which generates a potential field for the ergodic search. We demonstrate that this approach generalizes previous results, which consider radial basis functions and the solution of the heat equation to represent the difference between the goal density distribution and the covered trajectories. In our solution, the agent movement is directed using the gradient of the solution of the Perona-Malik diffusion, and our formulation includes the heat equation as a special case. We demonstrate the methodology with a series of simulations in different scenarios.
Search-and-Rescue Network (SARNet) is a lightweight hybrid routing protocol for unmanned aerial vehicle (UAV) swarms deployed in search-and-rescue missions. SARNet enhances the Optimized Link State Routing protocol (OLSR) with event-driven topology updates and signal-to-interference-plus-noise-ratio (SINR)-based link-quality triggers. This paper evaluates a partial implementation of SARNet, which currently includes these two mechanisms, while flow-aware QoS modules remain future work. Using the KogakuinMobility model to generate realistic, mission-oriented UAV trajectories, we compare SARNet and OLSR in five urban maps and an open-field setting with OMNeT++/INET simulations. In dense city environments, both protocols exhibit low packet delivery ratios (PDR < 7%), with OLSR maintaining slightly higher PDR and consistently lower overhead due to its timer-driven control. In open terrain, both sustain high PDR (similar to 85- 95%) under light and moderate loads but degrade sharply at 3.93 Mb/s due to MAC/PHY saturation rather than routing. SARNet's reactive SINR triggers increase control traffic as network density grows, reflecting sensitivity to parameter tuning. These findings indicate that while SARNet's event-driven design improves adaptivity, further refinements, such as adaptive hysteresis or threshold adjustment, are required for stable and scalable communication in diverse search and rescue flying ad hoc network (SAR-FANET) deployments.
This paper addresses the routing and scheduling problem for a marine marsupial system, where an unmanned surface vehicle (USV) and multiple autonomous underwater vehicles (AUVs) cooperate to efficiently perform distributed subsea tasks. Due to the requirement that the USV must return to deployment locations for AUV retrieval, the problem involves unique cross-schedule dependencies. We formulate this problem, termed the Marsupial-TSP, as a mixed-integer linear program (MILP) that explicitly incorporates these constraints while minimizing the overall mission completion time. To address the NP-hard nature of the problem and enhance scalability, we introduce a divide-and-conquer planning framework. In the divide phase, we introduce the Marsupial-aware Agglomerative Clustering (MAC) algorithm, designed to capture key operational characteristics of the marsupial system and effectively partition the task set. In the conquer phase, we solve the resulting open-ended Marsupial-TSP subproblems to construct a complete solution. The proposed approach is validated through Monte Carlo simulations and qualitative analyses, including Gantt charts and route visualizations. Simulation results demonstrate that our method significantly improves both computational efficiency and solution quality.
Multi-robot exploration of unknown environments is challenging, especially in scenarios with constrained communication. Decentralised Monte Carlo Tree Search (Dec-MCTS) is a promising online planning approach that enables robots to collaboratively explore an environment while handling communication failures. In this work, we propose a communication strategy for a frontier-guided version of the Dec-MCTS that is robust to communication loss and limited range while maintaining efficient exploration. Through extensive experimental evaluations, we demonstrate that limiting communication does not significantly impact performance and that discarding outdated information can negatively affect exploration efficiency. We also study how information propagation can positively or negatively impact the mission depending on the environment.
Decentralised multi-robot coordination strategies often assume that all robots communicate with one another to plan jointly. While effective in small teams, this all-toall communication model becomes infeasible under realistic bandwidth constraints, presenting a fundamental challenge to scalability. To address this, we propose a coordination framework that includes a novel coalition formation step, which partitions robots into sub-teams, grouping together those whose optimal actions strongly depend on one another. At the core of our approach is a sampling-based, information-theoretic utility measure that captures higher-order dependenciesbeyond simple pairwise interactions-between robots' joint action spaces. This utility guides a greedy agglomerative clustering algorithm to form coalitions tailored to the specific coordination demands of the task. We evaluate our method on a multi-robot information-gathering scenario with complex interrobot dependencies. Results show that our approach achieves comparable or better performance than full connectivity while significantly reducing communication, demonstrating improved coordination efficiency crucial for scaling multi-robot systems.
This paper presents a decentralized control approach for extending the operational lifetime of aerial communication networks via perching. We propose a switching-controller-based algorithm that enables UAVs to maintain connectivity while minimizing energy consumption by opportunistically perching. The algorithm is evaluated in diverse simulated scenarios using the Gymnasium framework, highlighting its performance and limitations. The approach is evaluated in real-world conditions, using Bitcraze Crazyflie UAVs and an AgileX Scout Mini UGV, integrating a ROS 2 wrapper for seamless interaction between simulation and physical platforms. Experimental results demonstrate the practicality of the approach and reveal key challenges posed by hardware constraints. The proposed framework offers a scalable and energy-aware solution for persistent aerial communication networks. Code: https://github.com/larics/mrs_connectivity_perch Video: https://youtu.be/u8JEi_p27BQ
We propose a cooperative active search framework for localizing people in industrial environments with a team of heterogeneous robots. In contrast to exhaustive search methods, our approach leverages prior information about the initial target distribution and potential behavior to enable primed probabilistic search strategies. To estimate a person's motion trajectory, we combine global path planning with the Social Force Model. We formulate the multi-robot coordination problem as a Mixed Integer Linear Programming (MILP) optimization that accounts for visibility constraints and heterogeneous robot capabilities. Monte Carlo simulation experiments in two real-world industrial scenarios based on simulated and real human trajectory data show that our approach outperforms several baselines. The results demonstrate the effectiveness of combining predictive target modeling for coordinating robot search teams.
Autonomous surface vehicles (ASVs) need to navigate in the sea while following COLREG rules. The collision avoidance planners should provide guarantees on collision avoidance while maintaining near-optimal plans. Due to the inclusion of the COLREGs, most of the planners are reactive and do not provide guarantees. To meet these requirements, in this paper, we present a search-based planning strategy that is computationally efficient, COLREGs-compliant, complete, and provides guarantees on the collision-free path. The proposed strategy is evaluated through simulations for different scenarios. We also perform real-world experiments to demonstrate the performance of our approach.
The area coverage problemw, here robots collectively monitor a region, arises in environmental monitoring, surveillance, and search-and-rescue. Centroidal Voronoi Tessellations (CVTs) offer a widely studied solution, partitioning space and positioning robots at cell centroids for optimal coverage. Existing work assumes perfect sensing and communication, or treats imperfections only as performance degraders. We take a different view: sensing inaccuracies and communication disruptions meaningfully shape system dynamics. Through rigorous analysis, we show that small position errors and Delaunay graph disruptions prevent convergence to static CVT equilibria, instead creating persistent dynamic behavior. This quantitative characterization reveals that sensitivity to imperfect conditions, often seen as fragility, can inspire new algorithmic strategies. Our findings challenge common theoretical assumptions and enable robust, adaptive multi-robot systems suited to unpredictable real-world environments.
Multi-agent pursuit-evasion scenarios in the presence of learning targets pose a significant challenge for robot swarms. In this work, we investigate a cooperative strategy among decentralized unmanned aerial vehicles to keep track of a rapidly evading target afforded with learning capabilities. To deal with the non-stationarity of the environment while solely depending on local observations, we propose a multi-agent reinforcement learning framework enhanced with dynamic intraswarm communication protocols available during the training process. Specifically, we propose, (a) a coverage-range-based communication network, and (b) a $k$-nearest neighborhood communication network which are both developed based on reward policies aimed at maintaining network connectivity despite its dynamic nature. Both of these protocols are integrated into the training process as additional reward-shaping terms, which allow the agents to be deployed in a decentralized manner without requiring any information exchange based on a learned joint policy that adapts to the presence of a communication channel. The multi-agent system is trained in a simulated environment using the Multiple Particle Environment and deployed on Crazyflie drones in a real-world controlled environment. Our experimental results in simulation and real-world show that swarms trained without our proposed networking protocols fail to learn any effective policy. In addition, our proposed dynamic communication strategy during training yields two notable benefits to the system's performance during deployment: (i) it reduces the average distance to the target agent, and (ii) it enables the learning of the pursuit task in larger environments.
The arrangement of pods (mobile storage units) is critical to the throughput of a Robotic Mobile Fulfillment System, where robots travel between pods and workstations to complete orders. Our focus is on positioning a specified number of pods while maximizing system throughput and guaranteeing connectivity between every workstation and every pod. Existing methods either fail to scale to large warehouses or yield layouts with highly unpredictable performance. To overcome these limitations, we propose a two-step framework that optimizes individual pod positions, by first computing high-density connected layouts, and then successively removing pods chosen to best improve throughput. The framework generates layouts that enable a multi-robot system to robustly achieve 12-34% higher maximum throughput compared to traditional rectangular pod island designs. The method scales well, effectively handling warehouses up to 80 x 80 grid tiles.
Pursuit-evasion games can model strategic interactions between teams with conflicting objectives. Solving these problems is difficult due to the curse of dimensionality. One way to overcome this is by dividing the game into sub-games and solving them individually. This method has yielded effective solutions. However, computing the sub-games every timestep requires excessive computational effort. To address this, we studied a metareasoning approach that determines when to compute sub-games during the episode. At each timestep, the metareasoning policy uses a learning-based model to determine whether to compute the sub-games. Results indicate that metareasoning can significantly reduce the number of subgames computed while maintaining performance.
We address the problem of coordinating a team of robots to cover an unknown environment while ensuring safe operation and avoiding collisions with non-cooperative agents. Traditional coverage strategies often rely on simplified assumptions, such as known or convex environments and static density functions, and struggle to adapt to real-world scenarios, especially when humans are involved. In this work, we propose a human-aware coverage framework based on Model Predictive Control (MPC), namely HMPCC, where human motion predictions are integrated into the planning process. By anticipating human trajectories within the MPC horizon, robots can proactively coordinate their actions and adapt to dynamic conditions. The environment is modeled as a Gaussian Mixture Model (GMM), representing regions of interest. Team members operate in a fully decentralized manner, without relying on explicit communication-an essential feature in hostile or communication-limited scenarios. Our results show that human trajectory forecasting enables more efficient and adaptive coverage, improving coordination between human and robotic agents.
Effective collective decision-making in swarm robotics often requires balancing exploration, communication and individual uncertainty estimation, especially in hazardous environments where direct measurements are limited or costly. We propose a decentralized Bayesian framework that enables a swarm of simple robots to identify the safer of two areas, each characterized by an unknown rate of hazardous events governed by a Poisson process. Robots employ a conjugate prior to gradually predict the times between events and derive confidence estimates to adapt their behavior. Our simulation results show that the robot swarm consistently chooses the correct area while reducing exposure to hazardous events by being sample-efficient. Compared to baseline heuristics, our proposed approach shows better performance in terms of safety and speed of convergence. The proposed scenario has potential to extend the current set of benchmarks in collective decision-making and our method has applications in adaptive risk-aware sampling and exploration in hazardous, dynamic environments.
Maze-like environments, such as cave and pipe networks, pose unique challenges for multiple robots to coordinate, including communication constraints and congestion. To address these challenges, we propose a distributed multi-agent maze traversal algorithm for environments that can be represented by acyclic graphs. It uses a leader-switching mechanism where one agent, assuming a head role, employs any single-agent maze solver while the other agents each choose an agent to follow. The head role gets transferred to neighboring agents where necessary, ensuring it follows the same path as a single agent would. The multi-agent maze traversal algorithm is evaluated in simulations with groups of up to 300 agents, various maze sizes, and multiple single-agent maze solvers. It is compared against strategies that are naïve, or assume either global communication or full knowledge of the environment. The algorithm outperforms the naïve strategy in terms of makespan and sum-of-fuel. It is superior to the global-communication strategy in terms of makespan but is inferior to it in terms of sum-of-fuel. The findings suggest it is asymptotically equivalent to the full-knowledge strategy with respect to either metric. Moreover, real-world experiments with up to 20 Pi-puck robots confirm the feasibility of the approach.
This paper presents a novel approach to avoiding jackknifing and mutual collisions in Heavy Articulated Vehicles (HAVs) by leveraging decentralized swarm intelligence. In contrast to typical swarm robotics research, our robots are elongated and exhibit complex kinematics, introducing unique challenges. Despite its relevance to real-world applications such as logistics automation, remote mining, airport baggage transport, and agricultural operations, this problem has not been addressed in the existing literature. To tackle this new class of swarm robotics problems, we propose a purely reaction-based, decentralized swarm intelligence strategy tailored to automate elongated, articulated vehicles. The method presented in this paper prioritizes jackknifing avoidance and establishes a foundation for mutual collision avoidance. We validate our approach through extensive simulation experiments and provide a comprehensive analysis of its performance. For the experiments with a single HAV, we observe that for 99.8% jackknifing was successfully avoided and that 86.7% and 83.4% reach their first and second goals, respectively. With two HAVs interacting, we observe 98.9%, 79.4%, and 65.1%, respectively, while 99.7% of the HAVs do not experience mutual collisions.
