This paper introduces a novel distributed approach for forming UAV-based multi-hop relay networks by adapting traditional flocking models to create relay chains between remote points. Our method modifies the standard flocking paradigm by incorporating dynamic agent roles, allowing UAVs to self-organize based solely on local state and neighbor information, and integrates networking information such as routing decisions directly into mobility control. A side contribution is the introduction of a Line-of-Sight (LOS) conservation force, which mitigates communication failures due to obstacles and is easily adaptable to the flocking model. The proposed algorithm is evaluated using a joint robotics and network co-simulator that combines realistic multi-rotor physics with ns-3-based network simulations. Simulation results across diverse environments and varying mission complexities demonstrate that our approach effectively maintains connectivity, enhances Quality of Service (QoS), and scales robustly, thereby bridging the gap between robotic control and aerial wireless network design.

Successfully addressing a wide variety of tasks is a core ability of autonomous agents, requiring flexibly adapting the underlying decision-making strategies and, as we ar-gue in this work, also adapting the perception modules. An analogical argument would be the human visual system, which uses top-down signals to focus attention determined by the current task. Similarly, we adapt pretrained large vision models conditioned on specific downstream tasks in the context of multitask policy learning. We introduce task-conditioned adapters that do not require finetuning any pretrained weights, combined with a single policy trained with behavior cloning and capable of addressing multiple tasks. We condition the visual adapters on task embeddings, which can be selected at inference if the task is known, or alternatively inferred from a set of example demonstrations. To this end, we propose a new optimization-based estimator. We evaluate the method on a wide variety of tasks from the CortexBench benchmark and show that, compared to existing work, it can be addressed with a single policy. In particular, we demonstrate that adapting visual features is a key design choice and that the method generalizes to unseen tasks given a few demonstrations.

Implicit representations such as Neural Radiance Fields (NeRF) allow to map color, density and semantics in a 3D scene through a continuous neural function. However, these models typically require manual and careful human data collection for training. This paper addresses the problem of active exploration for autonomous NeRF construction. We study how an agent can learn to efficiently explore an unknown 3D environment so that the data collected during autonomous exploration enables the learning of a high-quality neural implicit map representation. The quality of the learned representation is evaluated on four robotics-related downstream tasks: classical viewpoint rendering, map reconstruction, planning, and pose refinement. We compare the impact of different exploration strategies including frontier-based and learning-based approaches (end-to-end and modular) with different reward functions tailored to this problem. Empirical results show that NeRFs can be trained on actively collected data using just a single episode of experience in an unseen environment and that AutoNeRF, a modular exploration policy trained with reinforcement learning, enables obtaining a higher-quality NeRF for the considered downstream robotic tasks. Finally, we show that with AutoNeRF an agent can be deployed to a previously unknown scene and then automatically improve its navigation performance by adapting to the scene through a cycle of exploration, reconstruction, and policy finetuning.

How to coordinate multiple robots moving in modifiable cluttered environments? In this paper, we introduce the multi-robot version of the NAMO problem (Navigation Among Movable Obstacles). In MR-NAMO, robots must not only plan for the possibility of displacing obstacles as needed to facilitate their navigation, but also solve conflicts that may arise when trying to simultaneously access a location or obstacle. After identifying all different types of conflicts, we define and compare variants of an implicit coordination strategy allowing the use of existing NAMO algorithms [1] in a Multi-Robot context. We also show how our previously introduced social occupation cost model [2] can improve the efficiency of multirobot plans with better obstacle placement choices, and how it can be applied in a novel way to find relevant robot placement choices to solve deadlock situations.

Understanding and mapping a new environment are core abilities of any autonomously navigating agent. While classical robotics usually estimates maps in a stand-alone manner with SLAM variants, which maintain a topological or metric representation, end-to-end learning of navigation keeps some form of memory in a neural network. Networks are typically imbued with inductive biases, which can range from vectorial representations to birds-eye metric tensors or topological structures. In this work, we propose to structure neural networks with two neural implicit representations, which are learned dynamically during each episode and map the content of the scene: (i) the Semantic Finder predicts the position of a previously seen queried object; (ii) the Occupancy and Exploration Implicit Representation encapsulates information about explored area and obstacles, and is queried with a novel global read mechanism which directly maps from function space to a usable embedding space. Both representations are leveraged by an agent trained with Reinforcement Learning (RL) and learned online during each episode. We evaluate the agent on Multi-Object Navigation and show the high impact of using neural implicit representations as a memory source.

This paper deals with the anonymous multi-agent path finding (MAPF) problem for a team of tethered robots. The goal is to find a set of non-crossing paths such that the makespan is minimal. A difficulty comes from the fact that a safety distance must be maintained between two robots when they pass through the same subpath, to avoid collisions and cable entanglements. Hence, robots must be synchronized and waiting times must be added when computing the makespan. We show that bounds can be efficiently computed by solving linear assignment problems. We introduce a variable neighborhood search method to improve upper bounds, and a Constraint Programming model to compute optimal solutions. We experimentally evaluate our approach on three different kinds of instances.

This paper addresses a path planning problem for a fleet of Unmanned Aerial Vehicles (UAVs) that uses both prior information and online gathered data to efficiently inspect large surfaces such as ship hulls and water tanks. UAVs can detect corrosion patches and other defects on the surface from low-resolution images. If defects are detected, they get closer to the surface for a high-resolution inspection. The prior information provides expected defects locations and is affected by both false positives and false negatives. The mission objective is to prioritize the close-up inspection of defected areas while keeping a reasonable time for the coverage of the entire surface. We propose two solutions to this problem: a coverage algorithm that divides the problem into a set of Traveling Salesman Problems (Part-TSP) and a cooperative frontier approach that introduces frontier utilities to incorporate the prior information (Coop-Frontier). We finally provide extensive simulation results to analyze the performance of these approaches and compare them with alternative solutions. These results suggest that both Part-Tspand Coop-Frontier perform better than the baseline solution. Part-Tsphas the best performance in most cases. However, coop-Frontier is preferable in extreme cases because more robust to inhomogeneous corrosion distribution and imperfect information.

In the context of visual navigation, the capacity to map a novel environment is necessary for an agent to exploit its observation history in the considered place and efficiently reach known goals. This ability can be associated with spatial rea-soning, where an agent is able to perceive spatial relationships and regularities, and discover object characteristics. Recent work introduces learnable policies parametrized by deep neural networks and trained with Reinforcement Learning (RL). In classical RL setups, the capacity to map and reason spatially is learned end-to-end, from reward alone. In this setting, we introduce supplementary supervision in the form of auxiliary tasks designed to favor the emergence of spatial perception capabilities in agents trained for a goal-reaching downstream objective. We show that learning to estimate metrics quantifying the spatial relationships between an agent at a given location and a goal to reach has a high positive impact in Multi-Object Navigation settings. Our method significantly improves the performance of different baseline agents, that either build an explicit or implicit representation of the environment, even matching the performance of incomparable oracle agents taking ground-truth maps as input. A learning-based agent from the literature trained with the proposed auxiliary losses was the winning entry to the Multi-Object Navigation Challenge, part of the CVPR 2021 Embodied AI Workshop.

We introduce a new Multi-Agent Path Finding (MAPF) problem which is motivated by an industrial application. Given a fleet of robots that move on a workspace that may contain static obstacles, we must find paths from their current positions to a set of destinations, and the goal is to minimise the length of the longest path. The originality of our problem comes from the fact that each robot is attached with a cable to an anchor point, and that robots are not able to cross these cables. We formally define the Non-Crossing MAPF (NC-MAPF) problem and show how to compute lower and upper bounds by solving well known assignment problems. We introduce a Variable Neighbourhood Search (VNS) approach for improving the upper bound, and a Constraint Programming (CP) model for solving the problem to optimality. We experimentally evaluate these approaches on randomly generated instances.

In this paper, we study existing flocking models and propose extensions to improve their abilities to deal with environments having obstacles impacting the communication quality. Often depicted as robust systems, there is yet a lack of understanding how flocking models compare and how they are impacted by the communication quality when they exchange control data. We extend two standard models to improve their ability to stay connected while evolving in environments with different obstacles distributions. By taking into account the radio propagation, we model the obstacles impact on communications in a simulator that we use to optimize flocking parameters. The simulation results show the efficiency of the proposed models and how they adapt to different environmental constraints.

Routing delivery vehicles to serve customers in dynamic and uncertain environments like dense city centers is a challenging task that requires robustness and flexibility. Most existing approaches to routing problems produce solutions offline in the form of plans, which only apply to the situation they have been optimized for. Instead, we propose to learn a policy that provides decision rules to build the routes from online measurements of the environment state, including the customers configuration itself. Doing so, we can generalize from past experiences and quickly provide decision rules for new instances of the problem without re-optimizing any parameters of our policy. The difficulty with this approach comes from the complexity to represent this state. In this paper, we introduce a sequential multi-agent decision-making model to formalize the description and the temporal evolution of a Dynamic and Stochastic Vehicle Routing Problem. We propose a variation of Deep Neural Network using Attention Mechanisms to learn generalizable representation of the state and output online decision rules adapted to dynamic and stochastic information. Using artificially-generated data, we show promising results in these dynamic and stochastic environments, while staying competitive in deterministic ones compared to offline classical heuristics.

An important goal of research in Deep Reinforcement Learning in mobile robotics is to train agents capable of solving complex tasks, which require a high level of scene understanding and reasoning from an egocentric perspective. When trained from simulations, optimal environments should satisfy a currently unobtainable combination of high-fidelity photographic observations, massive amounts of different environment configurations and fast simulation speeds. In this paper we argue that research on training agents capable of complex reasoning can be simplified by decoupling from the requirement of high fidelity photographic observations. We present a suite of tasks requiring complex reasoning and exploration in continuous, partially observable 3D environments. The objective is to provide challenging scenarios and a robust baseline agent architecture that can be trained on mid-range consumer hardware in under 24h. Our scenarios combine two key advantages: (i) they are based on a simple but highly efficient 3D environment (ViZDoom) which allows high speed simulation (12000fps); (ii) the scenarios provide the user with a range of difficulty settings, in order to identify the limitations of current state of the art algorithms and network architectures. We aim to increase accessibility to the field of Deep-RL by providing baselines for challenging scenarios where new ideas can be iterated on quickly. We argue that the community should be able to address challenging problems in reasoning of mobile agents without the need for a large compute infrastructure. Code for the generation of scenarios and training of baselines is available online at the following repository1.1https://github.com/edbeeching/3d_control_deep_rl.

Current Navigation Among Movable Obstacles (NAMO) algorithms focus on finding a path for the robot that only optimizes the displacement cost of navigating and moving obstacles out of its way. However, in a human environment, this focus may lead the robot to leave the space in a socially inappropriate state that may hamper human activity (i.e. by blocking access to doors, corridors, rooms or objects of interest). In this paper, we tackle this problem of "Social Placement Choice" by building a social occupation costmap, built using only geometrical information. We present how existing NAMO algorithms can be extended by exploiting this new cost map. Then, we show the effectiveness of this approach with simulations, and provide additional evaluation criteria to assess the social acceptability of plans.

In this paper, we investigate the problem of optimizing the network performance of a fleet of unmanned aerial vehicles (UAVs) in static positions. More precisely, we allow each UAV to change its orientation in order to improve the quality of communication with its neighbours. This form of controlled mobility takes advantage of the effective radiation pattern of each UAV. We build a decentralized scheme based on the hill climbing optimization approach without a priori knowledge of the antennas radiation patterns. Then, we propose a simulation framework, based on ns--3, allowing to evaluate the gain in network performance. We provide results in several deployment scenarios involving different rate adaptation algorithms and network sizes.

This paper addresses the problem of exploring unknown terrains with a fleet of cooperating aerial vehicles. We present a novel decentralized approach which alternates gradient-free stochastic optimization and a frontier-based approach. Our method allows each robot to generate its trajectory based on the collected data and the local map built integrating the information shared by its teammates. Whenever a local optimum is reached, which corresponds to a location surrounded by already explored areas, the algorithm identifies the closest frontier to get over it and restarts the local optimization. Its low computational cost, the capability to deal with constraints and the decentralized decision-making make it particularly suitable for multi-robot applications in complex 3D environments. Simulation results show that our approach generates feasible trajectories which drive multiple robots to completely explore realistic environments. Furthermore, in terms of exploration time, our algorithm significantly outperforms a standard solution based on closest frontier points while providing similar performances compared to a computationally more expensive centralized greedy solution.

Automotive systems must undergo a strict process of validation before their release on commercial vehicles. With the increased use of probabilistic approaches in autonomous systems, standard validation methods are not applicable to this end. Furthermore, real life validation, when even possible, implies costs which can be obstructive. New methods for validation and testing are thus necessary. In this paper, we propose a generic method to evaluate complex probabilistic frameworks for autonomous driving. The method is based on Statistical Model Checking (SMC), using specifically defined Key Performance Indicators (KPIs), as temporal properties depending on a set of identified metrics. By studying the behavior of these metrics during a large number of simulations via our statistical model checker, we finally evaluate the probability for the system to meet the KPIs. We show how this method can be applied to two different subsystems of an autonomous vehicle: a perception system and a decision-making approach. An overview of these two systems is given to understand related validation challenges. Extensive validation results are then provided for the decision-making case.

Deploying fleets of mobile robots in real scenarios and environments raises several scientific challenges. One of them concerns the ability of the robots to adapt to the dynamics of their environment. We introduce DynFloR , a dynamic network flow based approach for finding optimal policies for data delivery in multi-robot network patrolling where the robots can communicate instantly and free of charge one to another when they meet, there is a periodicity of the robot meetings and the distribution of the data collected during the patrol is regular. Experiments on randomly generated synthetic examples are performed for evaluating the performance of the DynFloR method. The performed experiments empirically show that independent of the problem setting (such as number of robots, memory of the robots) the amount of data transferred to a base station per unit of time converges to an equilibrium state. The case of lost data has been also examined through various experiments, but it requires further experimentation as well as in-depth analysis.

With the rise of the complexity of the IEEE 802.11 standard, rate adaptation algorithms have to deal with a large set of values for all the different parameters which impact the network throughput. Simple trial-and-error algorithms can no longer explore solution space in reasonable time and smart solutions are required. Most of the WiFi controllers rely on proprietary code and the used rate adaptation algorithms in these controllers are unknown. Very few WiFi controllers provide their rate adaptation algorithms when they do not rely on the Minstrel-HT algorithm, which is implemented in the Linux kernel. Intel WiFi controllers come with their own rate adaptation algorithms that are implemented in the Intel IwlWifi Linux Driver which is open-source. In this paper, we have reverse-engineered the Intel rate adaptation mechanism from the source code of the IwlWifi Linux driver, and we give, in a comprehensive form, the underlying rate adaptation algorithm named Iwl-Mvm-Rs. We describe the different mechanisms used to seek the best throughput adapted to the network conditions. We have also implemented the Iwl-Mvm-Rs algorithm in the ns-3 simulator. Thanks to this implementation, we can evaluate the performance of Iwl-Mvm-Rs in different scenarios (static and with mobility, with and without fast fading). We also compare the performances of Iwl-Mvm-Rs with the ones of Minstrel-HT and IdealWifi, also implemented in the ns-3 simulator.

As drivers approach a road intersection, they must decide whether to cross it or to come to a stop. For this purpose, drivers make a situation assessment and adapt their behaviour accordingly. When this task is performed by a computer, the available information is partial and uncertain. Any decision requires the system to use this information as well as taking into account the behaviour of other drivers to avoid collisions. Common metrics such as collision rate can remain low in an interactive environment because of other driver's actions. Consequently, evaluation metrics must depend on other driving aspects. In this paper a decision-making mechanism and metrics to evaluate such a system at road intersection crossing are presented. For the former, a Partially Observable Markov Decision Process is used to model the system with respect to uncertainties in the behaviour of other drivers. For the latter, different key performance indicators are defined to evaluate the resulting behaviour of the system with different configurations and scenarios. The approach is demonstrated within an automotive grade simulator. It has showed at times, that whilst the vehicle can cross safely the intersection, it might not satisfy other key performance indicators related to highway code.

When approaching a road intersection, drivers consider several factors and choose amongst different likely manoeuvres. For an autonomous agent, it is fundamental to understand what other drivers are doing before deciding their own manoeuvres. These are seldom be the same as intersections differ and the situations too. Whilst, learning techniques can be used to process features of trajectories and to predict manoeuvres of others cars. The problem with such approaches is the difficult process of recording data for each intersection, not only of the subject vehicle but of the other vehicles. To address this problem, a hybrid data set was constructed. It is built in a simulated environment and completed with real data after has driven multiple times across the intersection. To analyze these data, classification technique is used to find the common range of features for each manoeuvre. Random forest classifiers are used in conjunction with our functional discretization to analyze the trajectories of cars approaching an intersection. The classifiers can determine the longitudinal manoeuvre as well as the direction. We show how our approach performs compared to other classifiers and space discretization. In addition, we demonstrate the impact and the usefulness of the mixture between simulated and real data. An improvement of 30% accuracy is obtained with the hybrid data set, and 5% using our functional discretization with respect to baseline approach.

Multi-robots systems require dedicated tools and models for their design and the deployment. Our approach proposes service-oriented architecture that can simplify the development and deployment. In order to solve the problem of neighbors and service discovery in an ad-hoc network, the fleet robot needs a protocol that is able to constantly discover new robots in its coverage area. To this end we propose a robotic middleware, SDfR, that is able to provide service discovery. This protocol is an extension of the Simple Service Discovery Protocol (SSDP) used in Universal Plug and Play (UPnP) to dynamic networks generated by the mobility of the robots. Even if SDfR is platform independent, we propose a ROS (ROS, 2014) integration in order to facilitate the usage. We evaluate a series of overhead benchmarking across static and dynamic scenarios. We also present some use-cases where our proposal was successfully tested.

This paper aims to use a fleet of mobile robots, each embedding a camera, to optimize the observation of a human dynamic scene. The scene is defined as a sequence of activities, performed by a person in a same place. Mobile robots have to cooperate to find a spatial configuration around the scene that maximizes the joint observation of the human pose skeleton. It is assumed that the robots can communicate but have no map of the environment and no external localisation. This paper presents a concentric navigation topology allowing to keep easily each robot camera towards the scene. This topology is combined with an incremental mapping of the environment in order to limit the complexity of the exploration state space. We also introduce the marginal contribution of each robot observation, to facilitate stability in the search, while the exploration is guided by a meta-heuristics. We developped a simulator that uses skeleton data from real human pose captures. It allows to compare the variants of the approach and to show its features such as adaptation to the dynamic of the scene and robustness to the noise in the observations.

Patrolling is mainly used in situations where the need of repeatedly visiting certain places is critical. In this paper, we consider a deployment of a wireless sensor network (WSN) that cannot be fully meshed because of the distance or obstacles. Several robots are then in charge of getting close enough to the nodes in order to connect to them, and perform a patrol to collect all the data in time. We discuss the problem of multi-robot patrolling within the constrained wireless networking settings. We show that this is fundamentally a problem of vertex coverage with bounded simple cycles (CBSC). We offer a formalization of the CBSC problem and prove it is NP-hard and at least as hard as the Traveling Salesman Problem (TSP). Then, we provide and analyze heuristics relying on clusterings and geometric techniques. The performances of our solutions are assessed in regards to networking parameters, robot energy, but also to random and particular graph models.

In ambient intelligence, pressure sensing can be used for detecting and recognizing objects based on their load profile. This paper presents a pressure scanning technique that improves weight-based object recognition, by adding information about the surface of the object in contact with the floor. The new high-resolution pressure scanning technique employs sub-pixel shifting to assemble a series of low-resolution scans into an aggregated high-resolution scan. The proposed scanning device is composed of 4 load-sensing tiles, on which the scanned object slides in regular movements. The result is a regular grid image of the object's contact surface, containing the weight of each section of the grid, as well as the corresponding centers of mass. A formal proof-of-concept is provided, together with experimental results obtained both on a noiseless simulated platform, and on a noisy physical platform.

Traffic management in large networks remains an important challenge in transportation systems. The best approach would be to use existing infrastructure and find a solution to manage the increasing flows of vehicles. Multi-agent systems and autonomous vehicles are today considered as a promising approach to deal with traffic control. In this paper, we propose a two-level decentralized multi-agent system which allows autonomous vehicles crossing the network intersections without stopping. At the first level, we use a control agent at each intersection which (1) lets the vehicles from each road pass alternately, and (2) allows them to optimally regulate their speed in its vicinity. At the second level, each agent coordinates with its neighboring agents in order to optimize the flows inside the network. We evaluate this approach empirically, with a comparison with a more opportunistic First-Come First-Served strategy. Experimental results (in simulation) are presented (measuring energy consumption), showing the advantages and disadvantages of each approach.

This article addresses the problem of computing a Voronoi diagram in a distributed fashion without any synchronization heuristic. To our knowledge, no previous work asynchronously solves this problem. We investigate a simple case of a synchronism and tackle this challenge in a decentralized fashion on a grid of communicating cells with a von Neumann neighborhood. We describe algorithms for extracting single site, area and pseudo line Voronoi diagrams. These algorithms are implemented and executed on maps in which we consider different kinds of sites defined as simple polygonal shapes to extract roadmaps.

In real-world multi-agent systems, as in the context of the automatic transportation of goods, autonomous vehicles can face unexpected events like the failure of a vehicle, the presence of obstacles on the road, etc. Such events can generate first local congestions, and then, if they persist, global phenomena and complex traffic congestions (such as traffic jams). We want to manage space sharing conflicts at the local level, when they appear, to allow a quick (real-time) regulation, i.e., without requiring to re-plan the routes of all involved agents. Our approach relies on reactive coordination between vehicles using simple interactions between neighboring agents, using perceptions and little or no communication. We consider in particular a scenario where two queues of vehicles share a single lane, describing the model of the network as well as the agents, and proposing simple coordination rules that only involve the two vehicles at the front of each queue. We then conduct experiments that allow the analysis and the comparison of the proposed self-regulation rules.

In this paper we propose a new experimental device for defining and studying self-organized systems, especially those including physical or chemical interactions such as those encountered in collective natural phenomena. We want to be able to reproduce with real robots several paradigms such as stigmergy. In these phenomena the environment stores, diffuses, evaporates chemical substances (pheromones) that drive the behavior of each entity. The proposed device is a smart surface which relies on a graphical environment on top of which robots can move but also read/write information thanks colorimetric sensors and infrared emitters. The surface itself is able to perform some computation, implementing e.g. diffusion/evaporation mechanisms. More generally, the proposed robotic system allows to re-examine theoretical/simulated models in the perspective of defining self-organized robots. We consider in this paper the foraging problem as a case study. In particular we re-examine the expression of the model proposed by Drogoul&Ferber to implement pheromone-based exploration and transport with robots. We then analyze self-organized behaviors, as emergence of chains of robots, and their robustness.

This paper considers the platooning problem: we aim to steer a train of vehicles along an unknown path generated by the first vehicle, which is human driven. Among existing approaches, we study a decentralised local approach to avoid robustness issues due to communication failure which are common to centralised approaches. However, decentralised control rises up the problem of lateral deviation which is accumulated along the platoon.

This paper deals with the platooning problem that can be defined as the automatic following of a manned driven vehicle by a convoy of automatic ones. Different approaches have been proposed so far. Some require the localisation of each vehicle and a communication infrastructure, others called near-to-near approach only needs vehicle on-board sensors. However, to our knowledge, they do not provide any proof of non collision. We propose a novel near-to-near longitudinal platooning building a collision-free platooning whatever the number of vehicles. The model is derived from the study of the most dangerous interaction between two vehicles, i.e. considering the maximum acceptable acceleration when the previous vehicles brakes at maximum capacity. Collision avoidance of this model is proved. Finally, we show that this model can be combined to existing ones, keeping this collision-free property while allowing more various behaviors.

This paper addresses the multi-agent patrolling problem, which consists for a set of autonomous agents to visit all the places of an unknown environment as regularly as possible. The proposed approach is based on the ant paradigm. Each agent can only mark and move according to its local perception of the environment. We study EVAW, a pheromone-based variant of the EVAP [3] and VAW [12]. The main novelty of the paper is the proof of some emergent spatial properties of the proposed algorithm. In particular we show that obtained cycles are necessarily of same length, which ensures an efficient spatial distribution of the agents. We also report some experimental results and discuss open questions concerning the proposed algorithm.

This paper deals with the multi-agent patrolling problem in unknown environment using two collective approaches exploiting environmental dynamics. After specifying criteria of performances, we define a first algorithm based only on the evaporation of a pheromone dropped by reactive agents (EVAP). Then we present the model CLInG [10] proposed in 2003 which introduces the diffusion of the idleness of areas to visit. We systematically compare by simulations the performances of these two models on growing- complexity environments. The analysis is supplemented by a comparison with the theoretical optimum performances, allowing to identify topologies for which methods are the most adapted.

This article deals with the problem of dynamic role-playing in Multi-Agent organisations. The approach presented uses a formal specification notation and is based upon a formal framework which defines the concepts of role, interaction and organisation. Within this framework the problem of dynamic role-playing specification is related to the merging of specifications. The formal notation used composes Object-Z and Statecharts. The main features of this approach are: enough expressive power to represent Multi-Agents dynamic aspects, tools for specification analysis and mechanisms allowing the refinement of a high level specification into a low level specification which can be easily implemented. The last part of this paper presents an application with the specification of a reactive and cooperative MAS model named Satisfaction Altruism. An analysis of the specification validates the agents' behaviours.

We have developed an agent's architecture towards the goal of building efficient, robust and safe multi-robot systems considered as cooperating distributed reactive agents. This architecture is based on satisfaction and altruism allowing the agents to amend their low-level behavior like goal seeking and collision avoidance in order to solve more complex problems. We demonstrate in particular that local conflicting and locking situations are automatically avoided or made repulsive. Computer simulations of tasks in complex environments confirm it. The designed mini-robots, the implementation of their architecture, and the communication protocol are described. The same hardware is shared between communication, collision avoidance, and task achievement. Experiments using two mobile robots and a test bed confirm the theoretical and simulation results.

This paper shows how a probabilistic model is able to predict the evolution of most multi-robot systems and thus to save a lot of simulation time. To demonstrate the performance of this approach, a complex heterogeneous system is considered. It is composed of two populations of robots, having different but complementary abilities. They must survive by finding supply centers in the environment. It is shown how to model the process by a stochastic Petri net and its associated Markov chain. The latter allows one to compute the time evolution of the system. The process includes several sink states, which correspond to a singular problem. However, comparisons of simulation and theoretical results show very close values of the state probabilities when the agents are initially located at random positions. The number of agents is varied in order to obtain the most favorable terminal state: the Markovian analysis is shown to help one to determine the best parameters. Finally, the hardware used in experiments is described.