One key to more effective cooperative interaction in a multi-robot team is the ability to understand the behavior and intent of other robots. Observed teammate action sequences can be learned to perform trajectory recognition which can be used to determine their current task. Previously, we have applied behavior histograms, hidden Markov models (HMMs), and conditional random fields (CRFs) to perform trajectory recognition as an approach to task monitoring in the absence of commu- nication. To demonstrate trajectory recognition of various autonomous vehicles, we used trajectory-based techniques for model generation and trajectory discrimination in experiments using actual data. In addition to recognition of trajectories, we in- troduced strategies, based on the honeybee’s waggle dance, in which cooperating autonomous teammates could leverage recognition during periods of communication loss. While the recognition methods were able to discriminate between the standard trajectories performed in a typical survey mission, there were inaccuracies and delays in identifying new trajectories after a transition had occurred. Inaccuracies in recog- nition lead to inefficiencies as cooperating teammates acted on incorrect data. We then introduce the Trajectory Adaptation for Recognition (TAR) framework which seeks to directly address difficulties in recognizing the trajectories of autonomous vehicles by modifying the trajectories they follow to perform them. Optimization techniques are used to modify the trajectories to increase the accuracy of recognition while also improving task objectives and maintaining vehicle dynamics. Experiments are performed which demonstrate that using trajectories optimized in this manner lead to improved recognition accuracy.

Agents in most types of societies use information about potential partners to determine whether to form mutually beneficial partnerships. We can say that when this information is used to decide to form a partnership that one agent trusts another, and when agents work together for mutual benefit in a partnership, we refer to this as a form of cooperation. Current multi-robot teams typically have the team's goals either explicitly or implicitly encoded into each robot's utility function and are expected to cooperate and perform as designed. However, there are many situations in which robots may not be interested in full cooperation, or may not be capable of performing as expected. In addition, the control strategy for robots may be fixed with no mechanism for modifying the team structure if teammate performance deteriorates. This dissertation investigates the application of trust to multi-robot teams. This research also addresses the problem of how cooperation can be enabled through the use of incentive mechanisms. We posit a framework wherein robot teams may be formed dynamically, using models of trust. These models are used to improve performance on the team, through evolution of the team dynamics. In this context, robots learn online which of their peers are capable and trustworthy to dynamically adjust their teaming strategy. We apply this framework to multi-robot task allocation and patrolling domains and show that performance is improved when this approach is used on teams that may have poorly performing or untrustworthy members. The contributions of this dissertation include algorithms for applying performance characteristics of individual robots to task allocation, methods for monitoring performance of robot team members, and a framework for modeling trust of robot team members. This work also includes experimental results gathered using simulations and on a team of indoor mobile robots to show that the use of a trust model can improve performance on multi-robot teams in the patrolling task.

Mobile sensor networks have been shown to be a powerful tool for enabling a number of activities that require recording of environmental parameters at various spatial and temporal distributions. These mobile sensor networks could be implemented using a team of robots, usually called robotic sensor networks. This type of sensor network involves the coordinated control of multiple robots to achieve specific measurements separated by varied distances. In most formation measurement applications, initialization involves identifying a number of interesting sites to which mobility platforms, instrumented with a variety of sensors, are tasked. This process of determining which instrumented robot should be tasked to which location can be viewed as solving the task allocation problem. Unfortunately, a centralized approach does not fit in this type of application due to the fault tolerance requirements. Moreover, as the size of the network grows, limitations in bandwidth severely limits the possibility of conveying and using global information. As such, the utilization of decentralized techniques for forming new sensor topologies and configurations is a highly desired quality of robotic sensor networks. In this thesis, several distributed task allocation algorithms will be explained and compared in different scenarios. They are based on a market approach since our interest is not only to obtain a feasible solution, but also an efficient one. Also, an analysis of the efficiency of those algorithms using probabilistic techniques will be explained. Finally, the task allocation algorithms will be implemented on a real system consisted of a team of six robots and integrated in a complete robotic system that considers obstacle avoidance and path planning. The results will be validated in both simulations and real experiments.

Cette thèse présente l’étude d’une stratégie de coordination hybride d’un groupe de robots sous-marins pour la recherche d’objets de petites dimensions ou de singularités sur les fonds marins. Chaque robot est équipé d’un module de perception utilisant la librairie de traitement d’image OpenCV qui lui permet d’apercevoir les autres éléments de la meute ainsi que l’environnement d’évolution de la mission.Cette stratégie hybride est constituée de deux phases : une phase de mise en formation géométrique et une phase d’acquisition des données vidéo. La première phase s’appuie sur des algorithmes de type "essaims" alors que la seconde se fonde sur une méthode hiérarchique de coordination. En cas de perte de la formation, le groupe de robots quitte le mode hiérarchique et reprend le mode essaim pour se reformer. Ces changements de modes sont contrôlés par une machine à états finis. Avant d’entamer une expérimentation en grandeur nature, la méthodologie et les algorithmes de coordination doivent être testés et validés par simulation.Dans ce contexte, un simulateur basé sur le logiciel Blender a été conçu de façon à ce qu’il tienne compte des différentes contraintes liées à l’évolution des robots dans l’environnement sous-marin. Les résultats de simulation d’une meute de 3 AUVs montrent la capacité de notre stratégie à optimiser l’exécution d’une mission d’acquisition vidéo par un groupe de robots autonomes contrôlés par la vision et coordonnés par une stratégie hybride. / In the underwater environment, the needs of data acquisition have significantly increased over the last decades. As electromagnetic waves show poor propagation in sea water, acoustical sensing is generally preferred. However, the emergence of small and low cost autonomous underwater vehicles (AUV) allow for rethinking the underwater use of optical sensors as their small coverage can be significantly improved by using a fleet of coordinated underwater robots.This paper presents a strategy to coordinate the group of robots in order to systematically survey the seabed to detect small objects or singularities. The proposed hybrid coordination strategy is defined by two main modes. The first mode relies on a swarm algorithm to organize the team in geometrical formation. In the second mode, the robot formation is maintained using a hierarchical coordination. A finite state machine controls the high level hybrid strategy by defining the appropriate coordination mode according to the evolution of the mission. Before sea validation, the behavior and the performance of the hybrid coordination strategy are first assessed in simulation. The control of individual robots relies on visual servoing, implemented with the OpenCV library, and the simulation tool is based on Blender software.The dynamics of the robots has been implemented in a realistic way in Blender by using the Bullet solver and the hydrodynamic coeficcients estimated on the actual robot. First results of the hybrid coordination strategy applied on a fleet of 3 AUV’s, show execution of a video acquisition task by a group of autonomous robots controlled by vision and coordinated by a hybrid strategy.