Decentralized Control and Estimation in Multi-robot Systems with Diverse Topological Requirements

This talk will give an overview of some theoretical and experimental results in the multi-robot field, with a special regard to the multi-UAV case. The major strengths of a multi-robot system are both the resilience to single point failures and the possibility of parallelizing the execution of a given task. These properties can be fully exploited in coverage-like tasks, e.g., exploration, pursuit-evasion (a.k.a. "clearing"), and periodical monitoring (a.k.a. "patrolling"). These tasks, in turn, contain several control and estimation subproblems, e.g., how to:- keep a certain optimal arrangement by using an appropriate formation controller, that should be decentralized and using of cheap and lightweight sensors; - keep some topological properties of the interaction, like the connectivity and the rigidity of the group, while still allowing for a flexible behavior of the robots, i.e., a time-varying topology, and decentralization; - mutually localize the robots in order to allow a proper fusion of the information gained by every single robot. The common and particularly challenging situation where the robot-to-robot sensor is anonymous, i.e., it does not retrieve the identity of the detected robot, will be considered; - allow the presence of one or more human co-operators in order to cope with particularly challenging tasks, where cognitive capabilities are required, e.g., in search and rescue operations. A relevant problem in this shared control case is how to balance the robot autonomy with the human assistance. - cope with the uncertainty in the robotic system and in the interacting environment in order to effectively employ actual multi-robot systems in the real-world scenarios