Learning applied to multi-robot systems

Cooperation in multi-robot systems is a challenging task. The desired outcome for the group must be achieve through agents that often perceive/act based on their own point of view and interests. To achieve a certain form of “agreement” between the agents, different algorithms have been studied, such as network control, auction, or optimization methods. These can be computed by a central entity, which simplifies modeling the problem, or be inherently decentralized, which scales well for large number of agents. In this presentation we explore the techniques that have been used to coordinate the action of multiple intelligent and autonomous agents. Also, we see how machine learning techniques (RL) or the state-of-the-art language models (e.g., ChatGPT) can be applied to improve system’s performance. ————————————————————— Dr. Kleber Cabral holds a doctorate from the Royal Military College of Canada (RMC, 2022). He is currently a Postdoctoral Fellow and an Adjunct Professor at Queen’s University. His research involves robotic and applied AI. He focused on multi-robotic systems, and the problem within, such as control and decision making. In the past, Dr. Cabral has investigated the use of network control, graph theory for decentralized autonomous construction problems. Currently, his research involves applying machine learning to multi-robot problems, exploring state-of-the-art modelling and solutions based on game theory and learning using language models. Speaker(s): Kleber Cabral, Virtual: https://events.vtools.ieee.org/m/502702