Cooperative multi-agent machine learning for swarm robotics

Multi-Agent Systems have become essential for solving complicated real-world problems where a single agent is no longer sufficient, due to their ability to either compete, cooperate, or both to solve these complicated problems and achieve a better performance. With the recent advancements of computing, communication, sensing, and control techniques, a large number of autonomous mobile robots connected via a communication network can be guided to achieve a common objective. In this project, intelligent coordination methods such as multi-agent deep reinforcement learning and distributed computing will be used to develop an advanced path-planning and control framework for networked swarm robotics. The effectiveness of the theoretical results will be validated by hardware experiments using real mobile robots.