We also study challenging decision and control problems in multi-robot systems. In this area, we mainly consider two classes of problems which can be viewed as dual to each other. One problem is on cooperative pursuit, namely, developing control strategies for a team of agents to capture an evader.
This is a classic problem in robotics with rich applications in surveillance, search and rescue, battle field automation, etc. It is essentially a multiagent adversarial game whose optimal solution is intractable due to the curse of dimensionality. We developed a framework based on open-loop reachability that can provide efficient solutions with guaranteed capture. The algorithms have been tested and used in human-automation team systems, and Game AI, etc.
We have also studied the multiagent collision avoidance problem, which develops decentralized planning/control strategy for each mobile robot to reach its desired target set/location while avoiding collisions with obstacles and other agents in the area. This is also a classic robotic problem, which has recently gained a renewed interest due to the popularity of service robots and AGVs. Our approach utilizes the state-of-the-art computer vision algorithms to track nearby moving obstacles and use reinforcement learning to design the collision avoidance strategies.