Invited Lecture: Advanced Control and Learning Theory for Robotic Applications

Abstract: Motion planning and control are essential parts of most robotic systems and have been studied extensively in both control and robotics communities. Recent advances in robotic applications pose many non-classical challenges that cannot be directly addressed using existing control tools. One common challenge lies in the interaction between continuous dynamics with discrete logic rules, making the overall system exhibit hybrid dynamic behavior. Another challenge lies in the interaction among multiple decision makers. In this talk, I will briefly go over our recent works on optimal control of hybrid systems, multiagent systems, and reinforcement learning that can potentially address the aforementioned non-classical challenges. The results will be presented with an emphasis on their applications in robotics and human-robot collaborative decision problems.