Teaching

Learning-Based Control for Robotics (ME461/561)

Learning-based methods have achieved remarkable progress in robot control, with applications spanning autonomous driving, quadrupeds, and humanoids. At the core of these advances lie optimal control and reinforcement learning (RL). This course develops the mathematical foundations of these methodologies and demonstrates how to apply them to real robotic systems. Topics range from core RL concepts to advanced algorithms (value/policy iteration, TD/Q-learning, actor–critic, DPG/DDPG, TRPO, PPO). The course emphasizes hands-on practice so you can use these tools in your own research.