Free Energy and Relative Entropy Dualities: Connections to Path Integral Control and Applications to Robotics

While optimal control and reinforcement learning are fundamental frameworks for learning and control applications, their application to high dimensional control systems of the complexity of humanoid a