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

Published on 2012-10-165574 Views

Evangelos Theodorou

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

Workshop on Statistical Physics 2012 - Granada

Related categories

Presentation

Free Energy and Relative Entropy Dualities: Connections to Path Integral Control and Applications to Robotics00:00

What makes robotic control difficult? - 103:11

What makes robotic control difficult? - 204:58

What makes robotic control difficult? - 304:59

What makes robotic control difficult? - 405:00

What makes robotic control difficult? - 505:01

Outline05:02

Dynamic Programming - 106:06

Dynamic Programming - 206:37

The Linear Bellman PDE07:30

Value Function Approximation/Policy Gradient Methods09:02

Policy Improvement with Path Integrals (PI2) - 112:49

Policy Improvement with Path Integrals (PI2) - 215:09

Policy Improvement with Path Integrals (PI2) - 317:11

Policy Improvement with Path Integrals (PI2) - 419:15

Policy Improvement with Path Integrals (PI2) - 519:16

Policy Improvement with Path Integrals (PI2) - 619:17

Policy Improvement with Path Integrals (PI2) - 721:18

Policy Improvement with Path Integrals (PI2) - 821:30

Policy Improvement with Path Integrals (PI2) - 921:31

Optimal Planning and Control - 122:20

Optimal Planning and Control - 222:21

Optimal Planning and Control - 322:22

Policy Improvement with Path Integrals (PI2) - 1022:57

Policy Improvement with Path Integrals (PI2) - 1124:01

Policy Improvement with Path Integrals (PI2) - 1225:00

Policy Improvement with Path Integrals (PI2) - 1325:06

Policy Improvement with Path Integrals (PI2) - 1425:14

Policy Improvement with Path Integrals (PI2) - 1525:15

Policy Improvement with Path Integrals (PI2) - 1625:15

Policy Improvement with Path Integrals (PI2) - 1725:16

Trajectory Optimization - 125:56

Applications to Robotics27:53

Trajectory Optimization and Gain Scheduling30:36

Trajectory Optimization - 232:55

Trajectory Optimization - 335:56

Trajectory Optimization - 438:57

Tendon Driven System - 140:42

Tendon Driven System - 241:11

Tendon Driven System - 341:37

Applications to Tendon Driven Robots - 142:59

Applications to Tendon Driven Robots - 246:24

Applications to Tendon Driven Robots - 349:36

The Information Theoretic View - 151:44

The Information Theoretic View - 253:25

The Information Theoretic View - 353:26

Application to Jump Diffusions53:29

Summary55:07