DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills

  • 2018-06-16 20:48:52
  • Xue Bin Peng, Pieter Abbeel, Sergey Levine, Michiel van de Panne
  • 0

Abstract

A longstanding goal in character animation is to combine data-drivenspecification of behavior with a system that can execute a similar behavior ina physical simulation, thus enabling realistic responses to perturbations andenvironmental variation. We show that well-known reinforcement learning (RL)methods can be adapted to learn robust control policies capable of imitating abroad range of example motion clips, while also learning complex recoveries,adapting to changes in morphology, and accomplishing user-specified goals. Ourmethod handles keyframed motions, highly-dynamic actions such asmotion-captured flips and spins, and retargeted motions. By combining amotion-imitation objective with a task objective, we can train characters thatreact intelligently in interactive settings, e.g., by walking in a desireddirection or throwing a ball at a user-specified target. This approach thuscombines the convenience and motion quality of using motion clips to define thedesired style and appearance, with the flexibility and generality afforded byRL methods and physics-based animation. We further explore a number of methodsfor integrating multiple clips into the learning process to developmulti-skilled agents capable of performing a rich repertoire of diverse skills.We demonstrate results using multiple characters (human, Atlas robot, bipedaldinosaur, dragon) and a large variety of skills, including locomotion,acrobatics, and martial arts.

 

Quick Read (beta)

loading the full paper ...