Memory Focused Proximal Policy Method for Adaptive Biped Locomotion
Submitted to IROS2019
- Attention mechanisem and RNN are introduced into PPO ,which makes a superior performence in control of biped robot walking. (So it's aslo called AM-RPPO)
-
Platform
- Torch
- OpenAI gym(four robots: BipedalWalker-V2, BipedalWalkerHardcore-v2, Humanoid-V2, Walker2d-V2)
- DDPG (refer to Morvan's)
- PPO (Openai Baseline)
- RDPG (refer to Doo Re Song's)
-
Performance
- Simulations
