silvery107 / rl-mpc-locomotion

Deep RL for MPC control of Quadruped Robot Locomotion
https://docs.google.com/presentation/d/18bznpYrkCPnhCisySPDz18hvL3Ytere7JiJEbdLvpgU/edit?usp=sharing
MIT License
417 stars 47 forks source link

Observation mismatch #20

Open silvery107 opened 3 weeks ago

silvery107 commented 3 weeks ago

The observation in training is mismatched with deployment. The base_pos should be removed.

I was trying to align the RL reward to MPC cost, but it turns out it's better to go without position tracking for both stages.