Software implementation of RL-augmented MPC for quadruped locomotion. See experiment video(https://youtu.be/HxSIxTnEw08) In this repo, two setups for RL-augmented MPC are implemented. One is foot placement offset to the Raiber heuristic and the other is joint angle offsets to the nominal swing trajectory (reported in the paper). Both setups are deployable to the robot and achieve similar performances.
- training_module Pybullet training environment for the project
- experiment_code Experiment code to run the trained policies
Note: The training module doesn't interface directly with hardware !!!
Go into README files of each folder for details
- code cleanup
- documentation
- Credits and Acknowledges to be added later