A functional and streamlined codebase for locomotion training in Genesis, featuring:
- Effective domain randomizations and useful rewards
- Fundamental rewards for executing a complete backflip
We use the same domain randomizations as in this repository in both single and double backflips. During deployment, we slightly changed Kps used on the real robot to do a clean backflip.
-
Install the modified version rsl_rl
cd rsl_rl && pip install -e . -
Train the Unitree Go2 for a backflip
This codebase includes only the fundamental rewards required for learning a backflip. You can observe how Go2 learns to perform a full backflip by running:
python train_backflip.py -e EXP_NAME -
Evaluate existing checkpoints
Run
python eval_backflip.py -e EXP_NAME --ckpt NUM_CKPT -
Generate a deployable checkpoint
To produce a deployable checkpoint comparable to the provided example in Genesis, additional rewards must be incorporated to regularize motion and minimize torque peaks.