Isaac sim environment to train the BDX robot

This repository contains the code to train the BDX robot to walk on flat or rough terrains using Isaac Lab Isaac Lab. The model of the robot was found on the AWD repository and the idea of this work comes from the Open_Duck_Mini repository. The robot is trained using the RSL-RL implementation of RL algorithms. This repository was created using the IsaacLab template.

bdx.mp4

Repository structure and important files

The repository contains the configuration files for a flat environment and a rough environment. We're training the robot to perform a task which in this case is a velocity task. The robot is trained to walk with a certain velocity. The configuration files define the environment and the task for the robot. The configuration files are located in the exts/bdx_walk/bdx_walk/tasks/locomotion/velocity folder. The repository is structured as follows:

    ├── data
    ├── exts/bdx_walk/bdx_walk/tasks/locomotion/velocity
    |   ├── velocity_env_cfg.py        # Configuration file for the velocity task
    │   ├── config/bdx/          
    │   │   ├── flat_env_cfg.py        # Configuration file for the flat terrain environment
    │   │   └── rough_env_cfg.py       # Configuration file for the rough terrain environment
    └── scripts/rsl_rl
        ├── train.py            # Train the BDX robot
        └── play.py             # Play the trained model

How to start training

Install Isaac Lab.

Install the package:

python -m pip install -e exts/bdx_walk/.

Train the model for the flat terrain using the following command:

python scripts/rsl_rl/train.py --task Isaac-Velocity-Flat-BDX-v0  --headless --num_envs 4096

During training, for debugging purposes it can be useful to output the full logs:

export HYDRA_FULL_ERROR=1

Visualize training information

For visualizing the training information, use tensorboard. To start tensorboard, run the following command:

tensorboard --logdir logs/rsl_rl/bdx_flat/

It will start a server on localhost:6006 where you can visualize the training information for the flat terrain.

Play the trained model

The training process will save checkpoints (pt files) inside the logs folder. To play the trained model, use the following command:

python scripts/rsl_rl/play.py --task Isaac-Velocity-Flat-BDX-Play-v0 --num_envs 32 --checkpoint [/PATH/TO/model.pt]