ARS-Bezier

Article: Zijian Zhao, Zitao Zhang, Kai Huang, "A Trajectory-based Reinforcement Learning Approach for Autonomous Locomotion of a Rat Robot"

Experiment Platform: Robot Rat NeRmo & MuJoCo

Conference Version of RS2002/RL-Rat: Repository for Final Design, "Motion Control Method for Small Quadruped Robots Based on Trajectory Modulation and Reinforcement Learning" (基于轨迹调制和强化学习的小型四足机器人运动控制方法).

1. Train

To train the model, use the following command. You can also use --help to find useful parameters such as the learning rate.

If you want to train the model in a given scenario, use the command:

python train.py --modelPath <environment path>

If you want to try our environment randomization method, use:

python train.py --rand_env

Here are some hyper-parameters:

--max_steps: maximum steps in each round.

--max_epoch: total training epochs.

--episode_num: amount of randomly sampled points in each epoch.

--max_epoch: top-k episodes in each epoch are used to update parameters.

--learning_rate: learning rate.

--exploration_noise: sampling range.

2. Eval

To evaluate the model in a specific scenario, use the command:

python eval.py --modelPath <environment path> --parameterPath <parameter path>

3. Citation

@misc{zhao2024ARS,
  title={A Trajectory-based Reinforcement Learning Approach for Autonomous Locomotion of a Rat Robot},
  author={Zijian Zhao and Zitao Zhang and Kai Huang},
  year={2024},
  howpublished={\url{https://github.com/RS2002/ARS-Bezier/blob/main/ARS_Bezier.pdf}},
  note={Github Blog}
}