The Panda robot model was made from these files.
We are trying to solve some interesting problems with reinforcement learning and finally deploy the models in the real world.
- Install Webots R2021a
- Install Python versions 3.8
- Follow the Using Python guide provided by Webots
- Install deepbots 0.1.3.dev2 through pip running the following command:
pip install -i https://test.pypi.org/simple/ deepbots - Install PyTorch via pip
The goal is to train an agent to reach the randomly selected goal within limited steps.
Here, the problem is solved with the Deep Deterministic Policy Gradient RL algorithm. The agent observes its seven motor positions and the Cartesian coordinates of the selected goal, and then controls the seven motor positions.
| Trained Agent Showcase | Reward Per Episode Plot |
|---|---|
 |
 |
This project is part of the System Integration Implementation teamwork, an undergraduate course supervised by Prof. Chih-Tsun Huang of Dept. of Computer Science, National Tsing Hua University.
We thank Manos Kirtas and Kostas Tsampazis, deepbots maintainers, for their help and feedback.
The Panda robot model was contributed by all the team members, Yung Tai Shih, Tsu Hsiang Chen, Chun Kai Yang, Yan Feng Su, Hsuan Yu Liao, and the author of this deepworlds Panda example, Jiun Kai Yang.
- Find more examples in deepworlds.
- Reach a Target via PPOAgent with Panda