This project is a complete pybullet robotc arm examples using a UR5 and reinforcment learning based on continuous reward PPO.
pip3 install pybullet, attrdict
Train neural network model with:
python3 train_rl.py
Use --render to display GUI. See code for more options.
Typical output:
pybullet build time: Feb 4 2021 14:39:03
Using device: cpu , device number: 0 , GPUs in system: 0
Environment name: PyBullet UR5 robot
Starting training with learning_param: 0.1
Episode 100 Avg length: 48 Avg reward: -141
...
Episode 1000 Avg length: 37 Avg reward: -83
...
Episode 2500 Avg length: 36 Avg reward: -81
Demo with:
python3 demo_rl.py /path_trained/model.pth --render