Gazebo ROS-based BD1 droid from Star Wars: The Fallen Order game. For testing reinforcement learning algorithms.
- Install ROS Noetic
- Install additional ROS packages
sudo apt install ros-noetic-velocity-controllers
-
If you want to use GPU, install CUDA and cuDNN, here is guide need to be followed until 'Finally to verify the correct installation'
-
Install pytorch
-
Install stable-baselines3
-
Clone this repo in your workspace and build it
- bd1_config - main launches
- bd1_description - robot model
- bd1_gazebo_env_interface - utils to wrap Gazebo in Gym environment
- bd1_gazebo_utils - some Gazebo additional utilities like contact handler
- bd1_manual_control - manual robot control for testing
- bd1_simple_moves - some handmade movements like deploy\undeploy
- bd1_train_sb3 - nodes implementing learning with stable-baselines3
roslaunch bd1_config bd1_gazebo.launch
Wait until Gazebo starts, then
roslaunch bd1_config bd1_train_sb3.launch
Run tensorboard with
tensorboard --logdir bd1_train_sb3/models/...