/blackbird_walking

Training a policy to make a bipedal robot walk in a Gazebo ROS2 enviornment with Reinforcement Learning

Primary LanguageC++

BlackBird Walking

this project is dedicated to creating a reinforcement learning environment (see: gz_blackbird) in order to create a policy to make the blackbird bipedal robot walk in a Gazebo enviornment using ros2_control as the middleware.

blackbird in gazebo simulation

see the separate directory descriptions below:

gz_blackbird

This is the training environment developed with Gazebo Harmonic. Written in C++, it features a python API for the gym enviornments. Please see BalanceEnv.py for a static balance environment, and BlackbirdEnv.py for the walking environment.

ai_model

this is where the AI models are developed and training is run in. try train_balance.py or test_balance.py to test the model

realtime_ws

this is the ros2 workspace dedicated for the test environment. see PosePublisher.cpp to see the topic created to publish the robot's state. Make sure gz_ros2_control is installed by running the following command:

sudo apt install ros-jazzy-gz-ros2-control