The purpose of this project is to learn a pouring policy for the robot using reinforcement learning. Two pouring situations are considered in our project, including pouring all liquid in the container and pouring an accurate amount of liquid. The pouring policy is represented by a neural network, which is trained on a fluid simulation environment based on Unity. The policy network can be trained using different policy gradient algorithms. Besides, we implemented an interface which allows the pouring policy to be applied to the robot directly.
Include Fetch environment and interface to policy model.
Include A3C code and saved model.
The Unity pouring environment is uploaded to Google Drive: https://drive.google.com/file/d/1KCLKcPBpiEU7_4GAtki_Y4yCghJ7A_co/view?usp=sharing
Multi-agent training scene: https://youtu.be/rOD9qzz6RC8
Pouring: https://youtu.be/B17nsjaB2KU
Fetch simulation: https://youtu.be/te8icsWMvW0
