This repository leverages the Genesis Simulator to perform reinforcement learning for developing control agents in a human-robot interaction context. The project focuses on designing RL-based policies to control:
- Robot Agent: To execute human-robot interactive movements that aim to teach or transfer knowledge to humans safely (e.g., boxing training).
- Humanoid Agent: To serve as a "virtual assistant," demonstrating the correct sequences of movements to humans based on the robot's actions and the human's posture.
The ultimate goal is to enable collaborative and interactive tasks between robots and humans, emphasizing safety (e.g., collision avoidance) and effective skill transfer.