jonzia / Robotarium

Training networked agents to perform herding tasks using reinforcement learning in the GT Robotarium (MATLAB).

Robotarium

Training autonomous agents to perform herding tasks using reinforcement learning.

User Guide (MATLAB)

1. Clone the Robotarium MATLAB repository.
2. Run init.m in the Robotarium MATLAB simulator folder.
3. To train a model using reinforcement learning, run RL.mlx. Alternatively, a pre-trained model net.m is provided for a configuration with one leader agent and five follower agents.
4. The trained model may be tested in the Robotarium MATLAB simulator with main.m.

The video files entitled ideal_simulation.avi, robotarium_simulation.avi, and robotarium_video provide full-access video described in the text, "Training Networked Agents to Perform Herding Tasks using Reinforcement Learning."

Model Training

Computer Simulation

Robotarium Implementation