We developed an open-source system that simplifies the training of robots to perform tasks through imitation learning. Our setup allows you to collect data, control robots, and train models in both real and simulated environments.
The system is built on the Diffusion Policy model and the ROS2 framework. To help you get started, we provide a pre-trained model and datasets, but you can also collect your own data for customized training.
We believe that imitation learning can be useful in dynamic environments where object positions can change, and our project aims to simplify this imitation learning process
We invite you to explore its capabilities and contribute to its growth! If you’re interested in training robots to perform tasks using imitation learning, check it out and let us know what you think!