Index of Embodied AI Reference Projects

This page lists open-source projects suitable as extended reading for courses. When using reference projects, you should first check the licenses, dependency versions, and hardware requirements, and then integrate their interfaces into your own experimental environment.

Matrix

• Project address: https://github.com/zsibot/matrix
• Key points to note: project directory organization, robot task interface, and the connection method between the model and control modules.
• Reading approach: First, review the project description and minimal examples, then locate the environment definition, action space, and evaluation entry. A separate environment should be created before running the example to avoid relying on version details that might affect the main course environment.

Project Filtering List

When introducing a new reference item, record at least the following information:

1. The robot tasks addressed by the project, along with their inputs and outputs;
2. Supported simulation engines, robot bodies, and operating systems;
3. Separate interfaces for training, inference, and evaluation;
4. Dataset, model weights, and license requirements;
5. Expected output of the minimum running example.

After organizing according to this list, the reference items can establish a clear correspondence with the environment setup, policy training, and evaluation sections in the textbook.