Visual-Policy Deployment on Real Unitree GO1 Robot Dog

• This project used the following two packages: unitree_ros2 and vision4leg
• The unitree_ros2 package comes from Katie Hughes and Nick Morales.
• The Vision4Leg package comes from Mehooz, the official implementation for the paper Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers and Vision-Guided Quadrupedal Locomotion in the Wild with Multi-Modal Delay Randomization.

Notes

The execution loop is called in execute_locotransformer.py →executor.execute(EXECUTION_TIME) “Execute” function is defined in control_loop_execution.main_executor In main_executor.execute function: main_execution function is called, this is where robot states, camera frames, and actions are computed; Then in the main_execution function, command is computed and sent using a robot interface The robot_controller thread is started in main_executor.execute function so command can be sent; after starting the thread robot_controller.control_function is called which starts a loop to repeatedly sending computed command