RoboCup Demo

The Booster T1 and K1 Robocup official demo allows the robot to make autonomous decisions to kick the ball and complete the full Robocup match. It includes three programs: vision, brain, and game_controller.

NOTICE The latest RoboCup rules do not allow unicast communication between robots and impose limits on the data size of communication packets during matches. This part of the rules has not yet been addressed in the current open‑source code. If robot communication is required in the competition, the implementation in braincommunication.cpp needs to be modified to comply with the rules.

introduction The Booster RoboCup demo allows the robot to make autonomous decisions to kick the ball and complete the full RoboCup match. It includes three programs: vision, brain, and gamecontroller.

- vision - The vision recognition program, based on Yolo-v8, detects objects such as robots, soccer balls, and the field, and calculates their positions in the robot's coordinate system using geometric relationships. - brain - The decision-making program reads visual data and GameController game control data, integrates all available information, makes judgments, and controls the robot to perform corresponding actions, completing the match process. - gamecontroller - Reads the game control data packets broadcast by the referee machine on the local area network, converts them into ROS2 topic messages, and makes them available for the brain to use.

Install extra dependency bash # Install ros-humble-backward-ros sudo apt-get install ros-humble-backward-ros

bash # Install onnxruntime (If you want to run without cuda)

./thirdpartyaarch64/installonnxruntime.sh # if aarch64

RUN ON Booster K1 or T1 This demo is designed for running on Booster K1, if you'd like to run this demo on Booster T1, few configs need to be updated. ### src/brain/config/config.yaml or src/brain/config/configlocal.yaml 1. set robot.robotheight to 1.12 2. set robot.odomfactor to 1.2 3. set RLVisionKick.enableAutoVisualKick to false 4. set vision.imagecamerainfotopic, vision.depthimagetopic, vision.depthcamerainfotopic to corresponding image topic ### src/vision/config/vision.yaml or src/vision/config/visionlocal.yaml 1. set camera.colortopic, camera.depthtopic, camera.intrintopic to corresponding ones 2. set detectionmodel.modelpath to ./src/vision/model/best122310.3.engine 3. set segmentationmodel.modelpath to ./src/vision/model/bestsegorin10.3.engine

About config: RLVisionKick.enableAutoVisualKick yaml RLVisionKick enableAutoVisualKick: true // This feature is only supported on the Booster K1, requiring firmware version 1.5.2 or higher. VisionKick is only support on K1 and requiring firmware version 1.5.2 or higher. See Firmware 1.5.2 Install Reference for install instruction.

Please make sure you have installed latest Booster Robotics SDK.

Note This repo support jetpack 6.2. Adapted to the default TRT model in src/vision/config/vision.yaml.