National Robotics Competition: RoboMaster University Series

What is robomaster?

RoboMaster is an internationally recognized robotics competition, organized by DJI Innovative Technologies. The competition requires teams to design, build, and program a series of robots to perform a variety of tasks and compete against each other in a specific venue. I was involved in the development of the navigation system as a member of the algorithm team.

Challenges

The main technical challenges we faced in the competition include:

1. Robot mechanical and electrical system design
2. real-time sensing and localization
3. path planning under undulating environment
4. whole body control for switch system
5. visual recognition and tracking
6. Robust strategies for adversarial environments

My contribution

As Head of Navigation Team for Season 23 and Head of Fully Automated Robotics Team for Season 24, I:

• Designed a navigation system for fully automated robots
• Developed a SLAM system based on LiDAR and IMU to realize autonomous localization and global positioning with centimeter-level accuracy
• Designed an optimal path planning algorithm to improve robot maneuverability in complex sites.
• Designed an optimal path planning algorithm to improve the maneuverability of the robot in complex sites.
• Implemented communication between robots
• Optimized visual recognition algorithms for omnidirectional sensing of the robot's environment

System Structure

Our system structure is based on ROS2 and navigation2

• Auto Aim Module:

The computer vision system that detects armor plates on enemy robots Uses camera inputs to identify and track targets using YOLO-V7 Includes armor detection, number classification, and 3D position estimation using EKF Supports multiple camera configurations for omnidirectional perception

• Navigation System

Localization: Uses LiDAR and IMU for robot self-positioning. Utilized SLAM algorithms like LiDAR-IMU fusion for EKF state estimation and Iterative Closest Point for point cloud registration Path Planning: Generates optimal paths for robot pose using A* Motion Planning: Generate optimal motion for robot using nonlinear model predictive control Obstacle Avoidance: Detects and navigates around obstacles in real-time using voxel mapping

• Decision Making System

Behavior Tree: Implements robot strategy and decision logic using behavior trees

• Hardware Interface

Cameras: Supports industrial cameras including Daheng and Hik Robot cameras for high-rate(120Hz) video inputs Serial Communication: Interfaces with lower-level control systems using Serial port LiDAR Drivers: Processes point cloud data for mapping and navigation

• Robot Description

Provides physical and kinematic models of the robots in URDF Supports various robot types (fully automated robot, mobile manipulator, etc.)

Competition Videos

Please refer to RoboMaster BiliBili Account and follow ARTINX.

Achievements

• 2023 National First Prize
• 2023 National First Prize for Fully Automated Robot
• 2024 National Second Prize