HKUST Robotics Team - The 14th Robot Design Contest 2022

Group 5 Fibotics - ARC (Automatic Racing Car)

Oct - Dec 2022

Technologies: STM32, Image Processing (Sobel), PID Control, RoboMaster M3508 Motors, C620 ESCs, Omniwheel Kinematics, CAN Bus, Embedded C

🎯 Game Task Overview

RDC2022 ARC Track

The Automatic Racing Car (ARC) must autonomously race on a designated track for four laps within 3 minutes. The faster team earns a Replacement Opportunity (a golden piece) for their TR. The ARC must follow the track without exiting; any lap where the robot leaves the track is not counted. The track may contain random obstacles (e.g., pieces dropped by TRs), and the ARC must avoid them.

📄 Download Full Game Rule (PDF) 📄 Game Introduction Video

Overview

The Automatic Racing Car (ARC) was designed to autonomously navigate a track and complete the tic-tac-toe challenge. As part of Team Fibotics (Group 5), I contributed to the software development, focusing on image processing, track sensing algorithms, and low-level motor control.

Fibotics ARC during testing

Image Processing

We used a Sobel filter to detect edges and filter out unwanted data for track sensing. Initially, filtering the entire image was too time-consuming and reduced the frame rate. To improve performance, we modified the algorithm to filter only the pixels needed for navigation, significantly increasing the ARC's responsiveness.

Sobel filter test (only for illustration; full-image filtering was avoided during movement)

Track Sensing & Control

The ARC continuously calculated two critical parameters:

The distance between the track's centre and the left/right borders.
The distance to the edge in front of the robot.

Using these measurements, a PID algorithm adjusted the steering angle to keep the robot centered. The system also detected upcoming turns by analyzing changes in border distances, then decelerated and executed a controlled drift through the turn. Additionally, the ARC counted the number of turns to determine whether it should follow the left or right border.

Track sensing concept: distance to borders and front edge (only for illustration; full-image filtering was avoided during movement)

Video Demonstrations

Demonstration of ARC navigation (Stable version)

Demonstration of ARC navigation (Fast version)

Awards & Recognition

🏆 HKUST Robotics Team - The 14th Robot Design Contest 2022 - 2nd Runner Up
🏅 HKUST Robotics Team - The 14th Robot Design Contest 2022 - Best Engineering Award

Resources

📄 Download Poster (PDF) 📊 Download Final Presentation (PPT) Scoreboard RDC2022 Google Drive Folder