Embedded Files
Anh Tung Ho, Binh Han, WorawateAusa, 박시영, 박해인, 유호진
Email: hoanhtung2000nd@gmail.com
Introduction
An autonomous racing car contest for 7 teams in which each team built up a racing car controlled by Jetson Nano. Navigated the car to finish 5 laps in the shortest amount of time while avoiding random obstacles and following traffic-light signals.
I led a team of 6 people to rank the first place in the contest.
Method
Developed the code for lidar-based mapping and localization with Iterative Closest Point (ICP) algorithm.
Optimized the Stanley lateral motion controller with adaptive looking ahead waypoint according to road curvature for smooth turning and quickly recovery.
Proposed a novel algorithm for local path planning to avoid obstacles, which tripled computer calculation rate compared to Frenet path-planning algorithm, from 10Hz to 30Hz.
Integrated YOLOv4 to detect traffic-light signals with confidence of 99%.
Result
Ranked 1st with the shortest amount of time record.
The racing car travelled robustly regardless of random obstacles.
Page updated
Google Sites
Report abuse