[Paper Review] Autonomous Driving System Design for Formula Student Driverless Racecar

Hanqing Tian, Jun Ni+1 more|arXiv (Cornell University)|Sep 19, 2018

Robotics and Sensor-Based Localization4 references4 citations

TL;DR

This paper presents an integrated autonomous driving system for a Formula Student driverless racecar, combining LIDAR and vision for traffic cone detection, fusing GPS-INS and LIDAR odometry for high-rate localization, and building a real-time track map with cone positions and colors. The system achieved stable path tracking on a closed-loop track, demonstrating robust performance in the 2017 Formula Student Autonomous Competition.

ABSTRACT

This paper summarizes the work of building the autonomous system including detection system and path tracking controller for a formula student autonomous racecar. A LIDAR-vision cooperating method of detecting traffic cone which is used as track mark is proposed. Detection algorithm of the racecar also implements a precise and high rate localization method which combines the GPS-INS data and LIDAR odometry. Besides, a track map including the location and color information of the cones is built simultaneously. Finally, the system and vehicle performance on a closed loop track is tested. This paper also briefly introduces the Formula Student Autonomous Competition (FSAC) in 2017.

Motivation & Objective

• To design a reliable autonomous driving system for a Formula Student driverless racecar competing in the 2017 Formula Student Autonomous Competition (FSAC).
• To enable accurate detection of traffic cones—used as track markers—using a LIDAR-vision cooperative detection method.
• To achieve high-rate, precise localization by fusing GPS-INS and LIDAR odometry data.
• To construct a dynamic track map that includes cone locations and color information for path planning.
• To validate the system's performance through real-world testing on a closed-loop track.

Proposed method

• A LIDAR-vision cooperating detection method was developed to identify traffic cones with high accuracy and update rate.
• GPS-INS and LIDAR odometry data were fused using a complementary filtering approach to improve localization precision and update frequency.
• A real-time track map was generated by processing detected cone positions and assigning color information from vision data.
• A path tracking controller was implemented to guide the vehicle along the predefined track using the map and localization data.
• The system was integrated into a racecar platform and tested on a closed-loop track to evaluate performance.
• The design was validated during the 2017 Formula Student Autonomous Competition, demonstrating system stability and reliability.

Experimental results

Research questions

• RQ1 How can LIDAR and vision data be effectively fused to detect traffic cones in real time for autonomous racing?
• RQ2 What is the optimal method for combining GPS-INS and LIDAR odometry to achieve high-rate, accurate localization?
• RQ3 Can a dynamic track map with both spatial and color information of cones be reliably constructed during autonomous operation?
• RQ4 How well does the integrated system perform in maintaining path tracking on a closed-loop track under real competition conditions?
• RQ5 What are the key system-level challenges in deploying an autonomous racecar in a Formula Student competition setting?

Key findings

• The LIDAR-vision fusion method enabled reliable and high-update-rate detection of traffic cones, essential for real-time navigation.
• The combined GPS-INS and LIDAR odometry solution achieved sub-decimeter localization accuracy with high update rates, critical for dynamic control.
• A real-time track map with cone positions and colors was successfully built and used for path planning during operation.
• The autonomous system demonstrated stable path tracking performance on a closed-loop track, confirming the system's robustness.
• The vehicle completed the competition track successfully, validating the overall system design and integration in a real-world autonomous racing environment.
• The system's performance highlights the effectiveness of sensor fusion and real-time mapping in low-latency autonomous driving applications.