LiDAR-Based Vehicle Detection and Tracking for Autonomous Racing

TL;DR

The paper addresses the need for fast, robust LiDAR-based perception to enable autonomous overtaking in racing. It proposes an online pipeline that combines Range Image ground removal, fast segmentation, multi-sensor cluster merging, a novel 2D pose estimator built from a fusion of trajectory-based and variance-based headings, and a variable-step EKF-based Multi-Target Tracking with GNN data association and M/N track management. Key contributions include a parallelizable segmentation approach capable of processing data from three LiDARs at high speeds, an effective 2D pose estimation method for irregular vehicle shapes, and a tracking system that delivers sub-sensor-latency performance (average ~26 ms) and reliable velocity and heading estimates up to tens of meters ahead, validated on the 2023 IAC CES dataset with overtakes at speeds above 250 km/h. The work demonstrates practical viability for fully autonomous overtaking in high-speed racing and informs design principles for perception pipelines under stringent latency and reliability constraints.

Abstract

Autonomous racing provides a controlled environment for testing the software and hardware of autonomous vehicles operating at their performance limits. Competitive interactions between multiple autonomous racecars however introduce challenging and potentially dangerous scenarios. Accurate and consistent vehicle detection and tracking is crucial for overtaking maneuvers, and low-latency sensor processing is essential to respond quickly to hazardous situations. This paper presents the LiDAR-based perception algorithms deployed on Team PoliMOVE's autonomous racecar, which won multiple competitions in the Indy Autonomous Challenge series. Our Vehicle Detection and Tracking pipeline is composed of a novel fast Point Cloud Segmentation technique and a specific Vehicle Pose Estimation methodology, together with a variable-step Multi-Target Tracking algorithm. Experimental results demonstrate the algorithm's performance, robustness, computational efficiency, and suitability for autonomous racing applications, enabling fully autonomous overtaking maneuvers at velocities exceeding 275 km/h.

Figures (12)

• Figure 1: Team PoliMOVE's Dallara AV21 "MinerVa" defending from an autonomous overtaking maneuver initiated by TUM Autonomous Motorsport during the final race of the Indy Autonomous Challenge event on January 7, 2023, at the Las Vegas Motor Speedway Credits: Indy Autonomous Challenge.
• Figure 2: Representation of the main research problems in Vehicle Detection and Tracking, the most common solutions in literature and their relationship. In bold, the research problems, in italic the approaches used in this work. Neighboring boxes should be considered as alternative methodologies.
• Figure 3: Block scheme of the LiDAR-Based Tracking algorithm
• Figure 4: The five steps of the Point Cloud segmentation algorithm and the resulting segmented Point Cloud. Images are scaled vertically for easier visualization.
• Figure 5: Input and Segmented Point Clouds from the three sensors capturing an opponent performing an overtaking maneuver at 250 km/h. The EGO vehicle is represented as a semi-transparent 3D graphical model. Grid size: 1m.