Target Tracking via LiDAR-RADAR Sensor Fusion for Autonomous Racing

TL;DR

This work tackles robust target tracking for high-speed, multi-vehicle autonomous racing using a latency-aware EKF-based multi-target tracking framework that fuses LiDAR and RADAR data. It introduces a reduced-state Constant Velocity Turn Rate model (CVTR) with a yaw-rate exogenous input derived from track curvature and a dual-buffer Out-Of-Sequence Measurement (OOSM) reprocessing scheme to cope with sensor latency, all implemented on a PoliMOVE racecar. A key contribution is the explicit integration of RADAR range-rate $R_r$ into the EKF measurement function, along with latency-aware online processing and ablation studies on real LVMS data and simulated Monza data, demonstrating improved tracking accuracy. The approach enables reliable autonomous overtakes up to $275$ km/h, underscoring its practical impact for high-speed racing and offering a basis for extending to cameras and V2V communications, as well as advanced tracking models.

Abstract

High Speed multi-vehicle Autonomous Racing will increase the safety and performance of road-going Autonomous Vehicles. Precise vehicle detection and dynamics estimation from a moving platform is a key requirement for planning and executing complex autonomous overtaking maneuvers. To address this requirement, we have developed a Latency-Aware EKF-based Multi Target Tracking algorithm fusing LiDAR and RADAR measurements. The algorithm explots the different sensor characteristics by explicitly integrating the Range Rate in the EKF Measurement Function, as well as a-priori knowledge of the racetrack during state prediction. It can handle Out-Of-Sequence Measurements via Reprocessing using a double State and Measurement Buffer, ensuring sensor delay compensation with no information loss. This algorithm has been implemented on Team PoliMOVE's autonomous racecar, and was proved experimentally by completing a number of fully autonomous overtaking maneuvers at speeds up to 275 km/h.

Figures (9)

• Figure 1: Team PoliMOVE's Dallara AV-21 (yellow) during an autonomous overtaking maneuver at the Las Vegas Motor Speedway, January 2023. Credits: Indy Autonomous Challenge
• Figure 2: The proposed algorithm architecture with its main components
• Figure 3: Scheme of the OOSM handling via Reprocessing: while in normal operation (left) the measurements are processed in the order of arrival, when an OOSM occurs (right) the state of the MTT is reset to the last valid measurement before the OOSM timestamp and then the measurements are processed in order.
• Figure 4: Computation of the Yaw Rate Exogenous Input using the vehicle estimated velocity $v_k$ and the track curvature $\rho_{closest}$
• Figure 5: Representation of the components of the RADAR Doppler measurement function: the range-rate is the projection of the relative velocity $v_{rel}$ over the Line-of-Sight angle $\alpha$.