Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing State Estimator for Autonomous Racing : Leveraging Multi-modal System and Managing Computing Resources

Daegyu Lee, Hyunwoo Nam, Chanhoe Ryu, Sungwon Nah, Seongwoo Moon, D. Hyunchul Shim

TL;DR

This work addresses robust state estimation for high-speed autonomous racing by integrating GPS/INS and LiDAR within a Bayesian fusion framework, complemented by a unified, georeferenced LiDAR-based state estimator accelerated with CUDA. It introduces a resilient wall-following navigation that maintains safe operation during localization outages and a unified frame map assembled via pose-graph optimization with loop closures and ICP alignment. The approach is validated through extensive real-world races (IAC and HAC) and high-speed simulations, showing improved localization resilience, reduced CPU bottlenecks, and safer operation under GPS degradation or erroneous path guidance. Overall, the methodology advances real-time, multi-sensor fusion for GPS-denied and resource-constrained autonomous racing, with clear practical implications for latency-sensitive, safety-critical navigation on high-speed tracks.

Abstract

This paper introduces an approach that enhances the state estimator for high-speed autonomous race cars, addressing challenges from unreliable measurements, localization failures, and computing resource management. The proposed robust localization system utilizes a Bayesian-based probabilistic approach to evaluate multimodal measurements, ensuring the use of credible data for accurate and reliable localization, even in harsh racing conditions. To tackle potential localization failures, we present a resilient navigation system which enables the race car to continue track-following by leveraging direct perception information in planning and execution, ensuring continuous performance despite localization disruptions. In addition, efficient computing is critical to avoid overload and system failure. Hence, we optimize computing resources using an efficient LiDAR-based state estimation method. Leveraging CUDA programming and GPU acceleration, we perform nearest points search and covariance computation efficiently, overcoming CPU bottlenecks. Simulation and real-world tests validate the system's performance and resilience. The proposed approach successfully recovers from failures, effectively preventing accidents and ensuring safety of the car.

Enhancing State Estimator for Autonomous Racing : Leveraging Multi-modal System and Managing Computing Resources

TL;DR

This work addresses robust state estimation for high-speed autonomous racing by integrating GPS/INS and LiDAR within a Bayesian fusion framework, complemented by a unified, georeferenced LiDAR-based state estimator accelerated with CUDA. It introduces a resilient wall-following navigation that maintains safe operation during localization outages and a unified frame map assembled via pose-graph optimization with loop closures and ICP alignment. The approach is validated through extensive real-world races (IAC and HAC) and high-speed simulations, showing improved localization resilience, reduced CPU bottlenecks, and safer operation under GPS degradation or erroneous path guidance. Overall, the methodology advances real-time, multi-sensor fusion for GPS-denied and resource-constrained autonomous racing, with clear practical implications for latency-sensitive, safety-critical navigation on high-speed tracks.

Abstract

This paper introduces an approach that enhances the state estimator for high-speed autonomous race cars, addressing challenges from unreliable measurements, localization failures, and computing resource management. The proposed robust localization system utilizes a Bayesian-based probabilistic approach to evaluate multimodal measurements, ensuring the use of credible data for accurate and reliable localization, even in harsh racing conditions. To tackle potential localization failures, we present a resilient navigation system which enables the race car to continue track-following by leveraging direct perception information in planning and execution, ensuring continuous performance despite localization disruptions. In addition, efficient computing is critical to avoid overload and system failure. Hence, we optimize computing resources using an efficient LiDAR-based state estimation method. Leveraging CUDA programming and GPU acceleration, we perform nearest points search and covariance computation efficiently, overcoming CPU bottlenecks. Simulation and real-world tests validate the system's performance and resilience. The proposed approach successfully recovers from failures, effectively preventing accidents and ensuring safety of the car.
Paper Structure (35 sections, 30 equations, 14 figures, 6 tables)

This paper contains 35 sections, 30 equations, 14 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Localization based on multi-sensor fusion
  4. Perception based navigation system
  5. Autonomous racing
  6. Methods
  7. Multimodal measurement fusion Kalman filter
  8. Vehicle state estimation formulation
  9. Problem definition
  10. Bayesian-based multimodal measurement update
  11. Unified frame map generation
  12. Efficient LiDAR-based state estimator
  13. Efficient registration method
  14. LiDAR-inertial-vehicle model(LIV) odometry for prediction
  15. Scan match for correction
  16. ...and 20 more sections

Figures (14)

  • Figure 1: (a)Dallara AV-21 of Team KAIST at the Las Vegas Motor Speedway (LVMS) (b)IONIQ 5 of Team KAIST for HAC 2023.
  • Figure 2: Proposed state estimator for high-speed race car
  • Figure 3: Diagram of Kalman filter for multimodal measurement and degradation identification
  • Figure 4: The corresponding pose graph is visualized, showcasing the relationship between the LiDAR point cloud and GPS/INS frames of National Automobile Racetrack of Monza, Italy.
  • Figure 5: Comparison of Covariance Estimation Methods: CPU-Based vs. GPU-Based Approaches
  • ...and 9 more figures