Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Spatially-Aware Adaptive Trajectory Optimization with Controller-Guided Feedback for Autonomous Racing

Alexander Wachter, Alexander Willert, Marc-Philip Ecker, Christian Hartl-Nesic

TL;DR

This paper addresses the minimum-lap-time problem in autonomous racing by exploiting the repetitive nature of circuits to iteratively refine trajectories. It introduces a closed-loop raceline optimization framework that couples a NURBS-based trajectory representation with CMA-ES global optimization and MPC tracking feedback, guided by a Kalman-inspired adaptive spatial constraint map and blame-region analysis. Key contributions include execution-feedback-driven trajectory shaping, spatial constraint learning without explicit sensing, and a closed-loop refinement architecture with an open-source implementation, achieving up to 17.38% simulation improvement and 7.60% real-world improvement across varying friction. The approach demonstrates robust performance under spatially varying track characteristics and demonstrates practical impact for autonomous racing in real-world conditions.

Abstract

We present a closed-loop framework for autonomous raceline optimization that combines NURBS-based trajectory representation, CMA-ES global trajectory optimization, and controller-guided spatial feedback. Instead of treating tracking errors as transient disturbances, our method exploits them as informative signals of local track characteristics via a Kalman-inspired spatial update. This enables the construction of an adaptive, acceleration-based constraint map that iteratively refines trajectories toward near-optimal performance under spatially varying track and vehicle behavior. In simulation, our approach achieves a 17.38% lap time reduction compared to a controller parametrized with maximum static acceleration. On real hardware, tested with different tire compounds ranging from high to low friction, we obtain a 7.60% lap time improvement without explicitly parametrizing friction. This demonstrates robustness to changing grip conditions in real-world scenarios.

Spatially-Aware Adaptive Trajectory Optimization with Controller-Guided Feedback for Autonomous Racing

TL;DR

This paper addresses the minimum-lap-time problem in autonomous racing by exploiting the repetitive nature of circuits to iteratively refine trajectories. It introduces a closed-loop raceline optimization framework that couples a NURBS-based trajectory representation with CMA-ES global optimization and MPC tracking feedback, guided by a Kalman-inspired adaptive spatial constraint map and blame-region analysis. Key contributions include execution-feedback-driven trajectory shaping, spatial constraint learning without explicit sensing, and a closed-loop refinement architecture with an open-source implementation, achieving up to 17.38% simulation improvement and 7.60% real-world improvement across varying friction. The approach demonstrates robust performance under spatially varying track characteristics and demonstrates practical impact for autonomous racing in real-world conditions.

Abstract

We present a closed-loop framework for autonomous raceline optimization that combines NURBS-based trajectory representation, CMA-ES global trajectory optimization, and controller-guided spatial feedback. Instead of treating tracking errors as transient disturbances, our method exploits them as informative signals of local track characteristics via a Kalman-inspired spatial update. This enables the construction of an adaptive, acceleration-based constraint map that iteratively refines trajectories toward near-optimal performance under spatially varying track and vehicle behavior. In simulation, our approach achieves a 17.38% lap time reduction compared to a controller parametrized with maximum static acceleration. On real hardware, tested with different tire compounds ranging from high to low friction, we obtain a 7.60% lap time improvement without explicitly parametrizing friction. This demonstrates robustness to changing grip conditions in real-world scenarios.
Paper Structure (20 sections, 17 equations, 8 figures, 1 table)

This paper contains 20 sections, 17 equations, 8 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Motion Planning and Trajectory Optimization
  4. Spatially-Aware Dynamics and Adaptive Methods
  5. Learning-Based Approaches and Optimization Methods
  6. Method
  7. NURBS-based Trajectory Representation
  8. Time-Optimal Parameterization with Spatial Adaptation
  9. Adaptive Constraint Map Framework
  10. Blame Region Computation Through Derivative Analysis
  11. Error Feedback and Constraint Learning
  12. Model Predictive Control Implementation
  13. Trajectory Design with Closure Constraints
  14. Global Trajectory Optimization
  15. Experimental Setup and Results
  16. ...and 5 more sections

Figures (8)

  • Figure 1: Closed-loop raceline optimization framework using NURBS trajectories, CMA-ES optimization, and MPC tracking feedback.
  • Figure 2: Constraint map visualization. Blue: reduced acceleration regions; Red: increased acceleration regions.
  • Figure 3: Track segmentation into acceleration (red) and deceleration (blue) zones for blame region identification. Since the trajectory does not contain any neutral phases, these are not represented in the figure.
  • Figure 4: F1Tenth experimental platform with Asus NUC and Hokuyo LiDAR.
  • Figure 5: Optimizer convergence on F1Aut track. Red line: Error feedback activation. Convergence is achieved within 8 laps, with 500 iterations/lap.
  • ...and 3 more figures