Online Velocity Profile Generation and Tracking for Sampling-Based Local Planning Algorithms in Autonomous Racing Environments

TL;DR

This work tackles the challenge of real-time velocity adaptation for autonomous racing by coupling an online velocity profile generator with a spatial-domain, sampling-based local planner on a three-dimensional track. It introduces an online apex-detection-based velocity optimization using a forward-backward (FW-BW) framework alongside a novel spatial trajectory sampling strategy that preserves track geometry and apex locations. The approach demonstrates improved adaptability to changing grip conditions and demonstrates advantages in single- and multi-vehicle simulations, while identifying limitations related to lateral deviations, transient dynamics, and potential jerkiness in acceleration. The results indicate that online velocity adaptation, combined with spatially aware local planning, can meaningfully enhance performance in high-speed racing scenarios and pave the way for more robust, real-world autonomous racing systems.

Abstract

This work presents an online velocity planner for autonomous racing that adapts to changing dynamic constraints, such as grip variations from tire temperature changes and rubber accumulation. The method combines a forward-backward solver for online velocity optimization with a novel spatial sampling strategy for local trajectory planning, utilizing a three-dimensional track representation. The computed velocity profile serves as a reference for the local planner, ensuring adaptability to environmental and vehicle dynamics. We demonstrate the approach's robust performance and computational efficiency in racing scenarios and discuss its limitations, including sensitivity to deviations from the predefined racing line and high jerk characteristics of the velocity profile.

Figures (7)

• Figure 1: An exemplary speed profile of a race car for turns 6 and 7 at Yas Marina Circuit, Abu Dhabi. Two apexes (red dots) are identified on the depicted section of the track.
• Figure 2: An overview of the proposed online velocity planning framework.
• Figure 3: Exemplary gg-diagrams for $v =$ 40m/s, $\tilde{g} =$ 9.81m/s^2 and $\rho = 1.3$, adjusted by three different scaling factors $\alpha$. Note that the longitudinal acceleration potential $\tilde{a}_{\mathrm{x, eng}}$ is not affected by $\alpha$.
• Figure 4: Comparison of relative longitudinal sampling strategies in the spatial and temporal domain in turns 6 and 7.
• Figure 5: The online velocity profile respecting an acceleration limit scaling of $\alpha=0.7$ and the offline generated profile. The lower plot shows the corresponding accelerations.