Autonomous racing is increasingly becoming a proving ground for autonomous
vehicle technology at the limits of its current capabilities. The most
prominent examples include the F1Tenth racing series, Formula Student
Driverless (FSD), Roborace, and the Indy Autonomous Challenge (IAC). Especially
necessary, in high speed autonomous racing, is the knowledge of accurate
racecar vehicle dynamics. The choice of the vehicle dynamics model has to be
made by balancing the increasing computational demands in contrast to improved
accuracy of more complex models. Recent studies have explored learning-based
methods, such as Gaussian Process (GP) regression for approximating the vehicle
dynamics model. However, these efforts focus on higher level constructs such as
motion planning, or predictive control and lack both in realism and rigor of
the GP modeling process, which is often over-simplified. This paper presents
the most detailed analysis of the applicability of GP models for approximating
vehicle dynamics for autonomous racing. In particular we construct dynamic, and
extended kinematic models for the popular F1TENTH racing platform. We
investigate the effect of kernel choices, sample sizes, racetrack layout,
racing lines, and velocity profiles on the efficacy and generalizability of the
learned dynamics. We conduct 400+ simulations on real F1 track layouts to
provide comprehensive recommendations to the research community for training
accurate GP regression for single-track vehicle dynamics of a racecar.

本研究针对 F1TENTH 竞赛平台，构建了动态和扩展运动模型，并进行了 400 多项实验以研究核选择、样本量、赛道布局、比赛路线和速度曲线对学习到的 GP 回归模型的准确性和泛化能力的影响。