Deep Dynamics: Vehicle Dynamics Modeling with a Physics-Constrained Neural Network for Autonomous Racing

TL;DR

This paper presents Deep Dynamics, a physics-constrained neural network that models high-speed vehicle dynamics accurately while ensuring physical plausibility, addressing the limitations of existing physics-based and data-driven models in autonomous racing.

Contribution

It introduces a novel PCNN with a Physics Guard layer that combines physics coefficients and dynamical equations for improved vehicle dynamics modeling at high speeds.

Findings

Accurately predicts vehicle states at speeds over 280 km/h.

Ensures physical plausibility of model predictions through the Physics Guard layer.

Demonstrates superior performance in simulation and real-world data assessments.

Abstract

Autonomous racing is a critical research area for autonomous driving, presenting significant challenges in vehicle dynamics modeling, such as balancing model precision and computational efficiency at high speeds (>280km/h), where minor errors in modeling have severe consequences. Existing physics-based models for vehicle dynamics require elaborate testing setups and tuning, which are hard to implement, time-intensive, and cost-prohibitive. Conversely, purely data-driven approaches do not generalize well and cannot adequately ensure physical constraints on predictions. This paper introduces Deep Dynamics, a physics-constrained neural network (PCNN) for vehicle dynamics modeling of an autonomous racecar. It combines physics coefficient estimation and dynamical equations to accurately predict vehicle states at high speeds and includes a unique Physics Guard layer to ensure internal coefficient estimates remain within their nominal physical ranges. Open-loop and closed-loop performance assessments, using a physics-based simulator and full-scale autonomous Indy racecar data, highlight Deep Dynamics as a promising approach for modeling racecar vehicle dynamics.