Fine-Tuning Hybrid Physics-Informed Neural Networks for Vehicle Dynamics Model Estimation

TL;DR

This work addresses the challenge of accurate, data-efficient vehicle dynamics modeling for high-speed autonomous racing by introducing Fine-Tuning Hybrid Dynamics (FTHD), a hybrid physics-informed neural network that fine-tunes a pre-trained Deep Dynamics Model (DDM) using a small labeled dataset and a complementary unsupervised physics-driven loss. A second contribution, EKF-FTHD, embeds an Extended Kalman Filter within the framework to denoise real-world sensor data while preserving the underlying physical dynamics, thereby improving parameter estimation under noise. The method combines two losses (Loss_1 and Loss_2) with weights $w_1$ and $w_2$ to enforce both vehicle-specific dynamics and general physical consistency, and uses a physics guard layer to keep Pacejka coefficients within physically plausible bounds. Empirical results on BayesRace-scale simulations and Indy Autonomous Challenge data show that FTHD outperforms the Deep Pacejka Model (DPM) and the original DDM, especially with limited data, and that EKF-FTHD enhances robustness by producing cleaner inputs for subsequent modeling. This combination advances real-time, high-fidelity vehicle dynamics estimation under noisy conditions, with implications for safer and more reliable high-speed autonomous racing control.

Abstract

Accurate dynamic modeling is critical for autonomous racing vehicles, especially during high-speed and agile maneuvers where precise motion prediction is essential for safety. Traditional parameter estimation methods face limitations such as reliance on initial guesses, labor-intensive fitting procedures, and complex testing setups. On the other hand, purely data-driven machine learning methods struggle to capture inherent physical constraints and typically require large datasets for optimal performance. To address these challenges, this paper introduces the Fine-Tuning Hybrid Dynamics (FTHD) method, which integrates supervised and unsupervised Physics-Informed Neural Networks (PINNs), combining physics-based modeling with data-driven techniques. FTHD fine-tunes a pre-trained Deep Dynamics Model (DDM) using a smaller training dataset, delivering superior performance compared to state-of-the-art methods such as the Deep Pacejka Model (DPM) and outperforming the original DDM. Furthermore, an Extended Kalman Filter (EKF) is embedded within FTHD (EKF-FTHD) to effectively manage noisy real-world data, ensuring accurate denoising while preserving the vehicle's essential physical characteristics. The proposed FTHD framework is validated through scaled simulations using the BayesRace Physics-based Simulator and full-scale real-world experiments from the Indy Autonomous Challenge. Results demonstrate that the hybrid approach significantly improves parameter estimation accuracy, even with reduced data, and outperforms existing models. EKF-FTHD enhances robustness by denoising real-world data while maintaining physical insights, representing a notable advancement in vehicle dynamics modeling for high-speed autonomous racing.

Figures (12)

• Figure 1: A schematic of the STM configuration.
• Figure 2: The structure of the proposed FTHD PINN model involves freezing layers of the pre-trained DDM and using a hybrid unsupervised loss for fine-tuning.
• Figure 3: The architecture of the proposed EKF-FTHD PINN model for data processing. This model incorporates an embedded EKF with dynamically adjustable covariance matrices to filter out noise from sensor measurements.
• Figure 4: Illustration of the process for updating the preset coefficient ranges using the EKF-FTHD method.
• Figure 5: The racetrack utilized for data collection.