Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neural Network Tire Force Modeling for Automated Drifting

Nicholas Drake Broadbent, Trey Weber, Daiki Mori, J. Christian Gerdes

TL;DR

The study tackles the challenge of accurately predicting front-tire lateral forces at the friction limit during automated drifting, where physics-based tire models struggle due to parameter variability and heating effects. It proposes a small three-layer neural network to predict front-tire lateral force as a drop-in predictor and integrates it into a nonlinear model predictive controller (NMPC), comparing performance against a Fiala brush model on a full-scale drift vehicle (Takumi). Results show that the neural-network-enabled NMPC achieves faster, less-oscillatory responses and lower mean/max lateral errors, especially when front-axle braking is applied, though initiation-region performance is limited by data imbalance and potential observer latency. The work demonstrates the potential of data-driven front-tire force modeling to capture latent drifting dynamics and improve autonomous drifting control, with future directions including data-balanced training, measurement-based labeling, and incorporating additional inputs such as tire temperature.

Abstract

Automated drifting presents a challenge problem for vehicle control, requiring models and control algorithms that can precisely handle nonlinear, coupled tire forces at the friction limits. We present a neural network architecture for predicting front tire lateral force as a drop-in replacement for physics-based approaches. With a full-scale automated vehicle purpose-built for the drifting application, we deploy these models in a nonlinear model predictive controller tuned for tracking a reference drifting trajectory, for direct comparisons of model performance. The neural network tire model exhibits significantly improved path tracking performance over the brush tire model in cases where front-axle braking force is applied, suggesting the neural network's ability to express previously unmodeled, latent dynamics in the drifting condition.

Neural Network Tire Force Modeling for Automated Drifting

TL;DR

The study tackles the challenge of accurately predicting front-tire lateral forces at the friction limit during automated drifting, where physics-based tire models struggle due to parameter variability and heating effects. It proposes a small three-layer neural network to predict front-tire lateral force as a drop-in predictor and integrates it into a nonlinear model predictive controller (NMPC), comparing performance against a Fiala brush model on a full-scale drift vehicle (Takumi). Results show that the neural-network-enabled NMPC achieves faster, less-oscillatory responses and lower mean/max lateral errors, especially when front-axle braking is applied, though initiation-region performance is limited by data imbalance and potential observer latency. The work demonstrates the potential of data-driven front-tire force modeling to capture latent drifting dynamics and improve autonomous drifting control, with future directions including data-balanced training, measurement-based labeling, and incorporating additional inputs such as tire temperature.

Abstract

Automated drifting presents a challenge problem for vehicle control, requiring models and control algorithms that can precisely handle nonlinear, coupled tire forces at the friction limits. We present a neural network architecture for predicting front tire lateral force as a drop-in replacement for physics-based approaches. With a full-scale automated vehicle purpose-built for the drifting application, we deploy these models in a nonlinear model predictive controller tuned for tracking a reference drifting trajectory, for direct comparisons of model performance. The neural network tire model exhibits significantly improved path tracking performance over the brush tire model in cases where front-axle braking force is applied, suggesting the neural network's ability to express previously unmodeled, latent dynamics in the drifting condition.
Paper Structure (7 sections, 2 figures)

This paper contains 7 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Experimental Setup
  3. Neural Network Model Development
  4. Trajectory Generation
  5. Control Architecture
  6. Results and Discussion
  7. Conclusion

Figures (2)

  • Figure 1: Neural network architecture for predicting lateral tire force
  • Figure 2: Tire model comparison of tracking performance under braking