One Model to Drift Them All: Physics-Informed Conditional Diffusion Model for Driving at the Limits

Franck Djeumou, Thomas Jonathan Lew, NAN DING, Michael Thompson, Makoto Suminaka, Marcus Greiff, John Subosits
Proceedings of The 8th Conference on Robot Learning, PMLR 270:604-630, 2025.

Abstract

Enabling autonomous vehicles to reliably operate at the limits of handling— where tire forces are saturated — would improve their safety, particularly in scenarios like emergency obstacle avoidance or adverse weather conditions. However, unlocking this capability is challenging due to the task’s dynamic nature and the high sensitivity to uncertain multimodal properties of the road, vehicle, and their dynamic interactions. Motivated by these challenges, we propose a framework to learn a conditional diffusion model for high-performance vehicle control using an unlabelled multimodal trajectory dataset. We design the diffusion model to capture the distribution of parameters of a physics-informed data-driven dynamics model. By conditioning the generation process on online measurements, we integrate the diffusion model into a real-time model predictive control framework for driving at the limits, and show that it can adapt on the fly to a given vehicle and environment. Extensive experiments on a Toyota Supra and a Lexus LC 500 show that a single diffusion model enables reliable autonomous drifting on both vehicles when operating with different tires in varying road conditions. The model matches the performance of task-specific expert models while outperforming them in generalization to unseen conditions, paving the way towards a general, reliable method for autonomous driving at the limits of handling.

Cite this Paper

BibTeX
@InProceedings{pmlr-v270-djeumou25a, title = {One Model to Drift Them All: Physics-Informed Conditional Diffusion Model for Driving at the Limits}, author = {Djeumou, Franck and Lew, Thomas Jonathan and DING, NAN and Thompson, Michael and Suminaka, Makoto and Greiff, Marcus and Subosits, John}, booktitle = {Proceedings of The 8th Conference on Robot Learning}, pages = {604--630}, year = {2025}, editor = {Agrawal, Pulkit and Kroemer, Oliver and Burgard, Wolfram}, volume = {270}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v270/main/assets/djeumou25a/djeumou25a.pdf}, url = {https://proceedings.mlr.press/v270/djeumou25a.html}, abstract = {Enabling autonomous vehicles to reliably operate at the limits of handling— where tire forces are saturated — would improve their safety, particularly in scenarios like emergency obstacle avoidance or adverse weather conditions. However, unlocking this capability is challenging due to the task’s dynamic nature and the high sensitivity to uncertain multimodal properties of the road, vehicle, and their dynamic interactions. Motivated by these challenges, we propose a framework to learn a conditional diffusion model for high-performance vehicle control using an unlabelled multimodal trajectory dataset. We design the diffusion model to capture the distribution of parameters of a physics-informed data-driven dynamics model. By conditioning the generation process on online measurements, we integrate the diffusion model into a real-time model predictive control framework for driving at the limits, and show that it can adapt on the fly to a given vehicle and environment. Extensive experiments on a Toyota Supra and a Lexus LC 500 show that a single diffusion model enables reliable autonomous drifting on both vehicles when operating with different tires in varying road conditions. The model matches the performance of task-specific expert models while outperforming them in generalization to unseen conditions, paving the way towards a general, reliable method for autonomous driving at the limits of handling.} }
Endnote
%0 Conference Paper %T One Model to Drift Them All: Physics-Informed Conditional Diffusion Model for Driving at the Limits %A Franck Djeumou %A Thomas Jonathan Lew %A NAN DING %A Michael Thompson %A Makoto Suminaka %A Marcus Greiff %A John Subosits %B Proceedings of The 8th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Pulkit Agrawal %E Oliver Kroemer %E Wolfram Burgard %F pmlr-v270-djeumou25a %I PMLR %P 604--630 %U https://proceedings.mlr.press/v270/djeumou25a.html %V 270 %X Enabling autonomous vehicles to reliably operate at the limits of handling— where tire forces are saturated — would improve their safety, particularly in scenarios like emergency obstacle avoidance or adverse weather conditions. However, unlocking this capability is challenging due to the task’s dynamic nature and the high sensitivity to uncertain multimodal properties of the road, vehicle, and their dynamic interactions. Motivated by these challenges, we propose a framework to learn a conditional diffusion model for high-performance vehicle control using an unlabelled multimodal trajectory dataset. We design the diffusion model to capture the distribution of parameters of a physics-informed data-driven dynamics model. By conditioning the generation process on online measurements, we integrate the diffusion model into a real-time model predictive control framework for driving at the limits, and show that it can adapt on the fly to a given vehicle and environment. Extensive experiments on a Toyota Supra and a Lexus LC 500 show that a single diffusion model enables reliable autonomous drifting on both vehicles when operating with different tires in varying road conditions. The model matches the performance of task-specific expert models while outperforming them in generalization to unseen conditions, paving the way towards a general, reliable method for autonomous driving at the limits of handling.
APA
Djeumou, F., Lew, T.J., DING, N., Thompson, M., Suminaka, M., Greiff, M. & Subosits, J.. (2025). One Model to Drift Them All: Physics-Informed Conditional Diffusion Model for Driving at the Limits. Proceedings of The 8th Conference on Robot Learning, in Proceedings of Machine Learning Research 270:604-630 Available from https://proceedings.mlr.press/v270/djeumou25a.html.

Related Material