EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Hybrid Physics–Data Digital Twin With Robust MPC for Real-Time Fault Compensation in Autonomous Vehicles

Romdhane Nasri, Majdi Mansouri, Jannet Jamii, Jawher Ghommam, Hazem Nounou, Zouhaier Affi

IEEE Open Journal of Vehicular Technology · 2026 · 7 · 3072-3087

DOI: 10.1109/ojvt.2026.3743008 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of Vehicular Technology, 2026. Verfasst von Romdhane Nasri und 5 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Qatar National Research Fund, Qatar National Library.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Autonomous vehicles operating near handling limits are vulnerable to sensor and actuator faults. A single biased measurement or a degraded steering actuator can turn a small tracking error into a loss of control. Existing fault-tolerant Model Predictive Control (MPC) methods usually assume perfect fault classification, treat one fault type at a time, or rely on offline digital twins that cannot adapt during a fault episode. This paper proposes a hybrid physics–data digital twin coupled with a robust MPC formulation. The data-driven part is an online Kernel Principal Component Analysis (KPCA) model that updates every 20 control cycles from a rolling window of recent state–input pairs. Fault detection and classification act on the raw measurements, upstream of any state reconstruction, through four mechanism-matched signature tests: a high-pass energy ratio of the speed signal (sensor noise), a latched step in the measured speed (sensor drift), a steering-feedback residual between the applied and commanded angle (actuator fault), and a sustained lateral-error offset (sensor bias). All thresholds are adapted from a 200-step fault-free baseline. Because the drift and actuator tests identify the fault magnitude at detection, the corresponding corrections are applied before the MPC rather than learned afterwards. The classifier separates the four fault classes at 97.2% accuracy (bias 93.2%, drift 100%, actuator 99.2%, noise 94.6%) with a mean cost of 0.7 ms per cycle. A confidence-weighted constraint-tightening scheme maps the real-time twin uncertainty into the size of the MPC tube. The framework is evaluated in simulation on a 1:10 scale autonomous racing platform (mass 1.98 kg, wheelbase 0.25 m, top speed 5 m/s) that follows the Verschueren racing circuit for 15.8 s at $T_{s} = 20$ ms under a fixed sequential fault schedule in which the bias and noise faults overlap for 56 cycles. Relative to uncompensated faulty operation the framework cuts lateral root-mean-square error (RMSE) from 0.164 m to 0.031 m (81%) and velocity RMSE from 0.157 m/s to 0.060 m/s (62%), bringing both to within 3% of the fault-free run; a per-window decomposition shows that the residual difference is confined to the 56 noisy speed samples, while the drift window is tracked at the fault-free level because the identified offset is removed before the speed loop sees it. Mean computation time stays below 10 ms per 20 ms control cycle. The paper also discusses what is needed to migrate the framework from the desktop-CPU simulation platform to a production automotive electronic control unit.

Förderung

  • Qatar National Research Fund
  • Qatar National Library

Zitieren

APA

Nasri, R., Mansouri, M., Jamii, J., Ghommam, J., Nounou, H., & Affi, Z. (2026). A Hybrid Physics–Data Digital Twin With Robust MPC for Real-Time Fault Compensation in Autonomous Vehicles. IEEE Open Journal of Vehicular Technology, 7, 3072-3087. https://doi.org/10.1109/ojvt.2026.3743008

Vancouver

Nasri R, Mansouri M, Jamii J, Ghommam J, Nounou H, Affi Z. A Hybrid Physics–Data Digital Twin With Robust MPC for Real-Time Fault Compensation in Autonomous Vehicles. IEEE Open Journal of Vehicular Technology. 2026;7:3072-3087. doi:10.1109/ojvt.2026.3743008

BibTeX
@article{Nasri2026,
  author = {Nasri, Romdhane and Mansouri, Majdi and Jamii, Jannet and Ghommam, Jawher and Nounou, Hazem and Affi, Zouhaier},
  title = {A Hybrid Physics–Data Digital Twin With Robust MPC for Real-Time Fault Compensation in Autonomous Vehicles},
  journal = {IEEE Open Journal of Vehicular Technology},
  year = {2026},
  volume = {7},
  pages = {3072--3087},
  doi = {10.1109/ojvt.2026.3743008}
}
Kostenlos starten →