EUVIMEDEUVIMED entdecken
Crossref

Hierarchical constrained reinforcement learning for coordinated vehicle-body vibration control with active suspension and active stabilizer bars

Kang Huang, Yingjie Zha, Fang Li, Hao Sun

Journal of Vibration and Control · 22.07.2026

DOI: 10.1177/10775463261471081 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Vibration and Control, online erschienen am 22.07.2026. Verfasst von Kang Huang und 3 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 37 Literaturangaben. Gefördert von: Fundamental Research Funds for the Central Universities, Anqing–Hefei University of Technology Industrial Innovation Guidance Fund.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents a constrained hierarchical reinforcement learning strategy for coordinated vehicle-body vibration control using active suspension and active stabilizer bars. Four fixed PID/skyhook modules remain as interpretable low-level controllers, while an upper-layer Augmented Lagrangian PPO coordinator adjusts bounded blending weights online. The policy is trained in CarSim–Simulink on a variable-speed Fishhook maneuver and tested on unseen sinusoidal path tracking and straight-line cruising. Compared with passive and PSO-tuned fixed-weight baselines, the proposed controller improves the stability–comfort trade-off while respecting actuator-capacity, suspension-travel, and tire-load constraints. The hierarchy also supports practical fallback to fixed calibrated weights because the learned policy supervises rather than replaces the low-level modules. In Fishhook, RMS roll angle and roll rate are reduced by 75.3 and 87.4% versus passive, and by 10.5 and 16.3% versus PSO. Relative to unconstrained PPO, AL-PPO also reduces total energy, ASB peak power, and total peak power by 2.2, 29.5, and 1.8%, respectively, while eliminating physical ASB moment violations. Actuator-ablation tests confirm the complementary roles of AS and ASB, and reduced-authority tests show that the method does not rely on frequent use of the 1000 N⋅m physical ASB limit.

Förderung

  • Fundamental Research Funds for the Central Universities
  • Anqing–Hefei University of Technology Industrial Innovation Guidance Fund

Zitieren

APA

Huang, K., Zha, Y., Li, F., & Sun, H. (2026). Hierarchical constrained reinforcement learning for coordinated vehicle-body vibration control with active suspension and active stabilizer bars. Journal of Vibration and Control. https://doi.org/10.1177/10775463261471081

Vancouver

Huang K, Zha Y, Li F, Sun H. Hierarchical constrained reinforcement learning for coordinated vehicle-body vibration control with active suspension and active stabilizer bars. Journal of Vibration and Control. 2026. doi:10.1177/10775463261471081

BibTeX
@article{Huang2026,
  author = {Huang, Kang and Zha, Yingjie and Li, Fang and Sun, Hao},
  title = {Hierarchical constrained reinforcement learning for coordinated vehicle-body vibration control with active suspension and active stabilizer bars},
  journal = {Journal of Vibration and Control},
  year = {2026},
  doi = {10.1177/10775463261471081}
}
Kostenlos starten →