A Frequency-Aware Reinforcement Learning Framework for Suspension Control with Integrated Temporal Features and Expert Guidance Tongji University

Format:

Book

Conference/Event

Author/Creator:

zhu, Zhehui, author.

Zhang, Lijun, author.

Meng, Dejian, author.

Hu, Xingyu, author.

Conference Name:

WCX SAE World Congress Experience (2026-04-14 : Detroit, Michigan, United States)

Language:

English

Subjects (All):

Active suspension systems.

Passive suspension systems.

Machine learning.

Vehicle ride.

Automated vehicles.

Comfort.

Local Subjects:

Active suspension systems.

Passive suspension systems.

Machine learning.

Vehicle ride.

Automated vehicles.

Comfort.

Physical Description:

1 online resource

Place of Publication:

Warrendale, PA SAE International 2026

Summary:

Active suspension systems play a crucial role in improving vehicle ride comfort and handling stability. However, most existing studies focus on the low-frequency range below 20 Hz, leaving the suppression of high-frequency vibrations within 50500 Hz largely unexplored, even though these vibrations strongly affect in-cabin noise and ride quality. To address this gap, this study introduces a quarter-car suspension model incorporating both bushing dynamics and a rigid-ring tire within a reinforcement learning (RL) framework. A major challenge for RL-based suspension control is its degradation in high-frequency performance. To overcome this issue, we design an innovative training framework that integrates multiple synergistic strategies. First, frequency-domain rewards are incorporated as auxiliary signals to explicitly guide policy optimization in the high-frequency band. Second, long short-term memory (LSTM) networks are embedded in both the Actor and Critic to capture the sequential dependencies of time-domain suspension signals, thereby enhancing temporal feature extraction. Finally, model predictive control (MPC) expert knowledge is injected through reward shaping, which accelerates convergence and stabilizes the learned policy. This combination allows the proposed controller to effectively exploit both data-driven learning and model-based insights for full-band suspension optimization. Simulation results show that the method achieves a 29.67% reduction in body acceleration RMS in the 020 Hz range compared with a passive suspension, and further achieves a 62.65% reduction in the 50500 Hz range relative to a baseline RL controller. By explicitly targeting vibration responses in the in-cabin acoustic control band (20500 Hz), this study establishes a foundation for integrated suspension-acoustic optimization, offering new insights into ride comfort and NVH enhancement in intelligent vehicles

Notes:

Vendor supplied data

Access Restriction:

Restricted for use by site license