Education, Science, Technology, Innovation and Life
Open Access
Sign In
  1. Home
  2. Journals
  3. Automation and Machine Learning
  4. Vol 6, Issue 2, 2025

Active Suspension Control Based on DQN-LSTM with Integrated Temporal Feature Extraction

Download as PDF

DOI: 10.23977/autml.2025.060210 | Downloads: 2 | Views: 62

Author(s)

Li Chenyang 1, Li Wei 1, Gao Yanfei 1, Zhang Hongjia 1

Affiliation(s)

1 School of Automotive Engineering, Shandong Jiaotong University, Jinan, 250357, China

Corresponding Author

Li Wei

ABSTRACT

To improve vehicle ride comfort and address the temporal dependency inherent in active suspension control, this study proposes a reinforcement learning–based control algorithm that integrates a Deep Q-Network (DQN) with a Long Short-Term Memory (LSTM) network, referred to as DQN-LSTM. A two-degree-of-freedom vertical dynamics model is first established as the interaction environment for the algorithm. A reward function is then designed to minimize the root-mean-square (RMS) value of the vehicle body vertical acceleration, where the DQN is responsible for policy optimization, and an LSTM layer is incorporated to extract temporal features embedded in historical state sequences, thereby enhancing the controller's capability to predict and respond to road excitations. Simulation tests on Class B and Class C random roads are conducted in MATLAB. The results indicate that, compared with the passive suspension, the DQN controller reduces the RMS of the body vertical acceleration by 12.78%, whereas the proposed DQN-LSTM controller further reduces it by 25.11%, yielding a notably smoother system response. These findings demonstrate that the proposed algorithm effectively captures temporal characteristics and exhibits strong adaptability, robustness, and application potential under stochastic road excitations.

KEYWORDS

Active Suspension, Eep Reinforcement Learning, DQN-LSTM, Ride Comfort Optimization, Intelligent Control

CITE THIS PAPER

Li Chenyang, Li Wei, Gao Yanfei, Zhang Hongjia, Active Suspension Control Based on DQN-LSTM with Integrated Temporal Feature Extraction. Automation and Machine Learning (2025) Vol. 6: 74-83. DOI: http://dx.doi.org/10.23977/autml.2025.060210.

REFERENCES

[1] Yu M, Evangelou S A, Dini D. Advances in Active Suspension Systems for Road Vehicles [J]. Engineering, 2023.
[2] Llopis-Albert C, Rubio F, Zeng S. Multiobjective optimization framework for designing a vehicle suspension system. A comparison of optimization algorithms [J]. Advances in Engineering Software, 2023, 176: 103375.
[3] Zeng Q, Zhao J. Event-triggered controller design for active suspension systems: An adaptive backstepping method with error-dependent gain [J]. Control Engineering Practice, 2023, 136: 105547.
[4] Sun W, Li C Y, Wang J N, et al. Study on ride comfort of a hybrid suspension based on multi-condition modes [J]. Automotive Engineering, 2024, 46(11): 2076–2090+2099.
[5] Lin B, Su X. Fault-tolerant controller design for active suspension system with proportional differential sliding mode observer [J]. International Journal of Control, Automation and Systems, 2019, 17: 1751-1761
[6] Sun W, Li C Y, Wang J N, et al. Study on ride comfort of hybrid suspension for off-road vehicles [J]. Automotive Engineering, 2022, 44(01): 105–114+122.
[7] Cao Y L, Zhang Q. Steering control of four-wheel steering vehicles based on adaptive neuro-fuzzy inference [J]. Mechanical Design and Manufacture, 2021(03): 224–228+233.
[8] Xu M, Huang Q S. Research status of intelligent control methods for vehicle semi-active suspension [J]. Machine Tool & Hydraulics, 2021, 49(01): 169–174.
[9] Nazemian H, Masih-Tehrani M. Hybrid Fuzzy-PID control development for a truck air suspension system [J]. SAE International Journal of Commercial Vehicles, 2020, 13(1): 55-70.
[10] Ming L, Yibin L, Xuewen R, et al. Semi-active suspension control based on deep reinforcement learning [J]. IEEE Access, 2020, 8: 9978-9986.
[11] Wang Z H. Research on semi-active suspension strategy based on deep Q-network [D]. Chengdu: Southwest Jiaotong University, 2021.
[12] Deng M, Sun D, Zhan L, et al. Advancing active suspension control with TD3-PSC: integrating physical safety constraints into deep reinforcement learning [J]. IEEE Access, 2024.
[13] Gu S Y, Jiang C H. Simulation study of model predictive control for vehicle active suspension based on RBF neural network [J]. Chinese Journal of Construction Machinery, 2025, 23(03): 410–414.
[14] Wei W Z, Xie Q Q, Sun J Z, et al. Research on a double-delay DDPG reinforcement learning control strategy for semi-active suspension [J]. Manufacturing Automation, 2025, 47(06): 85–92.
[15] Zhu X, Chen Z, Zhang S, et al. Intelligent Control Strategy of Vehicle Active Suspension Based on Deep Reinforcement Learning [C]//2022 China Automation Congress (CAC). IEEE, 2022: 4871-4876.
[16] Yu Y F. Research on control of four-corner interconnected air suspension based on multi-agent system [D]. Jinzhou: Liaoning University of Technology, 2023.
[17] Gao Z H, Bao M X, Gao F, et al. LSTM-based probabilistic multimodal anticipated trajectory prediction method [J]. Automotive Engineering, 2023, 45(07): 1145–1152+1162.
[18] Yadav H, Thakkar A. NOA-LSTM: An efficient LSTM cell architecture for time series forecasting [J]. Expert Systems with Applications, 2024, 238: 122333.

Subscription
E-Mail Alert
Most popular papers
Call for Special Issue
Publication Ethics & OA Statement
Plagiarism Policy
Contact Us
Downloads: 4423
Visits: 210940

Sponsors, Associates, and Links

All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.