A Study on Active Suspension System with Reinforcement Learning
- Publication Type:
- Thesis
- Issue Date:
- 2022
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
The research conducted in this paper focuses on providing a superior solution for controlling the active suspension system through machine learning algorithm. An active suspension system is embedded with an extra actuator and can provide more flexibility in dealing with different road situation than a passive suspension system that only contains a fixed set of spring and damper.
Among various control methods, the machine learning control strategy has demonstrated its optimality in dealing with different classes of roads depending on its self-learning capability. In this study, an advanced method of creating pavement signal has been presented to guarantee the quality of the simulation, together with Twin Delayed Deep Deterministic Policy Gradients (TD3) in suspension control that is an application of one of the cutting-edge algorithms in Reinforcement Learning (RL). To be able to realize the research, policy gradient algorithm, Markov decision process, and neural network modelling have been adopted.
To achieve such a proposed frame structure, a vehicle suspension model has been established together with a frame of reinforcement learning algorithm and an input signal of road pavement. The performance of the proposed twin delayed reinforcement agent has been compared against Deep Deterministic Policy Gradients (DDPG) and Deep Q-Learning (DQN) algorithms under different types of pavement input. The simulation result shows its superiority, robustness, and learning efficiency over other reinforcement learning algorithms.
Please use this identifier to cite or link to this item: