Research on Deep Reinforcement Learning Control Algorithm for Active Suspension Considering Uncertain Time Delay.

Area of Science:

• Automotive Engineering
• Control Systems
• Artificial Intelligence

Background:

• Actuator delays in active suspension systems critically impact control effectiveness and stability.
• Existing control algorithms often struggle with uncertain or variable system delays.
• Ensuring stable control performance requires addressing these inherent delays.

Purpose of the Study:

• To develop a novel active suspension control algorithm capable of handling uncertain actuator delays.
• To improve the stability and control performance of active suspension systems.
• To evaluate the algorithm's effectiveness under various delay conditions.

Main Methods:

• Utilized a twin-delayed deep deterministic policy gradient (TD3) algorithm tailored for systems with delay.
• Developed a deep reinforcement learning (DRL) approach to optimize control policies.
• Simulated performance across three conditions: deterministic, semi-regular, and uncertain delay.

Main Results:

• The proposed DRL algorithm demonstrated excellent control performance across all tested delay conditions.
• Achieved over 30% improvement in body vertical acceleration optimization compared to passive suspension.
• Effectively mitigated low-frequency body vibrations and enhanced ride comfort by over 30%.

Conclusions:

• The novel DRL-based control algorithm effectively addresses uncertain actuator delays in active suspension systems.
• The approach offers significant improvements in ride comfort and vibration reduction.
• The algorithm shows strong potential for practical application in automotive active suspension.