Ride Comfort Improvement with Preview Control Semi-active Suspension System Based on Supervised Deep Learning

As known to all, it is a challenging task to solve the delay of a controllable suspension system under the transient road. Thus, how to effectively and low-costly acquire road information and choose the reasonable control algorithm remains a hot topic in both academia and industry. With the rapid development and extensive application of the advanced intelligent driving system, a large number of sensors, such as cameras, have been installed on the vehicle, and deep learning technology has also been widely used to identify lane recognition, traffic direction signal, and pedestrian detection, but rarely used in semi-active suspension control. To address the above issues, a novel skyhook preview control (SPC) approach, which combines supervised deep learning, is proposed in the article. Firstly, a full vehicle dynamics model for semi-active suspension is established. Secondly, supervised deep learning (YOLOv3) is adopted to identify the transient road to preview the semi-active suspension. Finally, by comparing the vehicle road test results of the SPC with that of the skyhook wheelbase preview controller (SWPC), the skyhook controller (SC), and the passive suspension (PS), it is concluded that the designed algorithm can effectively improve the ride comfort of the vehicle.