Vibration Control of a Vehicle Active Suspension System Using a DDPG Algorithm

Controlling a vehicle suspension system is an essential technology to reduce vehicle vibrations. However, controlling the suspension system is difficult because it needs to control actively depending on the road conditions. To solve this problem, we propose our active suspension system using a DDPG algorithm, which is one of the reinforcement learning methods. A DDPG algorithm can give the active suspension force automatically. We chose the DDPG algorithm because the other reinforcement learning method, the DQN algorithm, can only output discrete actions, whereas the DDPG algorithm can output continuous actions. The results of our experiments show that the active suspension system we proposed is able to find policies that reduce vehicle vibration.