Active suspension reinforcement learning control method based on deep Q neural network

Abstract

The invention relates to an active suspension reinforcement learning control method based on a deep Q neural network, and belongs to the technical field of automobile dynamic control and artificial intelligence. A reinforcement learning controller main body obtains state observed quantities such as vehicle body acceleration and a suspension dynamic deflection from a suspension system, a strategy is used for determining a reasonable active force to be applied to the suspension system, the suspension system changes the state at a current moment according to the active force, and meanwhile an award value is generated to judge quality of the current active force. By setting a reasonable reward function and combining dynamic data obtained from an environment, an optimal strategy can be determined to determine a size of an active control force so that overall performance of the control system is more excellent under a large amount of training. According to the reinforcement learning controlmethod based on the deep Q neural network, the active suspension system can be dynamically and adaptively adjusted; and therefore, influences caused by factors such as parameter uncertainty, changeable road surface interference and the like which are difficult to solve in a traditional suspension control method are overcome, and riding comfort of passengers is improved as much as possible on the premise that the overall safety of a vehicle is guaranteed.

Description

technical field [0001] The invention mainly relates to the technical field of automobile dynamic control, in particular to an active suspension reinforcement learning control method based on a deep Q neural network. Background technique [0002] With the continuous development of science and technology and the improvement of people's living standards, in the near future, the number of cars in our country will meet the needs of one car for each household. The ride comfort and riding comfort of a car, as indicators that the human body can directly feel, directly determine the value of a car product to some extent. As a part of the vehicle driving system, the vehicle suspension system is very important to the vehicle's ride comfort, ride comfort and handling stability. Once the parameters of the traditional passive suspension are determined, they cannot be changed or adjusted according to the driving conditions of the car, and the performance is limited. The active suspension...

AI Technical Summary

Problems solved by technology

[0003] Traditional suspension control methods such as PID control, skyhook control, sliding mode control, etc. are conservative, rely on the specific model of the system, and even have linear model constraints, and cannot handle the suspension itself well during driving. In the case of uncertain parameters, it cannot adapt to complex and changeable road conditions

During the driving process of the car, the aging of the spring, the oxidation of the damping, and the change of the number of passengers will all bring about inevitable parameter changes, making the model complex and nonlinear, and the traditional control method cannot well solve the problems brought about by the uncertainty. influences

Images