基于增强学习的车辆转弯制动横向轨迹控制  

Lateral Trajectory Control of Vehicle Turning and Braking Based on Reinforcement Learning

作　　者：罗锦才 刘暾东 LUO Jin-cai;LIU Tun-dong(School of Aerospace Engineering,Xiamen Uniersity,Xiamen Fujian 361102,China)

机构地区：[1]厦门大学航空航天学院,福建厦门361102

出　　处：《计算机仿真》2021年第7期107-110,共4页Computer Simulation

摘　　要：针对传统的车辆转弯制动横向轨迹控制方法存在车辆整体滑移率高,控制后的车辆滑行时间较长、滑行路程较远的问题,提出基于增强学习的车辆转弯制动横向轨迹控制方法。构建车辆SAE坐标系,使用三角函数得到轮胎六分力的联合作用情况,通过拟合计算的形式,获取水平方向漂移数据及其侧向力,构建车辆动力模型,将车辆动力模型转化为线性的二自由度系统,设定车辆滑移率为目标函数,根据计算的函数最优解,使用增强学习算法选取最佳控制策略,从而完成基于增强学习的车辆转弯制动横向轨迹控制方法设计。构建仿真环节,通过与目前使用方法对比可知,上述方法的车辆整体滑移率较低,控制效果较为稳定。In the process of lateral trajectory control of vehicle turning and braking, the traditional method has the disadvantages of high overall slip rate, long sliding time and long sliding distance. Therefore, a novel lateral trajectory control method based on reinforcement learning was proposed in this paper. The vehicle SAE coordinate system was established. Through trigonometric function, the combined action state of tire six component force was obtained. The horizontal drift data and lateral force were obtained according to the form of fitting calculation, and the vehicle dynamic model was constructed. The vehicle dynamic model was transformed into a linear two degree of freedom system. The vehicle slip rate was set as the objective function. Based on the optimal solution of the calculated function, the reinforcement learning algorithm was applied to select the optimal control strategy, realizing the design of the lateral trajectory control method of vehicle turning and braking based on reinforcement learning. The simulation results show that the method has low overall slip rate and stable control effect.

关 键 词：转弯制动 横向轨迹 增强学习 三角函数 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]