基于改进MPC的自动驾驶轨迹跟踪控制  

Automatic Driving Trajectory Tracking Control Based on Improved MPC

作　　者：谢睿 刘广敏 朱凤华[2] 熊刚[2] 

机构地区：[1]山东交通学院,轨道交通学院,山东 济南 [2]中国科学院自动化研究所,多模态人工智能系统全国重点实验室,北京

出　　处：《交通技术》2023年第6期494-501,共8页Open Journal of Transportation Technologies

摘　　要：在自动驾驶车辆的运行过程中,轨迹跟踪控制发挥了十分重要的作用,使得自动驾驶技术更加高效、安全。在轨迹跟踪控制中,系统通常需要根据给定的轨迹或路径,实时调整自身的状态或输出,以使系统能够沿着轨迹进行运动。为了提高轨迹跟踪控制的精度,本文将给出一个基于粒子群优化算法(PSO)的变预测时域模型预测控制(MPC)模型。利用PSO算法计算使下一时刻跟踪精度最优的预测时域大小,并将其应用到MPC控制器模型中实时改变参数,从而达到更好的跟踪效果。在MATLAB软件上与传统MPC等其他轨迹跟踪方法进行对比分析,结果表明改进后的变预测时域MPC模型在提高轨迹跟踪精度方面的表现比传统MPC控制器模型及其他轨迹跟踪方法效果更好,说明本文所提方法能够提升车辆行驶的轨迹跟踪控制精度,并具有很高的实用价值。In the operation process of autonomous vehicles, trajectory tracking control plays a very impor-tant role, making autonomous driving technology more efficient and safe. In trajectory tracking control, the system usually needs to adjust its own state or output in real-time according to the given trajectory or path, so that the system can move along the trajectory. In order to improve the accuracy of trajectory tracking control, this paper presents a variable prediction time domain model predictive control (MPC) model based on particle swarm optimization (PSO). PSO algorithm is used to calculate the prediction time domain size that makes the next time tracking accuracy optimal, and it is applied to the MPC controller model to change the parameters in real-time, so as to achieve a better tracking effect. Compared with other trajectory tracking methods such as traditional MPC on MATLAB software, the results show that the improved variable prediction time domain MPC model has a better performance than the traditional MPC controller model and other trajectory tracking methods in improving trajectory tracking accuracy, indicating that the proposed method can improve the trajectory tracking control accuracy of vehicle running, and has a high practical value.

关 键 词：自动驾驶 轨迹跟踪 模型预测控制 粒子群算法 预测时域 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]