文献类型：期刊文献

中文题名：考虑轮胎力耦合约束的智能汽车轨迹跟踪控制算法

英文题名：Intelligent vehicle trajectory tracking algorithm taking into account tire force coupling constraints

作者：孙浩[1,2,3];杜煜[1,2];丁建文[1,2]

第一作者：孙浩

通讯作者：Du, Yu

机构：[1]北京联合大学,北京市信息服务工程重点实验室,北京100101;[2]北京联合大学机器人学院,北京100027;[3]清华大学汽车安全与节能国家重点实验室,北京100084

第一机构：北京联合大学北京市信息服务工程重点实验室

年份：2019

卷号：27

期号：6

起止页码：804-810

中文期刊名：中国惯性技术学报

外文期刊名：Journal of Chinese Inertial Technology

收录：CSTPCD;;EI(收录号：20200908236287);Scopus(收录号：2-s2.0-85080041510);北大核心:【北大核心2017】；CSCD:【CSCD2019_2020】；

基金：汽车安全与节能国家重点实验室开放基金课题(KF2012);国家自然科学基金青年基金(61803034)

语种：中文

中文关键词：智能汽车;轨迹跟踪;轮胎力耦合;模型预测控制;交叉方向乘子法

外文关键词：intelligent vehicle;path tracking;tire force coupling;model predictive control;alternating direction method of multipliers

摘要：针对在低附着变速工况下,忽略纵、侧向轮胎力耦合约束可能导致轨迹跟踪时车辆失稳问题,提出了一种模型预测控制框架下的考虑轮胎力耦合约束的车辆轨迹跟踪控制方法。首先,通过摩擦圆假设建立纵、侧向轮胎力的耦合关系,并推导与之等效的输入量边界约束,将该问题转化为约束二次规划问题;其次,提出了一种基于交叉方向乘子法的数值求解构型,降低了求解约束优化问题时Karush-Kuhn-Tucker方程的维数,实现了求解加速。仿真结果表明,在低附着变速工况下,所提出的算法能够实现最大0.166 m误差的稳定跟踪;同时数值求解过程最多仅需8次迭代,增强了控制过程的实时性。
Aiming at the problem of vehicle instability in trajectory tracking caused by neglecting the coupling constraint of longitudinal and lateral tire forces under the low-adhesion variable speed condition, a vehicle trajectory tracking control method considering the coupling constraint of tire forces under the model predictive control framework is proposed. Firstly, the coupling relationship between longitudinal and lateral tire forces is established based on the friction circle hypothesis, and the equivalent input boundary constraints are derived to transform the problem into a constrained quadratic programming problem. Then, a numerical solution configuration based on a cross-direction multiplier method is proposed, which reduces the dimension of Karush-Kuhn-Tucker equations when solving the constrained optimization problems to accelerate the solution. Simulation results show that the proposed algorithm can guarantee stable tracking with the maximum error of 0.166 m, and the numerical optimization process needs only eight iterations at most, which enhances the real-time performance of the control process.