【作者】 马迪；

【导师】 赵一兵；

【作者基本信息】 大连理工大学 ， 车辆工程（专业学位）， 2017， 硕士

【摘要】 为解决日益严重的交通、环境、能源问题,保证汽车产业可持续健康发展,人们对自主驾驶技术的研究如火如荼,众多科研机构和企业如雨后春笋,积极促进此项技术的蓬勃发展。汽车的自主驾驶技术象征着汽车行业未来发展方向,是新一轮科技革命背景下的新兴产业,而且融合了多项先进技术,譬如无线通信、智能互联、环境感知、车辆运动控制等等技术,因此研究自主驾驶技术对于积极促进汽车产业转型和升级富有重大长远的战略意义。本课题主要对其中的一项关键技术——轨迹跟踪的控制技术展开研究。对此,论文首先论述了课题的研究背景、意义以及现阶段的发展概况;详细分析轨迹跟踪控制这一模块的已有研究方法,找出现存的缺陷并提出一些改进措施;由于车辆系统与行驶道路环境的复杂性,单一的使用非完整约束的运动学模型从根本上无法保证自主车辆行驶的稳定性。因此,为使自主驾驶车辆具有更佳的操稳性,本文先根据MF公式建立轮胎模型并对模型进行仿真实验;然后联合轮胎模型与基于动力学模型建立的车辆系统作为全文所运用的模型基础。在研究控制算法的过程中,首先使用传统的模糊控制算法跟踪期望轨迹,利用第二章建立好的系统模型以及专家经验设计模糊控制器;考虑到单一的模糊控制算法仅按照已有经验这一局限性,本文提出运用PSO智能优化算法优化模糊控制器的参数,由第五章仿真结果知采用此算法后可以确保自主驾驶车辆在正常工况下轨迹跟踪效果理想,并且简要说明了计算流程。由于该方法并未考虑车辆与周边环境等之间的关系,例如执行机构的约束以及车轮和地面的附着关系因素,导致在极限工况时跟踪效果稍差(最大误差达到0.5m)。为解决极限工况时跟踪效果较差难题,论文进一步研究MPC轨迹跟踪控制算法的有效性。按照第二章的车辆系统建立预测方程,并对控制器进行优化求解,设计MPC控制器;鉴于非完整约束存在的缺陷,在建立控制器状态约束条件时把轮胎侧偏角约束以及地面附着系数也加入到了控制器的运算中。最后将本文所使用的两种控制方法进行仿真,结果表明在中低速时两种算法均能够较好的跟踪上期望的轨迹并且误差较小(0.3m以内);在高速时,两种算法在直线跟踪时均具备较好的效果,但在转弯时由于车速过快因此都会有一定的跟踪误差,单一采用FLC时达到1.2m,使用PSO优化后可以缩减至0.5m以内,利用MPC算法可保障误差在0.3m之内。总体来讲,本文采用的轨迹跟踪控制方法均具备较理想的跟踪性能和适应性能,能满足自主驾驶车辆的轨迹跟踪控制要求。更多还原

【Abstract】 In order to tackle the increasingly serious traffic,environment and energy problems and guarantee the sustainable development of auto industry,more and more people have been focusing on autonomous-driving technology.As a result,many scientific research institutions and enterprises have been established.This would be a positive trend to actively promote the vigorous development of the technology.The autonomous driving technology represents future tendency of the auto industry.In addition,it is an emerging industry that can be regarded as new revolution of science and technology,incorporating many advanced technologies such as wireless communication,intelligent interconnection,environment detection and dynamic control of vehicles,etc.Therefore,it is greatly important to put more attention on studying autonomous driving technology,and is of great strategic significance to promote transformation and upgrading of the auto industry.The topic of this thesis mainly concentrates on one of the key technologies-the study of trajectory tracking control.In this regard,the paper first elaborates the research background,significance and the present development situation,then concretely analyzes the trajectory tracking control of the existing research methods.Meanwhile,the existing drawbacks are concluded and some improving measures are proposed.As the complexity of the vehicle system and surrounding driving circumstances,one single kinematics model with non-holonomic constraint fundamentally cannot ensure the stability of the autonomous vehicle.Hence,it is necessary to establish tire model according to MF formula and then implement simulation under different conditions;In addition,tire model and the vehicle dynamics model are used to act as basic system model of this paper.During control algorithm study,the priority is to apply traditional fuzzy control algorithm to track the desired trajectory by adopting system model and fuzzy logic controller based on expert experience.Considering the limitations and drawbacks of fuzzy controller that is based on experience,this paper illustrates PSO intelligent optimization algorithm to optimize the parameters of the fuzzy controller and achieve adaptive fuzzy control.Also,the compute process is briefly introduced.This method,however,does not consider the relationship between the vehicle and the surrounding circumstance,including the constraints of the actuator and the wheel-ground adhesion factors,will result in poor tracking effect when vehicles encounter extreme conditions(with max error=0.2m).Therefore,in order to increase the performance of trajectory tracking control algorithm of autonomous driving vehicles,the paper further studies the MPC trajectory tracking control algorithm.Model predictive equation is established according to the vehicle system in Chapter two,the controller equations are also resolved and optimized.In view of the drawbacks of the non-holonomic constraint model,the constraint conditions with the tire side-slip angle and the coefficient of friction with the ground are added to the controller.The final step is to simulate and compare two control methods that have been proposed in this paper.The results convey that two algorithms can accurately track the desired trajectory at low & intermediate velocity and error is very small(approximately 0.3m).While,there still exists small error when vehicle making turns at high velocity.The error that utilizing FLC only reaches 1.2m and it decreases to 0.5m after using PSO to optimize parameters.Meanwhile,the MPC algorithm minimize error to 0.3m.In conclusion,the improved algorithms possess sound tracking performance and can satisfy the requirement of trajectory tracking control of autonomous driving vehicles.更多还原