基于模型预测控制的智能汽车轨迹跟踪研究开题报告

1. 研究目的与意义（文献综述）

1.1 Purpose and meaning of research

The overall purpose of the study is to designa autonomous light electric vehicle (reference model: a certain type ofhigh-speed imitation target vehicle), the vehicle parameters are shown in thefollowing table:

Thecar hopes to have three operating modes, auto mode, remote control mode andmanual driving mode. In automatic mode, the target car is able to travelautomatically at the preset speed on the set track. In remote control mode, thetarget car can be remotely controlled by the controlcommand. Manual driving mode, mainly used in the transfer and deployment of thetarget vehicle, the driver can directly drive the target vehicle to completethe preparation, commissioning, access to the warehouse, withdrawal and otherprocesses.

The part I am responsible for is the model predictivecontrol of intelligent car tracking research, which is one of the keytechnologies to achieve the target car automatic driving mode. On the premiseof maintaining the tracking accuracy and stability, the controller sendscontrol instructions such as steering wheel Angle and pedal depth to controlthe steering wheel of the vehicle's actuator, accelerator/brake pedal, etc., sothat the intelligent car can follow the reference track ^[1].

1.2 Analysis of thecurrent state of research

The purpose of trajectory tracking control isto allow autonomous vehicles to track the trajectory obtained by the planning algorithm ^[2]. Its main task is to output the corresponding control parametersaccording to the kinematic and dynamic constraints of the vehicle, such as frontwheel angle, wheel braking power, etc. Considering the multi-degree of freedomof the car, the complexity of the vehicle chassis, the coupling between thetire and the suspension, the interference between the suspension and thesteering system, and the complexity of the road conditions, all add to thecomplexity of the vehicle tracking process. Many algorithms are also working tosolve these problems ^[3].

At present, the main popular trajectorytracking control algorithms are: pre-scanning control algorithm, pure trackingalgorithm, PID control, feed-forward feedback control, linear secondaryregulator LQR tracking controller, model predictive control, etc. ^[4]

In the study of the trajectory tracking of autonomouscar, many researchers from enterprises and universities have made a lot ofresearch. As early as more than 30 years ago, many national scientific researchinstitutions and university scholars started research on the Imitation DriverModel. In the research of vehicle tracking technology in China, the horizontaland vertical speed control of vehicles, most enterprises or universities have adoptedthe optimal preview control theory put forward by Professor Guo Konghui andothers ^[5].

Based on the preview hypothesis and theprinciple of optimal curvature control is a common method of optimal previewtheory control. In principle, the preview control is modeled after the driver'sactual driving process, and based on the observation of the road conditionsaround the vehicle. The next action is decided, including the steering wheelsteering degree, accelerator/brake pedal. However, the steering degree of thevehicle steering wheel is usually determined by the bending degree of the roadahead, and the greater the bending degree, the greater the steering degree ^[6].The autonomous vehicle based on preview adjusts the vehicle's steering andspeed according to the above ideas. At present, the mostly applied puretracking algorithm is to use this principle to realize the trajectory trackingcontrol of the intelligent vehicle unmanned vehicle. Firstly, the error betweenthe current vehicle position and the reference trajectory is obtained, thepreview distance and the preview speed are determined again, the curvature ofthe optimal arc trajectory is calculated, and the front wheel and steeringwheel angle are obtained ^[7].However, when the vehicle is running athigh speed, the algorithm will have a large error ^[8].

The pure tracking algorithm uses the kinematicgeometric model of the vehicle. Only when the vehicle speed is relatively lowcan the algorithm achieve a better trajectory tracking effect ^[9].

PID control is a widely used control algorithmin the field of industrial control. If applied to autonomous vehicles, it hasthe advantage of not having to build a model. However, its control parametersneed to be tested again and again, which is a tedious and time-consuming work. Whenthe speed changes, the current control parameters are not suitable to controlthe speed, need to try again to gather the control parameters. Therefore, PIDis simple but has a poor adaptability to speed ^[10]. Other vehicleparameters or road environment parameters also have great influence on PIDcontrol. At present, the application of PID algorithm in intelligent car hasbeen reduced.

Feedforward feedback control usually does asupplementary control in a master algorithm. Using the information of the roadahead to compensate the disturbance amount to form feedforward control. The vehicle'scurrent state variable is used to adjust the control input to correct theactual trajectory, forming a feedback control ^[11].

The tracking controller based on linearquadratic regulator (LQR) is used to optimize the track tracking of vehicles athigh speed. Its principle is that in the control time domain, the trackingerror model in the whole system will be linearized to obtain a linear quadraticmodel for easy calculation. According to the requirements of the whole system,an optimal linear quadratic function is set up, which is optimally solvedglobally to obtain the optimal trajectory control input ^[12]. LQR isessentially a linear optimization algorithm, which does not consider the impactof vehicle dynamics constraints and external factors of vehicle environment. Vehicleside deflection may occur when driving in bad conditions. Moreover, due to thehigh precision required by the controller, the complexity of various parametersin real vehicle verification cannot achieve the ideal effect in simulation.

Model predictive control algorithm is known asMPC algorithm, which is a kind of optimization control algorithm widely used intransportation, manufacturing and other industrial fields. At present, afteryears of development, it has been applied in industrial field and intelligentcontrol field ^[13]. With the improvement of the algorithm, it hasbecome a kind of classical algorithm which is often used in the industrialfield. Although there are differences in the application of algorithms invarious fields. However, they all follow the basic principles of predictionmodel, rolling optimization and feedback correction ^[14]. In recentyears, the application of model prediction algorithm is more and more popularin automobile industry, especially in vehicle intelligent driving technology. Manyresearch institutions have used the principle of model prediction to study thetrajectory tracking control of intelligent vehicles. Model prediction algorithmis good at solving optimization problems with constraints, among which itsrolling optimization and feedback correction characteristics have theadvantages of reducing errors caused by system time lag and improving controlperformance ^[15].

2. 研究的基本内容与方案

2.1basic contents

(1) analyze the research status of intelligent vehicletrajectory tracking in domestic and abroad;

(2) compare the advantages and disadvantages andapplication scenarios of the existing intelligent vehicle trajectory trackingmethods;

3. 研究计划与安排

weeks 1-2 (weeks 18-19 of the 7th semester):determine graduation project topics, improve graduation project assignments(relevant parameters), and collect data inside and outside the school

weeks 3-4 (weeks 1-2 of the 8th semester):scheme idea, literature search, completion report

week 5 (3rd week of the 8th semester): foreignlanguage translation, data collection

4. 参考文献（12篇以上）

[1]由智恒. 基于mpc算法的无人驾驶车辆轨迹跟踪控制研究[d].吉林大学,2018:3-4.

[2]刘凯,陈慧岩,龚建伟,陈舒平,张玉.高速无人驾驶车辆的操控稳定性研究[j].汽车工程,2019,41(05):514-521.

[3]段建民,田晓生,夏天,宋志雪.基于模型预测控制的智能汽车目标路径跟踪方法研究[j].汽车技术,2017(08):6-11.