基于粒函數(shù)的智能無(wú)人駕駛泊車(chē)仿真

楊二狼 殷業(yè) 邵志情

摘要： 首先建立了小車(chē)的數(shù)學(xué)模型，采用傳統(tǒng)的模糊控制方法對(duì)小車(chē)的自動(dòng)倒車(chē)入庫(kù)進(jìn)行仿真，分析小車(chē)的行駛軌跡，以小車(chē)的車(chē)身參數(shù)、初始位置參數(shù)、終止位置參數(shù)為約束，對(duì)小車(chē)泊車(chē)過(guò)程的軸向角（前輪與車(chē)軸之間的夾角）進(jìn)行調(diào)整，并提出基于粒函數(shù)的方法，擬合出一條精確的點(diǎn)響應(yīng)函數(shù)，從而設(shè)計(jì)出一個(gè)粒函數(shù)控制系統(tǒng).實(shí)驗(yàn)證明粒函數(shù)控制系統(tǒng)能夠達(dá)到良好的效果.

關(guān)鍵詞：

智能控制； 自動(dòng)駕駛； 模糊控制； 粒函數(shù)控制； 泊車(chē)

Received date： 20171205

Biography： Yang Erlang （1993-），male，master，research area：Fuzzy information granular computation.Email：night_yang@163.com

*Corresponding author： Yin Ye （1961-），male，associate professor，research area：Fuzzy control and granular computation.Email：yinye@188.com

引用格式： 楊二狼，殷業(yè)，邵志情.基于粒函數(shù)的智能無(wú)人駕駛泊車(chē)仿真 [J].上海師范大學(xué)學(xué)報(bào)（自然科學(xué)版），2018，47（2）：263-266.

Citation format： Yang Erlang，Yin Ye，Shao Zhiqing.Parking simulation of intelligent SelfDriving based on the granular function [J].Journal of Shanghai Normal University （Natural Sciences），2018，47（2）：263-266.

Intelligent control has been widely used in various fields such as industry，agriculture，service industry and military[1].In recent years，artificial intelligence has been gradually becoming a hot topic.Automatic driving has also occupied one of the focuses in the field of artificial intelligence[2].With the number of vehicles increasing，it is hard to find parking space in the parking lot.If there is an intelligent system that assists the driver in parking，it will greatly reduce the difficulty of parking and the chance of collision during parking.

Mukherjee，et al.[3] studied the nonholonomic constrained path planning of automatic parking.Chang，et al.[4] adopted the method of fuzzy logic for automatic parking control.Zhao，et al.[5] used fuzzy logic method for small parking space by automatic parking control.Guo，et al.[6] combined fuzzy theory with expert experience to verify the feasibility of fuzzy control algorithm.In this paper，it set up the mathematical model in Matlab and simulated the car′s parking process，using a simple parking model.By taking the car′s body parameters，the initial position parameters and the termination position parameters as constraints，it designed a fuzzy controller and constructed a fuzzy control system of the car parking based on the traditional fuzzy control theory.Then，it fit the trajectory of the car，and eventually acquired the particle function[7] on the basis of the fuzzy control system，established the construction of the control system using the granular function，and made relevant parameter analysis of the experiment.

上海師范大學(xué)學(xué)報(bào)（自然科學(xué)版）J. Shanghai Normal Univ.（Nat. Sci.）

2018年

第2期楊二狼，殷業(yè)，邵志情：基于粒函數(shù)的智能無(wú)人駕駛泊車(chē)仿真

1System design

In the process of car parking，the whole control system is composed of various modules such as car body parameter constraint，parking space position，control system，trajectory planning and car movement.

Once the car automatic parking system is turned on，the car will detect the parking space around.If there is parking space，it will determine the initial position of the car and the parking end position，and will figure out whether the car is already in the parking space.If the car is in parking space，the system will inform the car to stop moving.Otherwise，the system will calculate the car body parameters and send them to the car′s control system.After the control system processing these parameters，the motion variables are sent to the car′s motion model， and the car moves until stops in an empty parking space.

2Granular Function

Granule is the core of granular computation.For different research objects，there are different granule layers and grain structures[8-9].In the same layer of granule，granular function is a mapping between a granule set and another one.

Granular function is represented by a set of precise functions.The number of these precise functions can be infinite.Point responding function is the function that could achieve the optimal output according to a certain rule.This paper tries to solve the conversion function between granular function and point responding function by the fuzzy set.X and Y are ndimensional European space，and {A1，A2…An′}，{B1，B2…Nn′} are respectively primitive elements in the set of X and Y.F is the fuzzy rule of mapping between X and Y.Because each fuzzy primitive element is a fuzzy granule，it can acqiure the granular function from fuzzy set X to Y by the fuzzy rule of F[X]→[Y].

However，the granular function obtained cannot be put into practice because it is a set of fuzzy function.Thus，it should be transformed into point responding function through granulation sampling.

3Simulation

This paper runs the system module established by Matlab platform.This system mainly includes parameter initialization，granular function controller，car movement model，animation and display and other major components.

Granular function can be obtained from fuzzy input and output by fuzzy rules.After granular sampling，it is able to acquire the point responding function which can be put into practice.In this paper，the fuzzy variables of the distance between the car and the target position and the axial angle are sampled，and are fitted through polynomial equation，and finally the point responding function is acquired.

The simulation system has three input variables：Distance between current position and target position（d），axial angel（a1），the front wheel rotation angel（a2），thus the point responding function is as follows：

F（d，a1，a2）=∑ni=0∑n-ij=0∑n-i-jpPijpdia1ja2p，（1）

i，j and p donates dimension of d，a1，and a2，n donates maximum dimension and Pijp donates current coefficient respectively.

The simulation sample adoptes fivefold fits（n=5），which can achieve a better fitting output.In order to acquire a perfect control effect，in this paper，the fuzzy variable of distance adopts the piecewise fitting：d<=1.5，1.54.7.

4Experiment & Analysis

The car′s animation simulation inside the Animation & Human Controller module is achieved.Fig.1 is the Animation simulation.

Figure 1Animation simulation

As shown in the animation car simulation in Fig.1，the car can reach the parking space at the appointed target position from the specified starting position，and the parking trajectory of the car is smooth，which means that the simulation is practical and effective.

5Conclusion

On the basis of the original fuzzy control system，a new method of particle function is proposed in this paper.This method not only eliminates the process of establishing a large number of fuzzy rules，but also obtains good control effect.It provides a new method for parking in the field of intelligent control.The car parking system proposed in this paper is based on the car model and parking space model established by theoretical information，however，in reality，the models of vehicles and the parking spaces are varied and diverse，the system should be changed with the circumstances changing.

References：

[1]Wang Y N.Integrated and integrated intelligent control of complex industrial processes [J].Information And Control，1999，28（4）：198-304.

[2]Yang R.Autonomous driving into artificial intelligence focus [J].Computer and Network，2017（5）：14-15.

[3]Mukherjee R，Anderson D P.A surface integral approach to the motion planning of nonholonomic systems [C].American Control Conference，San Francisco：IEEE，1993.

[4]Chang S J，Li T H S.Design and implementation of fuzzy parallelparking control or a cartype mobile robot [J].Journal of Intelligent and Robotic Systems，2004，34（2）：175-194.

[5]Zhao Y N，JrCollins E G.Robust automatic parallel parking in tight space via fuzzy logic [J].Robotics and Autonomous Systems，2005，51：111-127.

[6]Guo K H.Automatic parallel parking steering controller based on fuzzy logic [J].Journal of Jilin University （Engineering Edition）.2009，39（S2）：236-240.

[7]Yin Y.The essence of fuzzy control [J].Scientific frontier，2015，33（9）：16-27.

[8]Tang W.Design and simulation of fuzzy control system based on granular function [J].Journal of Shanghai Normal University （Natural Sciences）.2017，46（2）：255-261.

[9]Ke D Y.Research on fuzzy control algorithm based on granular function [D].Shanghai：Shanghai Normal University，2016.

（責(zé)任編輯：包震宇）