模糊控制详细讲解实例.docx

1、模糊控制详细讲解实例一、速度控制算法：首先定义速度偏差-50 km/he（k）50km/h，-20ec（i）= e（k）- e（k-1）20，阀值eswith=10km/h设计思想：油门控制采用增量式PID控制算法,刹车控制采用模糊控制算法，最后通过选择规则进行选择控制量输入。选择规则：e（k）- eswith and throttlr_10 选择油门控制 否则：先将油门控制量置0，再选择刹车控制0e（k） 先选择刹车控制，再选择油门控制e（k）=0 直接跳出选择刹车控制：刹车采用模糊控制算法1.确定模糊语言变量e基本论域取-50,50，ec基本论域取-20,20，刹车控制量输出u基本论域取-

2、30,30，这里我将这三个变量按照下面的公式进行离散化： 其中，n为离散度。E、ec和u均取离散度n=3，离散化后得到三个量的语言值论域分别为：E=EC=U=-3,-2,-1,0,1,2,3其对应语言值为 NB,NM,NS,ZO, PS,PM,PB 2.确定隶属度函数E/EC和U取相同的隶属度函数即： 说明：边界选择钟形隶属度函数，中间选用三角形隶属度函数，图像略实际EC和E输入值若超出论域范围，则取相应的端点值。3.模糊控制规则由隶属度函数可以得到语言值隶属度（通过图像直接可以看出）如下表：表1：E/EC和U语言值隶属度向量表-3-2-10123NB10.500000P0NM010.5000

3、0P1NS00.510.5000P2ZO000.510.500P3PS0000.510.50P4PM00000.510P5PB000000.51P6设置模糊规则库如下表：表2：模糊规则表UEECNBNMNSZOPSPMPBNBPBPBPMPMPSZOZONMPBPMPMPSZOZONSNSPMPMPSPSZONSNSZOPMPSPSZOZONSNMPSPSPSZOZOZONSNMPMPSZOZOZONSNMNBPBZOZOZONS*NMNMNB3.模糊推理由模糊规则表3可以知道输入E与EC和输出U的模糊关系，这里我取两个例子做模糊推理如下：if (E is NB) and (EC is NM)

4、 then (U is PB)那么他的模糊关系子矩阵为：其中，,即表1中NB对应行向量，同理可以得到， , if (E is NVB or NB) and (EC is NVB) then (U is PVB) 结果略按此法可得到27个关系子矩阵，对所有子矩阵取并集得到模糊关系矩阵如下：由R可以得到模拟量输出为：4.去模糊化由上面得到的模拟量输出为17的模糊向量，每一行的行元素（u（zij）对应相应的离散变量zj，则可通过加权平均法公式解模糊：从而得到实际刹车控制量的精确值u。油门控制：油门控制采用增量式PID控制，即：只需要设置、 、三个参数即可输出油门控制量。二、程序实现及参数调节clea

5、r all%*模糊算法%/*隶属度向量 *%P0=1,0.5,0,0,0,0,0;%*NBP1=0,1,0.5,0,0,0,0;%*NMP2=0,0.5,1,0.5,0,0,0;%*NSP3=0,0,0.5,1,0.5,0,0;%*ZOP4=0,0,0,0.5,1,0.5,0;%*PSP5=0,0,0,0,0.5,1,0;%*PMP6=0,0,0,0,0,0.5,1;%*PB%*语言值 *%NB=-3;NM=-2;NS=-1;ZO=0;PS=1;PM=2;PB=3;%/*模糊规则表*%Pg=PB PB PM PM PS ZO ZO; PB PM PM PS ZO ZO NS; PM PM PS

6、 PS ZO NS NS; PM PS PS ZO ZO NS NM; PS PS ZO ZO ZO NS NM; PS ZO ZO ZO NS NM NB; ZO ZO ZO NS NM NM NB;%/*根据规则表计算模糊关系矩阵*%R1_=dikaer(xbing(P0,P1),7,P0,7);R1_=reshape(R1_,1,49);R1=dikaer(R1_,49,P6,7);R2_=dikaer(xbing(P2,P3),7,P0,7);R2_=reshape(R2_,1,49);R2=dikaer(R2_,49,P5,7);R3_=dikaer(P0,7,P1,7);R3_=r

7、eshape(R3_,1,49);R3=dikaer(R2_,49,P6,7);R4_=dikaer(xbing(P1,P2),7,P1,7);R4_=reshape(R4_,1,49);R4=dikaer(R4_,49,P5,7);R5_=dikaer(P3,7,P1,7);R5_=reshape(R5_,1,49);R5=dikaer(R5_,49,P4,7);R6_=dikaer(xbing(P0,P1),7,P2,7);R6_=reshape(R6_,1,49);R6=dikaer(R6_,49,P5,7);R7_=dikaer(xbing(P2,P3),7,P2,7);R7_=res

8、hape(R7_,1,49);R7=dikaer(R7_,49,P4,7);R8_=dikaer(P0,7,P3,7);R8_=reshape(R8_,1,49);R8=dikaer(R8_,49,P5,7);R9_=dikaer(xbing(P1,P2),7,P3,7);R9_=reshape(R9_,1,49);R9=dikaer(R9_,49,P4,7);R10_=dikaer(P3,7,P3,7);R10_=reshape(R10_,1,49);R10=dikaer(R10_,49,P3,7);R11_=dikaer(xbing(P0,P1),7,P4,7);R11_=reshape(

9、R11_,1,49);R11=dikaer(R11_,49,P4,7);P45=xbing(P4,P5);R12_=dikaer(xbing(P2,P3),7,P45,7);R12_=reshape(R12_,1,49);R12=dikaer(R12_,49,P3,7);R13_=dikaer(P0,7,P5,7);R13_=reshape(R13_,1,49);R13=dikaer(R13_,49,P4,7);R14_=dikaer(P1,7,P5,7);R14_=reshape(R14_,1,49);R14=dikaer(R14_,49,P3,7);P01=xbing(P0,P1);R15

10、_=dikaer(xbing(P01,P2),7,P6,7);R15_=reshape(R15_,1,49);R15=dikaer(R15_,49,P3,7);R16_=dikaer(P3,7,P6,7);R16_=reshape(R16_,1,49);R16=dikaer(R16_,49,P2,7);R17_=dikaer(P4,7,P0,7);R17_=reshape(R17_,1,49);R17=dikaer(R17_,49,P4,7);R18_=dikaer(xbing(P5,P6),7,P0,7);R18_=reshape(R18_,1,49);R18=dikaer(R18_,49,

11、P3,7);R19_=dikaer(xbing(P4,P5),7,P1,7);R19_=reshape(R19_,1,49);R19=dikaer(R19_,49,P3,7);R20_=dikaer(P6,7,xbing(P1,P2),7);R20_=reshape(R20_,1,49);R20=dikaer(R20_,49,P2,7);P23=xbing(P2,P3);R21_=dikaer(P4,7,xbing(P23,P4),7);R21_=reshape(R21_,1,49);R21=dikaer(R21_,49,P3,7);R22_=dikaer(P5,7,xbing(P23,P4)

12、,7);R22_=reshape(R22_,1,49);R22=dikaer(R22_,49,P2,7);R23_=dikaer(P6,7,xbing(P3,P4),7);R23_=reshape(R23_,1,49);R23=dikaer(R23_,49,P1,7);R24_=dikaer(P4,7,P5,7);R24_=reshape(R24_,1,49);R24=dikaer(R24_,49,P2,7);R25_=dikaer(P5,7,P5,7);R25_=reshape(R25_,1,49);R25=dikaer(R25_,49,P1,7);R26_=dikaer(P6,7,xbin

13、g(P6,P5),7);R26_=reshape(R26_,1,49);R26=dikaer(R26_,49,P0,7);R27_=dikaer(xbing(P4,P5),7,P6,7);R27_=reshape(R27_,1,49);R27=dikaer(R27_,49,P1,7);m=R1,R2,R3,R4,R5,R6,R7,R8,R9,R10,R11,R12,R13,R14,R15,R16,R17,R18,R19,R20,R21,R22,R23,R24,R25,R26,R27;R=bingji(m);%*初始化参量e=0;ec=0;y_1=0;y_2=0;u=0;u_1=0;u_2=0;

14、u_3=0;e_1=0;e_2=0;Eswith=10;throttle_1=0;brake_1=0;x=0 0 0;ts=0.001;sys=tf(1,1,2,1,inputdelay,0.5);dsys=c2d(sys,ts,zoh);num,den=tfdata(dsys,v);for k=1:1:40000%*控制系统time(k)=k*ts;if(k2000) throttle=2000;end%*刹车控制%/*压缩输入变量*%E=lisan(-50,50,3,e);EC=lisan(-20,20,3,ec);%/*计算实际输入变量隶属度向量*%E_R(1)=lbell(E,1,4,

15、-3);E_R(2)=trig(E,-3,-2,0);E_R(3)=trig(E,-3,-1,1);E_R(4)=trig(E,-2,0,2);E_R(5)=trig(E,-1,1,3);E_R(6)=trig(E,0,2,3);E_R(7)=rbell(E,1,4,3);EC_R(1)=lbell(EC,1,4,-3);EC_R(2)=trig(EC,-3,-2,0);EC_R(3)=trig(EC,-3,-1,1);EC_R(4)=trig(EC,-2,0,2);EC_R(5)=trig(EC,-1,1,3);EC_R(6)=trig(EC,0,2,3);EC_R(7)=rbell(EC,

16、1,4,3);%/*模糊推理过程*%U_R1=dikaer(E_R,7,EC_R,7);U_R1=reshape(U_R1,1,49);U_R2=jdikaer(U_R1,49,R,7);U_R=max(U_R2);u_L=mean(U_R);%/*去模糊化*%brake=-flisan(-150,150,3,u_L);e_2=e_1;e_1=e;%/*选择规则if (e-Eswith)|(throttle_1=0) if(throttle=throttle_1) throttle_1=throttle; u=throttle; Q(k)=u; W(k)=0; else throttle=0;

17、 throttle_1=throttle; u=throttle; Q(k)=u; W(k)=0; end else if(throttle_1=0) brake_1=brake; u=brake; W(k)=u; Q(k)=0; else throttle=0; throttle_1=throttle; u=throttle; Q(k)=u; W(k)=0; end endelse if(e=0) if(brake_1=0) throttle_1=throttle; u=throttle; Q(k)=u; W(k)=0; else brake=0; brake_1=brake; u=brak

18、e; end else u=0; W(k)=0; Q(k)=0; endend%*M(k)=u;%*迟滞环节if (time(k)=0.5) u=0; else u=M(k-0.5/ts);endend%*画图figure(1);plot(time,vd,r,time,y,k,linewidth,2);xlabel(time(s);ylabel(vd,y);legend(,);figure(2);plot(time,Q,r,linewidth,2);xlabel(time(s);ylabel(u);figure(3);plot(time,W,r,linewidth,2);xlabel(time(s);ylabel(u);figure(4);plot(time,N,b);xlabel(times);ylabel(e);程序说明：仿真分加速和加速两个阶段，加速阶段主要应用油门控制，加速阶段主要由刹车控制。加速过程中，调节PID的三个参数kp、Ti、Td，调节过程中不难发现，kp越大调节时间越长，Ti越小稳态误差越大，