createnet.m

用RBF神经网络和模糊控制方法控制二级倒立摆源码

%======================================================
% 我的神经网络控制器
%   See also: myneuro4.mdl
%   Lions Wang   BJUT    IAIR  tutor： Ruan Xiaogang.
%   $Revision: 1.00 $
% 好现在开始
%======================================================
%计算设计网络的时间
timer1=clock;
net=network;
%==========================================================================
%%定义网络然后开始初始化
%==========================================================================
% Update the pendulum animation.
% net =
%     Neural Network object:
%     architecture:
%          numInputs: 6
net.numInputs=6;
%          numLayers: 3
net.numLayers=3;
%        biasConnect: [0; 0; 0]
net.biasConnect=[0; 0; 0];
%       inputConnect: [3x6 boolean]
net.inputConnect=[1 1 1 1 1 1;0 0 0 0 0 0; 0 0 0 0 0 0];
%       layerConnect: [0 0 0; 0 0 0; 0 0 0]
net.layerConnect=[0 0 0;1 0 0;0 1 0];
%      outputConnect: [0 0 0]
net.outputConnect=[0 0 1];
%      targetConnect: [0 0 0]
net.targetConnect=[0 0 1];
% net.targets =  [1x1 struct]     []     []
% net.outputs=      []     []    [1x1 struct]
% net.biases
%      []
%      []
%      []
%         numOutputs: 0  (read-only)
%         numTargets: 0  (read-only)
%     numInputDelays: 0  (read-only)
%     numLayerDelays: 0  (read-only)
% 
%     subobject structures:
% 
%             inputs: {6x1 cell} of inputs
%sdandard input 
net.inputs{1}.range=[-7 7];
net.inputs{2}.range=[-0.5 0.5];
net.inputs{3}.range=[-0.05 0.05];
net.inputs{4}.range=[-5 5];
net.inputs{5}.range=[-5 5];
net.inputs{6}.range=[-0.5 0.5];
% > net.layers{1}
%      dimensions: 1
%     distanceFcn: ''
%       distances: []
%         initFcn: 'initwb'
%     netInputFcn: 'netsum'
%       positions: 0
%            size: 1
%     topologyFcn: 'hextop'
%     transferFcn: 'purelin'
%        userdata: [1x1 struct]
% 
net.layers{1}.size=6;
net.layers{1}.initFcn='initnw';
net.layers{1}.transferFcn='tansig';

% >> net.layers{2}
%      dimensions: 1
%     distanceFcn: ''
%       distances: []
%         initFcn: 'initwb'
%     netInputFcn: 'netsum'
%       positions: 0
%            size: 1
%     topologyFcn: 'hextop'
%     transferFcn: 'purelin'
%        userdata: [1x1 struct]
% 
net.layers{2}.size=10;
net.layers{2}.initFcn='initnw';
net.layers{2}.transferFcn='logsig';

% >> net.layers{3}
%      dimensions: 1
%     distanceFcn: ''
%       distances: []
%         initFcn: 'initwb'
%     netInputFcn: 'netsum'
%       positions: 0
%            size: 1
%     transferFcn: 'hextop'
%     transferFcn: 'purelin'
%        userdata: [1x1 struct]

%             layers: {3x1 cell} of layers
%            outputs: {1x3 cell} containing no outputs
%            targets: {1x3 cell} containing no targets
%             biases: {3x1 cell} containing no biases
%       inputWeights: {3x6 cell} containing no input weights
%       layerWeights: {3x3 cell} containing no layer weights
% 
%     functions:
% 
%           adaptFcn: (none)
%            initFcn: (none)
%         performFcn: (none)
%           trainFcn: (none)
% 
%     parameters:
% 
%         adaptParam: (none)
%          initParam: (none)
%       performParam: (none)
%         trainParam: (none)
%net.inputWeights{2,1}.delays=[0 0.005];
net.initFcn='initlay';
net.performFcn='mse';
net.trainFcn='trainlm';
%==========================================================================
%%初步完成初始化神经网络的参数如下
%==========================================================================
% net =
% 
%     Neural Network object:
% 
%     architecture:
% 
%          numInputs: 6
%          numLayers: 3
%        biasConnect: [0; 0; 0]
%       inputConnect: [3x6 boolean]
%       layerConnect: [0 0 0; 1 0 0; 0 1 0]
%      outputConnect: [0 0 1]
%      targetConnect: [1 0 0]
% 
%         numOutputs: 1  (read-only)
%         numTargets: 1  (read-only)
%     numInputDelays: 0  (read-only)
%     numLayerDelays: 0  (read-only)
% 
%     subobject structures:
% 
%             inputs: {6x1 cell} of inputs
%             layers: {3x1 cell} of layers
%            outputs: {1x3 cell} containing 1 output
%            targets: {1x3 cell} containing 1 target
%             biases: {3x1 cell} containing no biases
%       inputWeights: {3x6 cell} containing 6 input weights
%       layerWeights: {3x3 cell} containing 2 layer weights
% 
%     functions:
% 
%           adaptFcn: (none)
%            initFcn: 'initlay'
%         performFcn: 'mse'
%           trainFcn: 'trainlm'
% 
%     parameters:
% 
%         adaptParam: (none)
%          initParam: (none)
%       performParam: (none)
%         trainParam: .epochs, .goal, .max_fail, .mem_reduc, 
%                     .min_grad, .mu, .mu_dec, .mu_inc, 
%                     .mu_max, .show, .time
% 
%     weight and bias values:
% 
%                 IW: {3x6 cell} containing 6 input weight matrices
%                 LW: {3x3 cell} containing 2 layer weight matrices
%                  b: {3x1 cell} containing no bias vectors
% 
%     other:
%==========================================================================
%%对网络进行阈值，权值初始化
%==========================================================================
net=init(net);
%           userdata: (user stuff)
%     weight and bias values:
% 
%                 IW: {3x6 cell} containing no input weight matrices
%                 LW: {3x3 cell} containing no layer weight matrices
%                  b: {3x1 cell} containing no bias vectors
% 
%     other:
% % 
%  net.layerWeights=     []                   []     []
%                   [1x1 struct]              []     []
%                       []              [1x1 struct] []
% >> net.layerWeights{2,1}= delays: 0
%        initFcn: ''
%          learn: 1
%       learnFcn: ''
%     learnParam: ''
%           size: [6 6]
%       userdata: [1x1 struct]
%      weightFcn: 'dotprod'
%======================================================
%建成后先小训一下
%======================================================
for i=1:1:4
  for j=1:1:4
       for k=1:1:4
            for l=1:1:4
                 for m=1:1:4 
                     for n=1:1:4
  coun=j*4^4+k*4^3+l*4^2+m*4+n;
%    coun=i*7^5+j*7^4+k*7^3+l*7^2+m*7+n;
     p(1,coun)=1*(i-2)/4;
     p(2,coun)=0.4*(j-2)/4;
     p(3,coun)=0.04*(k-2)/4;
     p(4,coun)=4*(l-2)/4;
     p(5,coun)=4*(m-2)/4;
     p(6,coun)=0.4*(n-2)/4;
     t(coun)=p(1,coun)*1.00+p(2,coun)*(-231.110)+p(3,coun)*(326.3442)+p(4,coun)*0.9843+p(5,coun)*(-5.5640)+p(6,coun)*52.3631;
                     end
                end
            end
       end
  end
  net=train(net,p,t);
end
%计算设计网络的时间
datat=etime(clock,timer1)
save  net;
%测试一下
p=[0 0.1 0 0 0 0 ]';
output=sim(net,p)
goal=p(1)*1.0000+p(2)*(-231.1100)+p(3)*(326.3442)+p(4)*0.9843+p(5)*(-5.5640)+p(5)*52.3631
% 程序结束，byebye！