Prev: fmincon minimisation
Next: how to get similar result of pixel region
From: Manan on 15 Mar 2010 12:42
i am trying to write a code for anfis in a three link robotic arm which has:
1) a rotating base (with 360 degree rotation)
2) link1 : a link on top of rotating base which can have angles 0 to 90 degree
3) link2 : a link on top of link1 which can have angles from 0 to 180 degree with respect to link1

i used reference given on http://www.mathworks.com/products/fuzzylogic/demos.html?file=/products/demos/shipping/fuzzy/invkine_codepad.html for writing the code

please explain why THETA1P, THETA2P, THETA3P values are not accurate
given below is the code for the same which gives unsatisfactory result for theta values:

function [ data1, data2, data3, THETA1P, THETA2P, THETA3P ] = anfisdata(XYZ)

%UNTITLED Summary of this function goes here
% Detailed explanation goes here
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% theta1 : link1 angle
% theta2 : link2 angle
% theta3 : rotation angle
% a : link1 length
% b : link2 (end effector link) length
% c : base link length
%
% Training ANFIS data:
% anfis function parameters:
% The first parameter to anfis is the training data
% The second parameter is the number of membership functions used to characterize each input and output (increasing value of this parameter will give improved accuracy but takes a lot of time to train anfis data)
% The third parameter is the number of training epochs
% The last parameter is the options to display progress during training.
% The values for number of epochs and the number of membership functions have been arrived at after a fair amount of experimentation with different values.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

a = 1; % set length of first arm
b = 1; % set length of second arm
c = 0; % set length of base arm

theta1 = 0:1:pi/2; % all possible theta1 values
theta2 = 0:1:pi; % all possible theta2 values
theta3 = 0:1:2*pi; % all possible theta3 values

[THETA1, THETA2, THETA3] = meshgrid(theta1, theta2, theta3); % generate a grid of theta1,theta2 and theta3 values

Z = c + a * sin(THETA1) + b * sin(THETA1 + THETA2); % compute Z coordinates

r = (((a * sin(THETA1) + b * sin(THETA1 + THETA2)).^2 + (a * cos(THETA1) + b * cos(THETA1 + THETA2)).^2).^(.5));
theta = atan((a * sin(THETA1) + b * sin(THETA1 + THETA2))./(a * cos(THETA1) + b * cos(THETA1 + THETA2)));

X = r .* cos(THETA3) .* cos(theta); % compute x coordinates
Y = r .* sin(THETA3) .* cos(theta); % compute y coordinates

data1 = [X(:) Y(:) Z(:) THETA1(:)]; % create x-y-z-theta1 dataset
data2 = [X(:) Y(:) Z(:) THETA2(:)]; % create x-y-z-theta2 dataset
data3 = [X(:) Y(:) Z(:) THETA3(:)]; % create x-y-z-theta3 dataset

% plot(X(:), Y(:), Z(:));
% axis equal;
% xlabel('X')
% ylabel('Y')
% zlabel('Z')
% title('X-Y-Z co-ordinates generated for all theta1, theta2 and theta3 combinations using forward kinematics formulae')

fprintf('-->%s\n','Start training first ANFIS network. It may take one minute depending on your computer system.')
anfis1 = anfis(data1, 4, 100, [0,0,0,0]); % train first ANFIS network
fprintf('-->%s\n','Start training second ANFIS network. It may take one minute depending on your computer system.')
anfis2 = anfis(data2, 4, 100, [0,0,0,0]); % train second ANFIS network
fprintf('-->%s\n','Start training third ANFIS network. It may take one minute depending on your computer system.')
anfis3 = anfis(data3, 4, 100, [0,0,0,0]); % train third ANFIS network

XYZ=[sqrt(5) 0 0]

THETA1P = evalfis(XYZ, anfis1); % theta1 predicted by anfis1
THETA2P = evalfis(XYZ, anfis2); % theta2 predicted by anfis2
THETA3P = evalfis(XYZ, anfis3); % theta3 predicted by anfis3

end

please explain why THETA1P, THETA2P, THETA3P values are not accurate
 | 
Pages: 1
Prev: fmincon minimisation
Next: how to get similar result of pixel region