How to ensure Matrixes are matching size

I am doing a recursive least squares model to try and reduce sensor error for a project, but I keep getting the error that the left and right matrixes in the for loop are mismatched- the left is 4- by 1 and the right is 4 by four, and I'm not sure where it's going wrong, or how to fix it.

function [Qhat]= RLS_estimation(Z,Unite,T) %

% Call A matrix from gen sate, final matrix
n=0;
P(:,:,1)=eye(4);
R=fixed.forgettingFactor(Z);
A=[1,T,0,0;0,1,0,0;0,0,1,T;0,0,0,1];
K = zeros(4, 1,n); 
% AT is otranspose A, transpose A here
% Call matrix R with 'forgetting factors"
Qhat(:,:,1)=Unite;

for n= 2:Z
    K(:,:,n)=P(:,:,n-1)*A.'*inv(A*P(:,:,n-1)*A.'+R);
    P(:,:,n)=inv(R)-K(:,:,n)*A*inv(R)*P(:,:,n-1);
    Qhat(:,:,n)=Qhat(:,:,n-1)+K(:,:,n)*(Unite(n)-A*Qhat(:,:,n-1));
end

[Walter Roberson]

We do not know the size of Z or Unite or T (though we can deduce that T must be scalar)

You did not indicate which line the problem is occuring on.

[William]

The problem is occuring on the first line of the for loop, where we calculate K for a given N value. Though I belive it may occur for the next two lines as well.

T is a scalar between one and zero, Z is a scalar, which was 60 when I was getting this error.

Unite is a four by Z matrix

Answer 1

[Walter Roberson]

K = zeros(4, 1,n); 

K is initialized to 4 x 1 by something.

    K(:,:,n)=P(:,:,n-1)*A.'*inv(A*P(:,:,n-1)*A.'+R);

P is 4 x 4 by something.

A is 4 x 4.

P(:,:,n-1) * A.' is 4 x 4 x 1 mtimes transpose(4 x 4) so that part is (4 x 4) * (4 x 4) which gives a 4 x 4 result.

inv(A etc) is inv() of 4 x 4, so is 4 x 4 itself.

(4 x 4) * (4 x 4) gives a 4 x 4 result.

So the right hand side is 4 x 4.

You are attempting to store the 4 x 4 into a space that has been declared to be 4 x 1 x (1) . It is not going to fit.

Everything appears to work out if you initialized

K = zeros(4, 4,n); 

[William]

This appears to work, but when I attempt to plot it, everything goes fine until the line for this matrix, shown in red, goes back on itself.

Everthing else about the plot looks good, though.

[Walter Roberson]

Unite is a four by Z matrix

But your code has

 Qhat(:,:,n)=Qhat(:,:,n-1)+K(:,:,n)*(Unite(n)-A*Qhat(:,:,n-1));

which uses Unite as if it is a vector.