Return vector that minimizes a function

Hello,

I am writing an iterative algorithm that updates a vector "u" each iteration step. I also keep the previous vectors u, so I can control what happened previously, so each iteration there is a new row added to u. This row is the latest vector u if you want. So each iteration I try to update the latest u(k,:) I have by a new u(k+1,:).

The problem is not this iteration, but in each iteration step I find the new u(k+1,:) by minimizing a function "fun". However, this gives an error: "Assignment has more non-singleton rhs dimensions than non-singleton subscripts"

x0 = u(k,:);

fun = @(x)(r - (y0 + T.*x)).^2 + x-u(k,:);

u(k+1,:) = fminsearch(fun, x0);

Here is the full code: %% ILC algoritme clear all; close all; clc;

    % Begin input: assistentie exoskeleton
    u = zeros(1,5); %Begin met zero assistentie of je kan al een goede initiele guess maken eens je het systeem van het OpenSim model kent
    max_it = 100; % Kwestie van niet te lang zoeken
    lambda = 0.5; % Nu is dit nog een gewone constante, later verbeterd!
    y0 = [1 0 -1 3 4];
    y= []; 
    r = 5* ones(1,5);
    e = [];
    Norm = [];
    T= [1 5 3 4 5];

    for k = 1:1:max_it

        y = [y ; y0 + T.*u(k,:)]; % Vervangen door Torque_driven model met u als input voor exoskeleton
        e= [e ; r - y(k,:)];
        %e(k+1,:)= r - (y0 + T.*u(k+1,:));
        Norm = [ Norm ; norm(e(k,:))];

        if Norm(k) < 0.1
            break
        end

        x0 = u(k,:);
        fun = @(x)(r - (y0 + T.*x)).^2 + x-u(k,:);
        u(k+1,:) = fminsearch(fun, x0);

    end

    figure
    xlabel('Time_samples')
    ylabel('Norm error')
    title('Norm error in time')
    plot(Norm)

Answer 1

[Keerthana Chivukula]

The underlying reason for this error message is that the output expected from the objective function provided to "fminsearch" is supposed to be a scalar as mentioned in the following link:

http://www.mathworks.com/help/matlab/ref/fminsearch.html

However, in the code provided above, the function "fun" returned a matrix and this resulted in incorrect indexing. Modifying the function to return a scalar would fix the issue.