ode45 must return a column vector?

I keep getting the error:

??? Error using ==> odearguments at 113
CONTROLPAPERQ must return a column vector.

Error in ==> ode45 at 173
[neq, tspan, ntspan, next, t0, tfinal, tdir, y0, f0,
odeArgs, odeFcn, ...

Error in ==> controlPaper at 27
Q1=ode45('controlPaperQ',tB,0.0);

with the code:

%%Start by setting up all of the variables

t=0:.005:.5;

tB=fliplr(t);

save t

save tB

A=10;

B=1;

M1=ones(size(t));

for i=1:50

      M1(i)=0;

end

save M1

M1b=fliplr(M1);

save M1b

%%Now to find Q

Q1=ode45('controlPaperQ',tB,0.0);

save Q1

function Qdot=controlPaperQ(t, history)

load M1

load M1b

Qdot=zeros(1,1);

Qdot=1-20.*history.*M1b-history.^2;

Any ideas?

Answer 1

[Walter Roberson]

consider

t=0:.005:.5;    %creates a row vector
M1=ones(size(t)); %so also creates a row vector
M1b=fliplr(M1);   %continues to be a row vector

And then when we look at

Qdot=1-20.*history.*M1b-history.^2;

if "history" is not a scalar or a row vector, then 20.*history.*M1b would have a dimension error, so we deduce that "history" is a scalar or row vector. And from that we deduce that Qdot is going to be a row vector also.

Unfortunately for you, the function you pass as the first argument to ode45 must return a column vector.

When you are looking over your code trying to figure out what went wrong, please try to figure out why you are ignoring time in your calculation. The "t" you initialize in your code is not the same as the "t" that will be seen inside controlPaperQ : the "t" inside that routine will be the current time of integration.

Answer 2

[Star Strider]

I don't understand what 'history' is or what its dimensions are. I suggest you set a breakpoint at the second 'Qdot = ...' in 'controlPaperQ' to check its dimensions. If it's a row vector it would be easy enough to transpose it to make 'ode45' happy. If it's a matrix, you'll have to decide the appropriate way to fix it.

The documentation for 'ode45' states:

    ‘For a scalar T and a vector Y, ODEFUN(T,Y) must return a column vector 
    corresponding to f(t,y).’

so apparently it's just designed that way.