If you don't need to compile your code Matlab is very flexible with array sizes. For example if you have A = zeros([2 2]); you can then do A(2, 3) = 5; and Matlab will resize the array. You can then remove that extra column by A(2, 3) = [];
It's easier (for me anyways) to think of the operation in terms of a 2D array and then understand that the operation will perform the same way on an ND array.
You could:
B = zeros([r m n A(1)]);
And then if you assign something to
B(r, m , n, A(2)) = val;
Matlab will automatically resize the array. Using those commands you could both grow and shrink your array as needed for each timestep. Does this answer the question?
If you don't want to lose the data from B for each timestep you could make B a cell array of numeric arrays of different sizes such that:
for iter = 1:length(A)
B{i} = zeros([r m n A(i)]);
end
Then at t = 1;
B{1}(1, 1, 2, 1) = val1 + val2;
or whatever you need;
