Sum of unequal sized matrices

I have matrix A of fixed size 100x2, and B is a matrix growing in size (row-wise) from 1 to 300 rows. The distance criteria is to be used to find nearest neighbour for every row of A w.r.t. all rows in B. They will be stored as indexes a and b.

The modified A is calculated as A=A+B. Please demonstrate through code, consider A=rand(100,2). B gets formed as max row sum of C=rand(5,2) in every iteration from 1 to 300. I am confused of what to and how to use permute /reshape / repmat

[Image Analyst]

Please give, for some low iteration, an example of A, B, and C (use small sizes just to demonstrate). I don't even know how many columns B has or what you consider the "distance" to be. And I don't know if you want summing, or vertical concatenation. An example would surely help. Maybe you want pdist2() to compute distances - I don't know, there are still clouds in my crystal ball.

[Guillaume]

mukund comment mistakenly posted as an answer moved here:

It is for an idea, may not be accurate. Main iteration

   for itr=1:1000  %Main loop
    A=rand(100,2) %Fix sized matrix
    for i=1:300
    C=rand(5,2)   %random matrix
    [~,cid]=max(sum(C)) %row wise sum
    B(i,:)=C(cid,:)
    end
    for j=1:100
    for k=1:size(B,1)
    dist(j,k)=sqrt(A(j,:)-B(k,:).^2) % dist of each row in A to each row in B
    end
    [st,a]=min(dist) %min distance
    st(j)=st   
    b(j)=B(a,:) % index of nearest row in B
    end
    A=A+B        %this step is problematic for me
    end

The main aim is to resize B to match A for addition in case its size is less, equal or more.