How to Normalize the gradient matrices

I have a matrix F and initial[x1 y1] and final position[x2 y2]. I have to update the position [x y] using gradient of matrix F to get the desired position. My approach is like this:

[grad_x grad_y] = gradient(F);

[x_new y_new] = [x1 y1]+(step_size)*[grad_x(y1,x1), grad_y(y1,x1)];

I am taking step size 1. but when I tries to normalize the gradient matrices [grad_x grad_y] using formula norm_grad_x = grad_x/norm(grad_x), the values becomes very small and it is hard to update the position on every iteration. [x_new y_new] = [x1 y1]+(step_size)*[norm_grad_x(y1,x1), norm_grad_y(y1,x1)];

Can someone please guide me where I am having a mistake?