How to fix it?
You have fewer data points than you have unknowns. So at the very best, in general there will be infinitely many solutions. NO unique solution to the problem.
There are TWO choices of algorithm in lsqnonlin. The trust-region-reflective algorithm supplied does not handle that class of problem, i.e., with too few data points to have a unique solution.
The alternative algorithm is the levenberg-marquardt algorithm. It does not allow the use of bound constraints on a problem.
So how to fix it? Do one of...
1. Get more data
2. Drop the bound constraints
3. Find a different tool than lsqnonlin to solve the problem
Again, recall that the solution will generally not be unique, no matter what tool you do find. There may be zero solutions given the bound constraints, exactly one solution (very rarely), or infinitely many of them.
Best Answer