try a (or some) quick prototypes of the simplest possible form of the problem. Try and find if it maps to other, solved or proven unsolvable, problems. Split your problem into subproblems, see if those are solvable, if not completely, change course toward a more solvable, maybe similar, problem.