Multiple objective dynamic programming with forward filtering (Q1821702)

We present a heuristic method for reducing burden in multiple objective dynamic programming (MODP). Using techniques originally suggested for multiple objective linear programming, the solution set for each state (stage) are filtered, giving a representative subset of the set of efficient ways of attaining that state (stage). The method allows for considerable reductions in the amount of storage required to solve the problem, and in the dimensionality of the problem in solution space. It does not guarantee that all the identified solutions are non-dominated; however, the examples presented suggest that the representation of the subset of all efficient solutions is a good one.