Comparing Multi-agent System and Mixed-Integer Programming Approaches on a Simple Optimization Problem.

Mathematical programming approaches, like linear programming (LP) or mix-integer programming (MIP) are widely used as an optimization tool, since they can compute the best possible configuration of a constrained system. However, the needed computing resources and time cannot always be figured out in advance, and happen to be beyond available resources in some unexpected cases. Moreover, these approaches do not naturally fit with dynamically changing configurations. As an alternative, Multi-agent systems (MAS) prove to be more adaptive and predictable in terms of needed resources, but of course do not guarantee to provide the best solution, not even always a “good” solution, as far as one can define what a good solution is. In this paper we try to compare the two approaches on a simple optimization (warehouse location) problem.