Anytime frequency allocation with soft constraints

As is well known, the frequency allocation problem for mobile telephone networks can be approximated by a graph coloring problem where vertices model transmitters, colors model frequencies, and an edge corresponds to a pair of transmitters that must not use the same frequency. However, the graph coloring analogy doesn't suuce for mod-eling real world problems, which involve soft constraints and general distance constraints. Thus, general constraint satisfaction techniques are called for. In this paper, we model the frequency allocation problem as a partial constraint satisfaction problem, and give a branch & bound algorithm for solving it. We review a number of constraint solving techniques used in this problem and discuss their relevance to performance.