In this paper, we study a problem family inspired by a prominent network optimization problem (graph coloring), enriched and extended towards a real-world application (Wi-Fi channel assignment). We propose a utility model based on this scenario, and we generate an extensive set of test cases, against which we run both a complete information optimizer and two nonlinear negotiation approaches -a hill-climber and an approach based on simulated annealing (SA). We show that, for the larger-scale scenarios, the SA negotiation approach significantly outperforms the optimizer while running in roughly one tenth of the computation time. Also, we point out interesting patterns regarding the relative performance of the two approaches depending on the properties of the underlying graphs.
展开▼