Quasi-static dynamic channel assignment using a Linux PC cluster

In this paper, we propose a quasi-static approach which, using a powerful Linux/PC cluster, combines the merits of both traditional static and dynamic schemes for the channel assignment problem. The static component of our approach uses a parallel genetic algorithm to generate a suite of representative assignments based on a set of different estimated traffic scenarios. At on-line time, the dynamic component observes the actual traffic requirement and retrieve the representative assignment of the closest scenario from the off-line table. The retrieved assignment is then quickly refined by using a fast parallel local search algorithm. Our extensive simulation experiments have indicated that the proposed quasi-static system outperforms other dynamic channel assignment techniques significantly in terms of both blocking probabilities and computational overhead.