A quasi-static cluster-computing approach for dynamic channel assignment in cellular mobile communication systems

Efficient management of the radio spectrum can be accomplished by making use of channel assignment techniques, which work by allocating different channels of the spectrum to the cells of the network in a conflict-free manner (i.e., the co-channel interference is minimized). The problem of dynamically reallocating the channels in response to change in user location patterns, which occurs frequently for a microcell network architecture, is even more difficult to tackle in a timely manner. Most existing approaches use various sequential search-based heuristics which cannot produce high-quality allocation fast enough to cope with the frequent traffic requirement variations. In this paper, we propose a quasi-static approach which combines the merits of both static and dynamic schemes. 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 retrieves 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