A modified Hopfield neural network algorithm for cellular radio channel assignment

Since the frequency spectrum of the mobile radio communications is limited, the channel assignment problem deserves more attention in order to use the available frequency spectrum with optimum efficiency. A new channel assignment algorithm using a modified Hopfield neural network was proposed by Kim and Nasrabadi (see IEEE Trans. on Vehicular Technology, vol.46, no.4, p.957-67, 1997). In this paper, we propose various initialization techniques based on multilevel rearrangement of the channels before applying the algorithm of Kim et al. to decrease the number of iteration and improve the convergence rate. These techniques will guarantee that the neural network will skip the local minimum, and in all cases will converge to optimum arrangement of the channels. The specific characteristics of the channel assignment problem in cellular radio network such as co-site constraints, adjacent channel constraints, and co-channel constraints are considered with the implementation of the preassignment techniques. The results of the proposed techniques are compared with other prior reported techniques for the same eight benchmark problems. The comparison shows the merits of the proposed initialization techniques.