USE OF NEURAL NETWORKS FOR THE OPTIMUM FREQUENCY ASSIGNMENT PROBLEM

When two satellites in close proximity to each other use the same freq uency band, mutual signal interference is unavoidable. Especially, it is known in the FDMA communication system that the interference to eac h carrier and the total interference depends greatly on the carrier fr equency assignment. At the present stage, where a large number of sate llites have entered into operations, the assignment of the frequencies is becoming ever more important. This paper aims at a high-speed dete rmination of the signal carrier assignment so that the signal interfer ence is minimized and proposes a method of parallel solution using a H opfield-type neural network. In the proposed solution method, the give n frequency band is segmented into n frequency bands and the assignmen t is derived by rearranging the segments using the two-dimensional neu ral network array. The following properties were revealed as a result of simulation. The optimal solution or a solution dose to the optimal can be determined from a tremendous combination of possible solutions, while suppressing the explosive increase of the computational complex ity with the increase of the problem size. The power to derive the sol ution depends on the carrier configuration.