Practical channel assignment using neural networks

Abstract

The channel assignment problem (CAP) i.e. the task of assigning the channels to the radio base stations, is an NP-complete optimization problem occurring during design of cellular radio systems. An investigation is carried out to see how far a parallel neural network algorithm may be used for a real-world CAP. The neural network algorithm identified a channel assignment without heuristic rules generating the sequence of channel assignments. Formulating the optimization problems in terms of an energy function turned out to be very flexible, making it easy to include a variety of additional technical constraints and optimization criteria. Since a distinction between hard constraints and soft optimization criteria is only expressed by the relative strength of the terms in the energy function, the user is free to choose this relative strength according to his actual planning situation. Therefore, this neural network algorithm is well suited for the solution of real-world channel assignment problems.<>