On the performance of distributed neural network-based DCA schemes for packet-switched mobile communication systems

Emerging mobile communication systems will be required to manage the scarce spectrum more efficiently taking account of both temporal and spatial traffic variations. Moreover, these systems are expected to rely on packet-switched technology for provision of various services. Hence, new resource allocation schemes, capable of operating within a more demanding environment will be required. Centralised and distributed dynamic channel allocation (DCA) schemes based on Hopfield neural network (HNN) have been reported to perform favourably with other competing schemes with a potential of providing optimum channel assignment in a short period of time. The studies conducted have however mainly concentrated on performance provided within circuit-switched systems. This paper takes the analysis of distributed HNN-based DCA schemes a step further by means of evaluating the system capacity provided by these algorithms within packet-switched systems. The paper evaluates the system performance in terms of packet clipping and considers the fairness of the allocation strategies among users in different cells.