Author(s): Xu Yang (QMUL), John Bigham (QMUL), Laurie Cuthbert (QMUL)
This paper propose a novel learning approach that applies NeuroEvolution of Augmenting Topology (NEAT) based learning algorithm to resolve Call Admission Control (CAC) combined with resource allocation in adaptive multimedia wireless networks; this not only decides whether to accept or reject a request call, but also determines the allocated bandwidth to that requesting call. The objective is to maximize the network revenue and maintain predefined QoS constraints. The QoS constraints are classified as two categories: long period constraints and instantaneous constraints. Long period constraints are handled by a constraint handling method called Superiority of Feasible Points. Instantaneous constraints and system limitations are handled by an External Supervisor.