Network utility maximisation through particle swarm optimisation

Abstract

Given a network system with several nodes that are linked together in a certain topology, the network utility maximisation (NUM) problem is concerned with searching for the maximum aggregate utility function of all nodes under the constraint of limited link capacities. This paper proposes the use of a gradient-free, stochastic optimisation approach based on particle swarm optimisation (PSO) method for solving the NUM problem. In contrast to much of the prior solution methods which often assume the concavity of the network's utility function, the proposed PSO approach is capable of solving NUM problems with either concave or nonconcave utility functions. We show through simulation results of solving NUM problems in medium size networks that the optimal solutions obtained by the proposed PSO method indeed correspond to the exact global optima.