A novel multi-objective evolutionary algorithm for shortest path routing problem

Abstract

This paper presents an application of non-dominated sorting genetic algorithm-II (NSGA-II) technique for solving shortest path routing problems in computer networks. The problem is formulated as a non-linear constrained multi-objective optimisation problem. NSGA-II is applied to handle shortest path routing problem as a true multi-objective optimisation problem (MOOP) with competing and non-commensurable objectives. A priority-based encoding scheme is employed for population initialisation. Priorities are assigned to all the edges and NSGA-II is implemented to find the optimal solution. It is noted that this approach can find a diverse set of solutions and is converging near the true Pareto-optimal set. Results for a sample test network have been presented to demonstrate the capabilities of the NSGA-II algorithm to generate well-distributed Pareto-optimal solutions of shortest path routing problem in one single run. The results obtained by NSGA-II are compared with single objective weighting factor method for which genetic algorithm (GA) is applied.