Simulation-Based Fitness Landscape Analysis and Optimisation of Complex Problems

Abstract

Widespread hard optimisation problems in economics and logistics are characterised by large dimensions, uncertainty and nonlinearity and require more powerful methods of stochastic optimisation that traditional ones. Simulation optimisation is a powerful tool for solving these problems. Moreover, fitness landscape analysis techniques provide an efficient approach to better selection of a suitable optimisation algorithm. The concept and techniques of fitness landscape analysis are described. A formalised scheme for simulation optimisation enhanced with fitness landscape analysis is given. Benchmark fitness landscape analysis is performed to find relations between efficiency of an optimisation algorithm and structural features of a fitness landscape. Case study in simulation optimisation of vehicle routing and scheduling is described. Various optimisation scenarios with application of the fitness landscape analysis are discussed and investigated.