Experiments with a Plateau-Rich Solution Space

This paper will report results of various different metaheuristic search techniques applied to a complex student allocation problem. The problem has what might be considered a particularly difficult solution space when using the most natural definition of neighbourhood structure. The difficulty concerns the presence of many ”plateaux” large areas of solution space where all solutions have the same overall cost bordered by discrete steps up or down. The challenge for the neighbourhood search mechanism is to avoid becoming ”stuck” on a plateau by finding routes across these plateaux that will eventually lead to a step down, i.e. a reduction in overall cost.