Speciated evolutionary algorithm for dynamic constrained optimisation

Dynamic constrained optimisation problems (DCOPs) have specific characteristics that do not exist in dynamic optimisation problems with bounded constraints or without constraints. This poses difficulties for some existing dynamic optimisation strategies. The maintaining/ introducing diversity approaches might become less effective due to the presence of infeasible areas, and thus might not well handle with the switch of global optima between disconnected feasible regions. In this paper, a speciation-based approach was firstly proposed to overcome this, which utilizes deterministic crowding to maintain diversity, assortative mating and local search to promote exploitation, as well as feasibility rules to deal with constraints. The experimental studies demonstrate that the newly proposed method generally outperforms the state-of-theart algorithms on a benchmark set of DCOPs. © Springer International Publishing AG 2016.