A cluster-based clonal selection algorithm for optimization in dynamic environment

Comparing with stationary optimization problems, dynamic optimization problems pose more serious challenge to traditional optimization algorithms. For solving such problem, the algorithms should track the changing optima persistently. In this paper, a cluster based clonal selection algorithm for global optimization in a dynamic fitness landscape is presented. The population is partitioned into multiple clusters according to the spatial locations at first, and then each cluster is evolved separately by using a learning based clonal selection algorithm, in which, the learning strategy within the cluster and interaction among clusters are introduced to the hypermutation operator to improve search ability. In addition, memory mechanism is presented to deposit the previous searching information and reused for optima tracking after environmental change. Experimental results demonstrate that the proposed algorithm is superior to the immune based algorithms and is competitive with respect to the state of the art designs for dynamic optimization problems.