Dynamic clonal selection algorithm solving constrained multi-objective problems in dynamic environments

This work investigates a dynamic clonal selection algorithm suitable for time-varying nonlinear multi-objective problems with inequality constraints. In one such algorithm, several adaptive operators such as environmental detection, dynamic reproduction, adaptive mutation and reconstruction are designed specially. When the environment changes, the environmental detection operator related to the history information is first executed to generate an initial population helpful for rapidly capturing the time-varying Pareto front. Within a run period, the current population is divided into multiple layers as associated to the weak Pareto optimality concept. After so, different layers are required to evolve along different directions, relying upon their importance. The preliminary experiments through comparison demonstrate that the proposed algorithm can track adaptively the changing environment and also approximate rapidly the desired Pareto front for a given environment.