A new dynamic strategy for dynamic multi-objective optimization

Abstract

After detecting the change of the environment, it is effective to respond to the change of the environment. However, the majorities of these methods only respond to the change of the environment once, ignoring the use of new more environment information. In this paper, we propose a new algorithm for dynamic multi-objective optimization by combining the evolutionary algorithm and the dynamic strategy. The dynamic strategy consists of two parts which correspond to two responses to the environmental change: restart strategy (RS) and adjustment strategy (AS). RS is to use a small amount of the new environment information and local search to re-initialize the population which is expected to be close to the Pareto solutions in the new environment after the environment change. RS is beneficial for quickly responding to environment changes. AS is to adjust the current population with high quality solutions after getting more accurate environmental information. RS is expected to accelerate the convergence speed of the algorithm. The proposed algorithm is tested on a variety of test instances with different changing dynamics. Experimental results show that the proposed algorithm is very competitive for dynamic multi-objective optimization in comparison with state of-the-art methods.