MOEA/D with Adaptive Weights for Multi-objective Optimization Problems

Multiobjective optimization is always one of the most popular problems in evolutionary computation due to its wide existence in various practical applications. This paper proposes a new improved version of multiobjective evolutionary algorithms based on decomposition (MOEA/D) for solving multiobjective optimization problems, which is named MOEA/D-ADW including three significant strategies. Specifically, firstly, weight vector initialization mechanism is designed to generate high quality weights, which includes UR and WS-transformation strategies. Secondly, an external population is designed for storing the obtained individuals based on the neighborhoods of the weight vector and the search population. Finally, population agents stored in external set are reused for updating the weight adaptively during the evolutionary process. To evaluate the performance of the proposed algorithm, experimental results carried on a set of benchmark functions with various characteristics, demonstrate that the MOEA/D-ADW is competitive with respect to some existing multiobjective optimization algorithms.