A novel grid-based bidirectional local search algorithm for many-objective optimization
Qi Bing Jin, Yun Tao Li, Wu Cai · 2018
The balance between the convergence and diversity is the most important mission in evolutionary multiobjective optimization (EMO). Many studies pay attention to the convergence but neglect the diversity maintenance. In order to accomplish these two goals simultaneously, the proposed grid-based bidirectional local search algorithm (GrBLS) is capable of finding a set of solutions which have better convergence to the real Pareto optimal front as well as better distribution. The GrBLS uses grid-based framework to select the better individual in evolution generation. Through the different punishment degree in environmental selection process, the best solutions can be gotten from the previous population. Three state-of-the-art EMO algorithms are chosen to compare with GrBLS method by employing four classical performance indicators. The experimental results show that GrBLS method is able to obtain better spread of solutions and better convergence than the others algorithms. The results indicate that the excellent performance of the GrBLS method is suitable for handling multiobjective optimization problems.