A Hybrid Multi-Objective Teaching Learning-Based Optimization Using Reference Points and R2 Indicator

Abstract

Hybrid multi-objective evolutionary algorithms have recently become a hot topic in the domain of metaheuristics. Introducing new algorithms that inherit other algorithms’ operators and structures can improve the performance of the algorithm. Here, we proposed a hybrid multi-objective algorithm based on the operators of the genetic algorithm (GA) and teaching learning-based optimization (TLBO) and the structures of reference point-based (from NSGA-III) and R2 indicators methods. The new algorithm (R2-HMTLBO) improves diversity and convergence by using NSGA-III and R2-based TLBO, respectively. Also, to enhance the algorithm performance, an elite archive is proposed. The proposed multi-objective algorithm is evaluated on 19 benchmark test problems and compared to four state-of-the-art algorithms. IGD metric is applied for comparison, and the results reveal that the proposed R2-HMTLBO outperforms MOEA/D, MOMBI-II, and MOEA/IGD-NS significantly in 16/19 tests, 14/19 tests and 13/19 tests, respectively. Furthermore, R2-HMTLBO obtained considerably better results compared to all other algorithms in 4 test problems, although it does not outperform NSGA-III on a number of tests.