MOEMT-RPA: Reference Point-Assisted Explicit Multi-Objective Evolutionary Multi-Task Algorithm

The evolutionary multi-task optimization algorithm (EMT) enhances the algorithm's performance through information interaction. However, few studies can be found in the current literature utilizing reference points for knowledge transfer. This paper proposes a reference point-assisted explicit multi-objective evolutionary multi-task algorithm (MOEMT-RPA), in which the reference point method is utilized to measure the similarity of optimal solutions. The similarity of tasks is reflected in the number of repetitions of the reference point associated with optimal individuals in different tasks. Compared with advanced comparison algorithms based on the indicators of IGD value and HV value, the experiment on CEC2017 evolutionary multi-task optimization competition benchmarks verified the performance of MOEMT-RPA.