Genetic Programming with Adaptive Reference Points for Pareto Local Search in Many-Objective Job Shop Scheduling

Genetic Programming (GP) is a well-known technique for generating dispatching rules for scheduling problems. A simple and cost-effective local search technique for many-objective combinatorial optimization problems is Pareto Local Search (PLS). With some success, researchers have looked at how PLS can be applied to many-objective evolutionary algorithms (MOEAs). Many MOEAs'performance can be considerably enhanced by combining local and global searches. Despite initial success, PLS's practical application in GP still needs to be improved. The PLS is employed in the literature that uniformly distributes reference points. It is essential to maintain solution diversity when using evolutionary algorithms to solve many-objective optimization problems with disconnected and irregular Pareto-fronts. This study aims to improve the quality of developed dispatching rules for many-objective Job Shop Scheduling (JSS) by combining GP with PLS and adaptive reference point approaches. In this research, we propose a new GP-PLS-II-A (adaptive) method that verifies the hypothesis that PLS's fitness-based solution selection mechanism can increase the probability of finding extremely effective dispatching rules for many-objective JSS. The effectiveness of our new algorithm is assessed by comparing GP-PLS-II-A to the many-objective JSS algorithms that used PLS. The experimental findings show that the proposed method outperforms the four compared algorithms because of the effective use of local search strategies with adaptive reference points.