Genetic Programming with Adaptive Search Based on the Frequency of Features for Dynamic Flexible Job Shop Scheduling

Abstract

Dynamic flexible job shop scheduling (DFJSS) is a very valuable practical application problem that can be applied in many fields such as cloud computing and manufacturing. In DFJSS, machine assignment and operation sequencing decisions need to be made simultaneously in dynamic environments with unpredicted events such as new job arrivals. Scheduling heuristic is an ideal candidate for solving the DFJSS problem due to its efficiency and simplicity. Genetic programming (GP) has been successfully applied to evolve scheduling heuristics for job shop scheduling automatically. However, GP has a huge search space, and the traditional search algorithms do not utilise effectively the information obtained from the evolutionary process. This paper proposes a new method to make better use of the information during the evolutionary process of GP to further enhance the ability of GP. To be specific, this paper proposes two adaptive search strategies based on the frequency of features in promising individuals to guide GP to evolve effective rules. This paper examines the proposed algorithm on six different DFJSS scenarios. The results show that the proposed GP with adaptive search can converge faster and achieve significantly better performance than the GP without adaptive search in most scenarios while no worse in all other scenarios without increasing the computational cost.