Modified Adaptive Evolutionary Algorithm for Solving JSSP Problems

A job-shop scheduling problem is one of the classic scheduling problems considered to be NP-hard. In this paper, we presenta modified adaptiveevolutionary algorithm (EA) that uses speculative mutations, variable fitness functions and a pseudo-random number generator for solving job-shop scheduling problems. The algorithm was tested on well-known benchmark datainstances, such as Ft10, La01, Swv01, etc., with the goal of achieving the shortest make-span. The results show that using speculative mutations and interval placing reduces the number of steps and computational time to achieve a (near) optimal make-span. Some testing results on an early version of the proposed algorithm are also added,whichwere used to define the most effective types of mutations to generate better offspring. Key-words: evolutionary algorithm, scheduling, job shop, variable fitness function, speculative mutations