Empirical Performance Evaluation of a Parameter-Free GA for JSSP

The job-shop scheduling problem (JSSP) is a well known difficult NP-hard problem. Genetic Algorithms (GAs) for solving the JSSP have been proposed, and they perform well compared with other approaches [1]. However, the tuning of genetic parameters has to be performed by trial and error. To address this problem, Sawai et al. have proposed the Parameter-free GA (PfGA), for which no control parameters for genetic operation need to be set in advance [3]. We proposed an extension of the PfGA, a real-coded PfGA, for JSSP [2], and reported that the GA performed well without tedious parameter-tuning. This paper reports the performance of the GA to a wider range of problem instances. The simulation results show that the GA performs well for many problem instances, and the performance can be improved greatly by increasing the number of subpopulations in the parallel distributed version.