The optimization techniques such as Genetic algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Simulated Annealing (SA), etc. , were commonly used in solving job shop scheduling problem (JSSP). There are different variants of these algorithms that were addressed in several previous works. In previous literatures, it was commonly mentioned that the initial solution were generally guessed in a very random manner (such as random initialization of population in GA). In this work, we will address the impact of such random initialization on solving the JSSP while using an optimization technique - GA. The performance of this algorithm will be evaluated with different set of initial conditions. In one experiment, during initialization stage, the initial population will be initialized with random schedules. In another experiment, the initial population will be initialized with a known, worst case schedule. The impact of this initial condition on the performance of algorithm has been studied and achieved makespan. The arrived results proved that the conventional way of randomly selecting initial conditions of the evolutionary process has a worst effect on performance in JSSP of higher dimensions. While initializing with known, worst case solution, the evolutionary process was capable of converging into meaningful and more optimum solutions.

Scheduling Job Shop Scheduling Genetic Algorithm Gant-chart