Improving Multi-agent Evolutionary Techniques with Local Search for Job Shop Scheduling Problem

Abstract

Scheduling is the allocation of shared resources over time in order to perform a number of tasks. Job Shop Scheduling Problem (JSSP) is the most commonly encountered scheduling problem. A wide range of approaches have been proposed to solve it. In this paper two multi-agent based evolutionary models are proposed to tackle JSSP. The first one is Multi-Agent based Genetic Algorithm (MAGA) and the second model is a Multi-Agent Particle Swarm Optimization (MAPSO). A proposed local search technique as self-learning procedure for agents is hybridized with both of the multi-agent models to enhance their efficiency. The proposed models have been implemented using REPAST toolkit. Encouraging results from both models have been obtained for standard benchmarks from OR library.