Distributed Genetic Algorithm for Integrated Process Planning and Scheduling Based on Multi Agent System

Abstract

Process planning and scheduling are two crucial functions in manufacturing systems which are usually carried out sequentially. The scheduling is totally based on the outcomes of process planning, and the process planning may be restricted by manufacturing resources. Hence, conducting the process planning and scheduling separately is much likely to ruin the feasibility and optimality of both process planning and scheduling functions. The integration of process planning and scheduling (IPPS) is therefore fairly important for an efficient manufacturing system. In this paper, a distributed genetic algorithm (DGA) is suggested to cater for the IPPS problems domains, and the multi-agent system (MAS) is adopted to accommodate the algorithm. Here it is called the MAS-DGA system. Due to good properties of the MAS, a new agent-based architecture is proposed to accommodate subpopulations, support the traditional GA and provide channels for individuals' immigration. Furthermore, an negotiation mechanism is provided to support bilateral selections between individuals and subpopulations. Benchmark problems have been tested in the experiments, and the results are compared with a symbiotic evolutionary algorithm (SEA) and a cooperative co-evolutionary genetic algorithm (CCGA), which reveals that the proposed MAS-DGA system is feasible and efficient for the resolve of IPPS problems.