Incorporating transportation time in multi-agent production network scheduling

Abstract

The scheduling of factories that work in production network is a new type of scheduling problem that all of the developed single factory techniques are inappropriate for it. The aim of this paper is to propose the scheduling algorithm for such environment in which several factories disperse geographically in different places with parallel machines and each factory as a production agent may have a different objective function. We assume there are two types of production agent, i.e. some factories are interested in the sum of completion times and the remaining factories are interested in the makespan. In such system, a schedule should give enough flexibility to a local scheduler. This can be attained by transporting the jobs among factories from the overloaded machine to the machine which has fewer workloads. By incorporating the transportation assumption in problem definition, we first present a mathematical modelling for the new scheduling problem. We then used CPLEX solver to obtain Pareto solutions by applying ∈-constraint approach. Furthermore, in addition to a genetic algorithm (GA), we proposed a new evolutionary metaheuristic namely imperialist competitive algorithm (ICA) that armed with a new encoding scheme. Finally, the outputs obtained from mathematical algorithm, ICA and GA are reported.

Keyword:

• scheduling
• distributed multi-agent production network
• mathematical modelling
• imperialist competitive algorithm
• Pareto solutions
• transportation time