A Decentralized Scheduling Policy for a Dynamically Reconfigurable Production System

  • Authors:
  • Stefano Giordani;Marin Lujak;Francesco Martinelli

  • Affiliations:
  • Dip. Ingegneria dell'Impresa, University of Rome "Tor Vergata", Italy;Dip. Ingegneria dell'Impresa, University of Rome "Tor Vergata", Italy;Dip. Informatica Sistemi e Produzione, University of Rome "Tor Vergata", Italy

  • Venue:
  • HoloMAS '09 Proceedings of the 4th International Conference on Industrial Applications of Holonic and Multi-Agent Systems: Holonic and Multi-Agent Systems for Manufacturing
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, the static layout of a traditional multi-machine factory producing a set of distinct goods is integrated with a set of mobile production units - robots. The robots dynamically change their work position to increment the product rate of the different typologies of products in respect to the fluctuations of the demands and production costs during a given time horizon. Assuming that the planning time horizon is subdivided into a finite number of time periods, this particularly flexible layout requires the definition and the solution of a complex scheduling problem, involving for each period of the planning time horizon, the determination of the position of the robots, i.e., the assignment to the respective tasks in order to minimize production costs given the product demand rates during the planning time horizon. We propose a decentralized multi-agent system (MAS) scheduling model with as many agents as there are the tasks in the system, plus a resource (robot) owner which assigns the robots to the tasks in each time period on the basis of the requests coming from the competing task agents. The MAS model is coupled with an iterative auction based negotiation protocol to coordinate the agents' decisions. The resource prices are updated using a strategy inspired by the subgradient technique used in the Lagrangian relaxation approach. To measure the effectiveness of the results, the same are evaluated in respect to that of the benchmark centralized model.