Production networks can be dynamically structured and involving multiple production sites with different objectives. This organizational structure is able to match agility and efficiency to compete in the global market. In this environment is impossible for a single organization to control whole production networks; thus, a decentralized approach has been developed to manage the production networks. However, the coordinate mechanism in decentralized control is more important to obtain a high level of performance. The research proposes innovative coordination strategies for coordinate production networks by Multi Agent Architecture. A link between negotiation strategies and a production planning algorithm has been developed in order to support the coordination strategies proposed. In particular, two protocols to reach an agreement between customer and the production network have been proposed: negotiation and an expected profit approaches. Moreover, two coordination strategies have been proposed: index efficiency and ranking price approaches. Finally, the possibility of divide the orders in lots by the customer is proposed. A simulation environment based on open source code and Multi Agent Architecture has been developed to test the proposed approaches. The experiments have been conducted in different conditions of workload and mar-up; the results of the simulation provide the information necessary to select the suitable coordination and protocol mechanisms in a distributed production planning problem.
TopIntroduction
The market globalization and increasing of competition forces the manufacturing companies to adopt distributed production approach. Moreover, Small and Medium Enterprises can compete to a global level if they collaborate in production networks. The distributed approach of production planning requires coordination mechanism in order to obtain a high level of performance (Wiendahl and Lutz, 2002). Advanced Planning and Scheduling (APS) tools are considered by many as the state of the art manufacturing and scheduling practices. These tools are designed to support a centralized production, therefore multi-facility of a single company. The application of APS tools become very complex and difficult in distributed production system, especially when the unit are independent (Stadtler, 2005).
Decentralized production planning approaches lead to several advantages in managing production networks: i) data management is more suitable for independent unit; ii) production planning system is robust, scalable, extensible and easier to adapt when the strategies change; iii) the production planning problem is distributed and therefore more easier to solve; iv) production planning systems of the units are more easier to integrate.
However, distributed production planning approaches have some drawbacks: i) the performance of distributed approaches are lower than the centralized approaches; ii) the performance of the coordination approaches are difficult to foresee; iii) a third independent part is necessary to manage the network in order to avoid opportunistic behavior.
This research concerns the production planning problem in a production network characterized by independent unit, therefore without sharing of information among the plants involved. Small and Medium Enterprises can gain competitive advantages to participate in this kind of network.
Multi Agent Systems (MAS) is the appropriate framework for developing distributed applications, and this is particularly true in Distributed Production Planning problems (Swaminathan et al., 1996; Parunak andVanderbok, 1998). A MAS is a loosely coupled network of software agents that interact to solve problems that are beyond the individual capacities or knowledge of each problem solver (http://www.cs.cmu.edu/~softagents/). A MAS needs of proper coordination mechanism in order to guarantee goals achievement.
The main objectives of this research are:
- •
The development of the Multi Agent Architecture framework in static and dynamic views able to support the distributed production problem.
- •
Two coordination mechanisms have been proposed: the first mechanism is performed by an efficiency index computed on the proposal characteristics submitted by the plants. The second is based only on the evaluation of the price submitted by the plants.
- •
Two protocols to reach an agreement between the customer and the network have been proposed: a negotiation approach and an approach performed in a single step defined “expected profit”.
- •
The proposal computed by the plants is obtained by a link with the local production planning algorithm that provides a set of production planning alternatives.
- •
A simulation environment is developed in order to test the proposed approaches and evaluate the performance.
The chapter is structured as follows: section 2 reviews the literature in the multi-plant production planning context; section 3 describes the framework of the Multi Agent architecture; section 4 introduces the production planning model of the plants; in section 5 the two strategies of the plants is presented, while the coordination strategies are described in section. Section 7 explains the simulation environment developed and in section 8 the simulation results are discussed. Finally conclusions and further researches in this area are drawn in section 9.
TopLiterature Review
A multi facility production planning problem can be formulated as follows: given an external demand for items over a time horizon and a set of facilities able to produce those items, it needs to find a production planning over multiple facilities that maximizes customer and company satisfaction. This problem is subject to two categories of constraints: