Combining Multi–Agent Systems and Constraint Techniques in Production Logistics

The paper describes objectives and results of a research project conducted by Daimler–Benz AG. Intention of this project is decision support for a logistics division. In particular, the production flow control (PFC) in a factory has to be optimized by a decision support system. PFC is a sophisticated process influenced by complex factory structures and high dynamism of manufacturing processes. In order to support PFC we developed a system based on concepts of multi–agent systems and constraint techniques. With this system it is possible to simulate arbitrary production scenarios and to compute their effects on necessary production settings. The user thus receives decision support which may be used during operational planning in logistics as well as in development of new logistic strategies. 1.Application Domain and Problematic Nature The production flow control (PFC) department is part of a logistics division responsible for superordinate planning and coordination of the manufacturing process at a particular plant of Mercedes–Benz. PFC planning results in a production strategy. This strategy encompasses quantitative quotas for every manufacturing center in the factory to be entered in a table called daily production plan (DPP). Additionally, corresponding production settings have to be determined for the manufacturing centers. The general aim of production strategy is to achieve a production goal, that is a given number of products. PFC is a complex and sophisticated process influenced by a number of factors. Those have strong consequences on the manufacturing process and in so far on the logistic planning itself (see table 1 for an overview). Various factors and their effects on production and planning processes are described below: number of manufacturing lines in the factory: a product can be manufactured in several lines which are closely interconnected (see figure 1). This structure results in a multitude of alternatives during production planning and controlling. In addition, the number of alternatives is restricted by technical reasons and organizational requirements. number of manufacturing steps: the entire manufacturing process is divided into distinct steps. Each step can be performed at several manufacturing centers which are organizationally separated (cost centers). The installation of cost centers supports cost–optimal manufacturing on a local basis. However, decentralized planning and controlling leads to suboptimal global conditions since no sufficient global coordination is provided. manufacturing of different series variants, and individuality of every particular order due to a wealth of optional equippment: each product is at last an individual piece. Another aggravating effect towards the planning process within the logistics department is caused by the dynamic character of the manufacturing process. First, dynamism is yielded by a series of rebuildings at the present and in the future. Permanent reconstructions lead to serious alterations in factory structures. Due to this, alternatives for formulating a production strategy keep changing. Planning and controlling processes in PFC have to be adapted and rearranged. Second, it is typical for the manufacturing process to be disrupted by varied technical disturbances. Proneness to disruptions results in frequent plan deviations, requiring short– term plan modifications. This kind of dynamism results in high planning uncertainty and lack of time to formulate new production strategies. manufacturing center 2 types: C, D manufacturing center 1 types: A, B manufacturing center 3 types: A, D center 5 all types manufacturing center 9 types: A, D manufacturing center 7