A5 - Modellierung und Analyse der Dynamik selbststeuernder logistischer Prozesse

Motivation

We consider a large logistic network as a large number of autonomous objects with decentralized control. The local behaviour of small parts of such system is due interactions between single objects and is called micro-dynamics. A proper design of local controls may lead to self-organising structures of the whole system. Such structures appear in natural systems, e.g., insect colonies, where the behaviour of single insects helps to reach the global gaol of the colony. Dynamic behaviour of the system as a whole is called macro-dynamics. The motivating questions of the subproject are: How to reach a similar effectiveness and robustness for large scale logistic systems? How to design the local interactions to reach this goal?

Results Phase 1 (2004-2007)

A reference scenario for a production logistic shop floor model (MxN Matrix-Model) has been set up and several approaches for modelling and analysing were introduced: Discrete-event simulation models, System Dynamics models and analytical models in form of differential equations. Two different autonomous control methods, a queue length based and a pheromone based decision rule, where parts decide about their next production steps on the basis of local information only, have been implemented and analysed by doing multiple simulation experiments. It has been demonstrated that models with autonomous control methods are more effective in situations of high fluctuating demand and are more robust against unexpected disturbances than models which are controlled by conventional pre-planned schedules. Furthermore, it has been shown that...

Objectives Phase 2 (2008-2011)

The main objective of the second phase is modelling and analysis of the dynamics of production networks with autonomous control. Such networks contain several production plants and a transport network. Hence we deal with a large system of production processes and transport processes with autonomous objects. We focus on the dynamics of these processes which depends on local interactions in each plant and also between plants via transport network and information flows. The question is how to design appropriate control of participating objects to reach an effective and robust behaviour of the whole systems. The influence of the micro-dynamics on macro-behaviour will be studied. The interaction between different types of logistic processes will be investigated for this purpose.

Approach Phase 2 (2008-2011)

The interdisciplinary composition of the subproject allows usage of two different approaches simultaneously. One is mathematical modelling based on analytic methods, another is a software-based discrete event modelling. Both approaches have different advantages and will be used to compare results and to validate each other. Discrete event simulations can better depict the micro-dynamics, the analytic models are more helpful for the modelling of the global behaviour. However both modelling approaches can be used for different modelling (micro and macro) levels. It is also planed to extend the "Distributed Logistics Routing Protocol" developed for transport logistics processes in the first phase by the subproject B1 to the processes in production networks.