Autonomous Units as a Rule-Based Concept for the Modeling of Autonomous and Cooperating Processes

In this thesis autonomous units are presented as a concept to model autonomous processes. Autonomous units form a community with a common environment, in which they act and which they transform. They are based on rules, the applications of which yield changes in the environment. They are also equipped with an individual goal which they try to accomplish by applying their rules. A control condition enables autonomous units at any time and in any situation to select the rule that is actually applied from the set of all applicable rules. The formal semantics of a community as a whole and of each of its members is defined in two stages. In the sequential case only one unit can act at a time and the rule application of the involved units are interl eaved with each other. In order to illustrate the sequential case, the formal concept of Petri nets is modeled by a community of autonomous units. Here every transition of the Petrinet is realized as one autonomous unit. In the parallel case a number of actions take place in parallel at the same time. As an example, a colony of ants with a very simple foraging strategy is presented. In this case the parallel actions still occur in sequential order, so some preliminary ideas of a third stage are given. In this concurrent semantics, the autonomous units may act independently without chronological relations between them, unless a causal relationship demands a certain order of actions.As further illustration, communities of autonomous units are applied to the domain of transport logistics. A transport network is modeled which consists of depots and their connections, unit loads, and trucks. The load units have to be transported from a source depot to a target depot by trucks. Here the trucks as well as the load units are modeled as autonomous units. How the unit loads will actually be transported by the trucks results from negotiations between all involved entities. Two case studies that have actually been implemented using the graph transformation tool grgen are presented in detail. The first case study deals with a model of the board game Ludo and the sequential process semantics of the corresponding community. The second case study deals with a model of a foraging ant colony and the parallel process semantics of the corresponding community. Some fundamental aspects of the semantics of rule-based systems in relation to the semantics of visual models are discussed, which form the conceptional background of this thesis. Since control conditions are an essential part of the modeling with autonomous units, their efficient handling is the main challenge regarding the creation of a software tool. So some seemingly simple control conditions are investigated with respect to implementation.