International Workshop on Rule-Based Modeling and Simulation of Interacting Systems and Agents

15-21 February 2006, Cottbus, Germany

Organized by the REWERSE Working Group I1 on Rule Modeling and Markup and the Chair of Internet Technology, Brandenburg University of Technology at Cottbus

rule exampleRules can be used both in descriptive models for describing observable interaction patterns and in prescriptive design models for specifying the reactive behavior of individual systems and agents. They allow to connect events and internal states with actions and internal state effects in a natural and declarative way. Since they are also executable at the model level, they allow high-level simulations of interacting systems.

The main goal of this workshop is to discuss the state of the art of rule-based modeling and simulation with special consideration of Agent-Object-Relationship (AOR) modeling and simulation.

Location

Building: LG 1C, room 415. Address: Chair of Internet Technology, Institute of Informatics, Brandenburg University of Technology, Walther-Pauer-Str. 3, 03046 Cottbus, Germany

Preliminary Program

Wednesday, 15 February 2006

Time Presenter(s) Title
14:30 - 14:35 Gerd Wagner (Brandenburg University of Technology, Cottbus, DE) Opening Remarks
14:35 - 15:45 Giancarlo Guizzardi (Institute of Cognitive Science and Technology, Trento, IT) Ontologies and Conceptual Modeling
15:45 - 16:00 Coffee/Tea Break
16:00 - 17:00 Renata Guizzardi & Giancarlo Guizzardi The Unified Foundational Ontology with Applications to Agent-Oriented Modeling

Thursday, 16 February 2006

Time Presenter(s) Title
10:00 - 11:00 Gerd Wagner (Brandenburg University of Technology, Cottbus, DE) UML-Based Modeling of Rules
11:00 - 11:15 Coffee/Tea Break
11:15 - 12:15 Wolf-Ulrich Raffel (Berlin, DE) A Java-Based Simulator for Agent-Object-Relationship Simulation (abstract)
12:15 - 14:00 Lunch Break
14:00 - 15:00 Michael Wosko (Brandenburg University of Technology, Cottbus, DE) The Beer Game as a Simple Example of an Agent-Based Management Simulation (abstract)
15:00 - 15:15 Coffee/Tea Break
15:15 - ?? Work Sessions

Friday, 17 February 2006 (Location: Presentation room IKMZ, 7th floor)

Time Presenter(s) Title
10:00 - 11:00 Ioannis Athanasiadis (Istituto Dalle Molle di Studi sull'Intelligenza Artificiale, Manno, CH) Modelling Software Agents in Environmental Management Applications (abstract)
)11:00 - 11:15 Coffee/Tea Break
11:15 - 12:15 Renata Guizzardi (University of Twente, NL) ARKnowD: An Integrated Agent-oriented Methodology to Develop Knowledge Management Systems (abstract)
12:15 - 14:00 Lunch Break
14:00 - ?? Work Sessions

Monday, 20 February 2006

Time Presenter(s) Title
11:00 - 12:00 Kuldar Taveter (University of Melbourne, AU) What Kinds of Models are Needed for Agent-Oriented Modelling and Simulation? (abstract)
12:00 - 13:30 Lunch Break
13:30 - 14:30 Jens Dietrich (Massey University, NZ) The Mandarax Framework (abstract)
14:30 - 14:45 Coffee/Tea Break
14:45 - ?? Work Sessions

Tuesday, 21 February 2006

Time Presenter(s) Title
11:00 - 12:00 Sergey Lukichev (Brandenburg University of Technology, Cottbus, DE) A UML Rule Modeling Tool
12:00 - 13:30 Lunch Break
13:30 - 14:30 Jens Dietrich (Massey University, NZ) Towards ECA Server Pages (abstract)
14:30 - 14:45 Coffee/Tea Break
14:45 - ?? Work Sessions

Abstracts

Renata Guizzardi: ARKnowD - An Integrated Agent-oriented Methodology to Develop Knowledge Management Systems

Agents are suitable abstractions for modeling KM contexts, due to their cognitive characteristics, such as goals, beliefs, and reactivity. However, besides a good abstraction, the development of adequate solutions requires a consistent software engineering methodology. To help fill in this gap, we describe some results of our work on an agent-oriented methodology to develop KM system. Our approach is based on the integration of existing modeling methodologies and languages, namely the Tropos methodology for the analysis activity, and AORML aiming at the system design. Our model considers all humans, organizations and existing systems as agents. This enables the designer to understand and describe relations between entities before system development. Furthermore, different cognitive characteristics of agents are used in different phases of the development cycle. We pay special attention in understanding a) which of the characteristics are more appropriate for each development activity, and b) how to materialize such characteristics as concrete elements of information systems.

Ioannis Athanasiadis: Modelling Software Agents in Environmental Management Applications

The objective of this talk would be to present from a practical point of view the utilization of AORML in modelling software agent architectures in environmental management applications. Examples from building real-world applications in air quality assessment, meteorology, and water resource management will be presented in this talk. Focus will be given also on our experiences in sharing AORML diagrams with end-users (typically environmental scientist), who seam to be able to comprehand them quite easily.

Wolf-Ulrich Raffel: A Java-Based Simulator for Agent-Object-Relationship Simulation

We present a kind of agent-based simulation being a refinement of discrete-event simulation. It has a formal semantic for execution, which bases on object-based simulation, which itself is basing on discrete-event simulation. Our agent-based simulation contains distribution of the simulated system in active entities (agents) and passive entities (objects). Agents have an external (physical) as well as an internal (mental) state. In simulation there is an environment simulator, managing the environment the states of objects and the external states of agents. We also have an agent simulator for each agent, managing the internal agent state. Simulation cycles, whereby the environmental simulator computes the events for the current cycle, does changes in the environment and sends perceptions to each agent simulator. Agent simulators compute internal events for the current cycle, do changes to the internal state and send taken actions to the environmental simulator. We present a UML-based specification language to specify a concrete agent-based simulation system visually. This language uses reaction rules to specify the behaviour of simulation. We present an XML-based specification language called ABSimML as well. As proof of concept we implemented a simulation system, whose kernel is a Java program library. We used this system to simulate a complex system of Automatically Guided Vehicles with decentral control and to do studies on the efficiency of this system. Having a simulation system, which is visually modelled using the UML-based simulation language, one can generate a simulation model in the XML-based language XMI. Using XSL transformation this model can be transformed into a simulation model in the language ABSimML. This model can be transformed into Java code, which can be used by the simulator. So, we developed a system to transform a visually specified simulation model into a simulation program, which can be executed by the simulator.

Michael Wosko: The Beer Game as a Simple Example of an Agent-Based Management Simulation

This presentation deals with a synthesis of theoretical and conceptual work on a MAS-model of the MIT Beer Game, known from the paper [LTW03], with the implementation of a general simulation framework, called Absimulation, given in [Raf05]. Being this for the author the first approach to MAS applied, the goal of this study was to make the author acquainted with MAS modeling and implementation techniques, and with issues that arise within such processes. Though comparatively much work is associated here with implementation issues in the concrete case of the Beer Game, it was also attempted to discuss general problems, whenever the implementation (and also its limitations) offered such an opportunity. For sake of completeness, concepts related to the Game, like supply chains and their management are summarized shortly. A wider view of management and simulation games is offered in form of a comparative survey, aimed at finding a list of features which are offered by all or most of them, or are just repeatedly mentioned as needed, which pretty much define, what a general model of such a game should provide. REFERENCES: [Raf05] Raffel, W.-U. : “Agentenbasierte Simulation als Verfeinerung der Diskreten-Ereignis-Simulation unter besonderer Berücksichtigung des Beispiels Fahrerloser Transportsysteme”. [LTW03] Luin, Tulba, Wagner : “Remodeling the Beer Game as an Agent-Object-Relationship Simulation”.

Kuldar Taveter: What Kinds of Models are Needed for Agent-Oriented Modelling and Simulation?

The RAP/AOR methodology has been and is being used for modelling and simulation of different real-life domains, such as B2B e-commerce, ceramic factory and "air-to-air" business processes. The methodology is based on the RAP/AOR viewpoint modelling framework which promotes modelling from different perspectives and at different levels of abstraction. In my presentation, I will first describe the RAP/AOR viewpoint modelling framework. After that, I will propose one way of filling the cells of the framework with different kinds of models. Finally, I will make an attempt to analyse the roles played by different kinds of models in capturing the problem domain and its simulation. This analysis is to be continued in the discussion part related to my presentation.

Jens Dietrich: The Mandarax Framework

Mandarax is a framework for derivation rules designed to integrate knowledge from distributed, heterogeneous sources. We outline the design of mandarax, and present the basic data structures used. We emphasise the differences to other derivation rule implementations such as Prolog. We discuss how Mandarax can be used to access and process semantic web data and rules, focusing on the following aspects: the integration of RDF sources, test-case driven validation, rule persistency / markup languages and verbalisation.

Jens Dietrich: Towards ECA Server Pages

In this talk we present some ideas how to process ECA (event-condition-action) rules. This approach takes advantage of the fact that nowadays compilers are often part of the runtime environment and can be used at runtime when artefacts written in abstract description languages are deployed: they are translated into source code of the target language, this code is then transparently compiled into byte or machine code and the results are loaded into the execution environment (such as the java virtual machine). The prime example implementing this approach is Java Server Pages.

We present a prototype that compiles ECA rules into scalable procedural code, and discuss the following aspects: compiler architecture, deployment, modelling of events, handling of network events, event algebra, event lifecycle, integration of derivation rules and (relational) databases, expressions, referencing terms in conditions and actions, concurrency (dealing with multiple threads) and transaction handling.