Kurt Sundermeyer

Generic, configurable, cooperation protocols for multi-agent systems

In this paper we propose a unified and general mechanism for developing cooperation protocols in multi-agent systems. The protocols are essentially speech act based but have considerable advantages as compared to previous approaches: First, they are generic in the sense that a protocol execution algorithm can treat the domain independent parts separately from the application dependent reasoning and deciding processes involved. And second, they are recursively defined from primitives which allow a designer (or eventually the agents themselves) configure the appropriate general or domain-specific cooperation protocols.

Einführung der Agententechnologie in einem produzierenden Unternehmen - Ein Erfahrungsbericht

Four years ago, an engine plant of DaimlerChrysler AG initiated an industrial project, which had as a goal to design and implement a production system for large-series aggregate manufacturing that is more flexible and more robust than existing transfer lines. For the control of the newly designed flexible material flow system, the project considered to use an agent-based approach. In intensive cooperation, the DaimlerChrysler plant, external control suppliers, and the DaimlerChrysler research unit developed an agent-based approach for the control of the flexible material flow, showed its potential for improvement in simulation, demonstrated the technical feasibility with the help of a physical demonstrator, and implemented a prototype that was tested as a by-pass to an existing transfer line for cylinder heads and finally analyzed by the plant with respect to its profitability. This paper reports - from the point of view of the research unit - on the development steps necessary to implement the new production system. It is argued that the successful transfer of agent-oriented research results into a production environment requires several steps. In particular, the transfer is only successful if all units affected are integrated into the development process, if the technical feasibility is shown with the help of a physical demonstrator, and if the project provides a profitability analysis comparing existing and new technology. The goal of this paper is to motivate more research concerning agents in production and, based on the experience made, to highlight the critical factors for a successful transfer of agent-oriented research results.

Agent-oriented techniques for traffic and manufacturing applications: progress report

We view agent-oriented techniques (AOT) as a further development of techniques in object-oriented systems and distributed systems. AOT extend object-oriented techniques in that the analysis, design, and realization of complex systems are performed on a higher level of abstraction, and in that agents are active, concurrent objects, collectively embedded in a ‘society.’ Furthermore, agents are a specialization of objects in that the internal states of objects are typed by intentional states, for example, and in that messages exchanged among agents are classified by message types (Shoham, 1993). AOT can lean upon techniques provided by distributed systems (like sharing resources and synchronization), but they complement these techniques by making subsystems more autonomous and enabling them to co-ordinate their activities actively instead of being co-ordinated by design.

Cooperative problem-solving guided by intentions and perception (abstract)

The formalization of multi-agent autonomous systems requires a rich ontology for capturing a variety of collective behaviours and a powerful semantics for distinguishing between collective agents having, executing, and jointly intending a plan. In this paper, we introduce the notion of social agents and social plans. A definition of joint intentions is provided that avoids some of the problems encountered by previous formalizations. In particular, it models cooperation by requiring that agents adopt a joint goal and a joint plan of action before forming a joint intention. The paper also stresses the planning capability of agents and outlines a process for means-end reasoning by multiple agents.

Collaborative Supply Net Management

With the advent of web technologies the supply logistics in industry received a boost in what is nowadays called eSupply, Supply Chain Management, or - even more appropriate - Supply Net Management (SNM). In this contribution research in this domain is motivated by sketching the complexity of SNM especially in automotive industry and by showing up possible pathways into a visionary future. Then essentially two current research tracks in Daimler-Chrysler R&T are addressed, namely (1) supply net configuration, simulation and assessment, and (2) distributed supply net planning. In both domains agent technology is used, partly together with other AI techniques.

A Development and Simulation Environment for Cooperating Knowledge-Based Systems

This is a progress report on an experimental tool (DASEDIS) within a project serving for establishing a methodology for multi-agent systems. The cognitive skills of each agent are realized as a knowledge-based system. The knowledge-based systems are imbedded in DASEDIS and are accessible by its development and simulation component. In the development component DASEDIS provides procedures for implementing the knowledge-base and the problem- solving and cooperation component with various control and communication strategies. In its simulation component DASEDIS allows for simulating the intentional, robotic, and communicative aspects of each agent.

Workshop: Kommunikation, Koordination und Kooperation in Mehragenten Systemen

Fur die Verteilte Kunstliche Intelligenz ist die Kooperation zwischen den Agenten ein entscheidender Aspekt, welcher erst ermoglicht, Nutzen aus der Verteilung von Problemlosefahigkeiten zu ziehen. Jeder einzelne Agent kann zwar autonom gegenuber den anderen bzgl. seiner Existenz und seinen Fahigkeiten sein, wie dies bei den Mehragenten Systemen der Fall ist, doch nur durch die Zusammenarbeit erreicht die Gruppe ein hoheres Mas an Leistung als der Einzelne.