Commitment in Agent Cooperation, Applied to Agent-Based Simulation

Document type: Researchreports
Full text:
Author(s): Staffan Hägg
Title: Commitment in Agent Cooperation, Applied to Agent-Based Simulation
Translated title: "Commitment" mellan samarbetande agenter, tillämpat på agentbaserad simulering
Series: Research Report
Year: 1999
Issue: 11
ISSN: 1103-1581
Organization: Blekinge Institute of Technology
Department: Department of Software Engineering and Computer Science (Institutionen för programvaruteknik och datavetenskap)
Dept. of Software Engineering and Computer Science S-372 25 Ronneby
+46 455 780 00
Authors e-mail:
Language: English
Abstract: We regard simulation as a set of interaction episodes between partaking agents in an agent-based simulation. In order to structure the interaction we use commitments, and the semantics of such commitments is an analogy to a two-phase commit scheme used in distributed databases. A commitment is a contract between the involved agents to reach and preserve a specified goal, during a specified time. Here we define the semantics, describe the messages and outline administrative issues.
The full strength of the model is reached when an initiator needs to make commitments with a number of other agents. It then awaits the replies, and then, depending on the replies, it either confirms or rejects proposed commitments. Hereby it accomplishes an analogy to an atomic action when a two-phase commit scheme is used.
When the commitment scheme is applied to a simulation, a crucial part of the design is to model the requirements for accepting commitments during simulation. We introduce the concepts of an acceptance function and acceptance domains that are used to express these requirements. An example from battle field simulation is given.
With the proposed model simulated objects may quite easily be replaced by real objects. With the right kind of interface, real trucks and other entities may partake in real time operations, while other entities are simulated.
Summary in Swedish: I agentbaserad simulering definierar vi ett scenario som en mängd interaktionstillfällen. Varje sådant interaktionstillfälle mellan agenter (och de enheter de representerar) uttrycker vi med hjälp av begreppet "Commitment". För distribuerade databaser är detta ett väl beprövat koncept, och här utvidgar vi det till att gälla generellt för agentinteraktion.
Subject: Computer Science\Artificial Intelligence
Computer Science\Distributed Computing
Keywords: Agent-based simulation, Agent interaction, Commitment, two-phase commit
Note: The paper is also published in Proceedings of the 5th International Conference on Intelligent Autonomous Systems (IAS-5), Sapporo, Japan, June 1-4, 1998.
URN: urn:nbn:se:bth-00137