Title: Commitment-based service coordination

Authors: Stefan J. Witwicki, Edmund H. Durfee

Addresses: Computer Science and Engineering, University of Michigan, 2260 Hayward Street, Ann Arbor, MI, 48109, USA. ' Computer Science and Engineering, University of Michigan, 2260 Hayward Street, Ann Arbor, MI, 48109, USA

Abstract: We present a methodology for composing large-grained services that exhibit temporal uncertainty and complex task dependencies. Our multi-agent approach incorporates temporal and stochastic planning paradigms and commitment-based negotiation to achieve the coordinated provision of services with stochastic outcomes. This is all captured within a service-choreography protocol, by which agents can request future service provisions and receive probabilistic temporal service promises in return, converging on coordinated decisions about how and when to work together. We hypothesise that our approach can scalably converge on good coordination solutions because it partially decouples the problems of negotiating service interactions and computing service policies. Our empirical evaluation provides initial confirmation of this hypothesis, demonstrating the scalability and effectiveness of our approach in producing coordinated, approximately optimal agent policies in a small fraction of the time that an optimal approach requires.

Keywords: multi-agent systems; MAS; service-oriented computing; large-grained services; commitments; service coordination; negotiation; stochastic planning; Markov decision processes; linear programming; heuristic search; multi-agent systems.

DOI: 10.1504/IJAOSE.2009.022946

International Journal of Agent-Oriented Software Engineering, 2009 Vol.3 No.1, pp.59 - 87

Published online: 04 Feb 2009 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article