Challenges in exception handling in multi-agent systems
Proceedings of the 2006 international workshop on Software engineering for large-scale multi-agent systems
An architecture for exception management in multiagent systems
International Journal of Agent-Oriented Software Engineering
Challenges for Exception Handling in Multi-Agent Systems
Software Engineering for Multi-Agent Systems V
A definition of exceptions in agent-oriented computing
ESAW'06 Proceedings of the 7th international conference on Engineering societies in the agents world VII
| Hi-index | 0.00 |
Multiagent systems involve a rich variety of interactions among agents-situated computations that are autonomous in their behavior and heterogeneous in structure. These interactions can be realized unambiguously if they are governed by published protocols, since agents diverse in their structure and behavior can interact as long as they respect the protocols. However, traditional protocol specifications are unduly rigid for application in open settings involving autonomous entities. They represent protocols simply as an ordering of steps and stifle the participants' autonomy due to a lack of flexibility during enactment. Commitments among agents, which are akin to contractual obligations among businesses, are a powerful abstraction for modeling flexible protocols. Commitment-based design enables a more faithful model of the openness of the business world. However, modeling business interactions requires a rich variety of interaction protocols that can capture the needs of different applications. Whereas general (business) protocols might most flexibly characterize the interactions of their participants, protocols often must be refined based on the environment in which they are to be deployed, so as to yield improvements along various properties such as performance and risk outlay, when applied to real-world tasks such as in e-business. We introduce a formal semantics and an operational characterization for commitment-based protocols wherein traditional software engineering notions such as refinement and aggregation are extended to apply to protocols. We also develop a principled approach for the design of such protocols in addition to methodologies for modeling and handling exceptions in them. We demonstrate, with appropriate examples, the benefits of this approach over traditional ones when applied to business process modeling and enactment. Our chief contributions are (1) A theoretical basis for describing protocol refinement using subsumption hierarchies and an algebra for composing protocols using existing ones. (2) A methodology for modeling and handling exceptions in commitment protocols that incorporates the preferences of the protocol designer and policies of the participants and enables specification of exceptions independent of the protocol specification. (3) Two methodologies for designing commitment protocols, one by enhancing an existing agent-oriented software engineering methodology, and another by deriving protocols from agent conversations. Our work draws from and contributes to agent communication, business process modeling and enactment, service-oriented computing, and software engineering.