A Soft Constraints-Based Approach for Reconciliation of Non-Functional Requirements in Web Services-Based Multi-Agent Systems

摘要

Over the last years agents have proved to be a powerful mechanism for solving a range of problem sets. In order to fully leverage their power in a multi-agent environment with heterogeneous participants, the establishment of barrier-free communication between different parties is of prime concern. In particular, such an environment may consist of numerous entities, based on different implementations and with different non-functional requirements (NFRs), e.g. encryption key length or latency. Currently, interoperability between the involved parties in multi-agent systems is limited as there is no established way of elaborating a common set of NFRs. In this paper, we outline a Web services-based architecture addressing this issue of interoperability in multi-agent systems. Furthermore, we introduce a soft constraints-based approach for modeling and reconciling the NFRs. In particular, Web services serve as a tool for establishing an open system providing reliable communication between a variety of agent implementations and possible non-agent based entities while lacking the limitations of Agent Communication Languages (ACLs), which are asynchronous in nature and closed in implementation. In this way, Web services enable the desired level of openness and robustness needed for maximum interoperability. Adding a soft constraints-based approach to the Web services agent layer provides an extensible and flexible mechanism for the reconciliation of interdependent NFRs. We present a prototype implementation of an open, Web services-based framework optimizing unary soft constraints in a pareto optimal way and therefore enabling barrier-free communication in a multi-agent scenario. Due to its open nature, the framework is applicable to any heterogeneous environment in need of reconciling NFRs. To illustrate its functioning, we briefly demonstrate the framework in the potential use case of a Demand Driven Supply Network (DDSN).