Semantic Weblog

Semantic Web Services and Agents

I was thinking these days about the possible ways of combining the multiagent paradigms and the semantic Web services approach.

A lot of research in the multiagent area dealt with the issues related to distributed architectures like discovery, matchmaking, composition, etc. The multiagent approach to distributed systems is generally driven by two main ideas:
- Every component (agent) is autonomous, i.e. capable of making decisions and acting proactively to achieve a given goal.
- Agents can communicate, exchange knowledge, collaborate or negotiate, to efficiently achieve a common goal.

While results from the multiagent research domain are promising in theory, their application in the real world was not very successful. I personally think that one of the major obstacles to a wide spread of multiagent techniques is the lack of standardisation. Many efforts were made in this sense (cf. FIPA), but there is still something missing I think: a low level standard of communication between agents. That is the problem Web services technology may resolve.

I had a discussion with Peter last week about these ideas. He told me that I should see if it is better to incorporate multiagent paradigms into Web service architectures (i.e. adding planning, knowledge sharing, autonomy, ... to Web services), or incorporating Web services standards into existing multiagent architectures (i.e. SOAP protocol, UDDI registry, DAML-S, ... ).

Some researchers are investigating the possibilities of incorporating agent behaviours into Web services. Others are exploring the concepts of autonomous Web services, and agents as Web services... What I have to do now is to read some papers discussing these ideas, then write a state of the art draft summarizing my readings. In this draft I should present an in-depth comparison of agent-based systems and Web services architectures strengths and weaknesses, and propose some possible solutions to the two paradigms integration.