Combining FIPA ACL With DAML+OIL – A Case Study Agent Communication with CIA, FIPA ACL and Techniques of the Semantic Web Michael Schalk, Thorsten Liebig, Torsten Illmann, Frank Kargl University of Ulm James-Franck-Ring D-89069 Ulm, Germany michael.schalk@informatik.uni-ulm.de ABSTRACT of the next generation Web. The Collaboration and Coordination Infrastructure for per- Looking at current agent systems, the following limita- sonal Agents (CIA) is a Java-based multi-agent framework tions can be discovered: for personal assistance. Until now, inter-agent communi- • If agent communication is restricted to specific and cation in CIA is done via topic-based communication chan- proprietary protocols, ad hoc communication of un- nels with Java-based event classes. Information within these acquainted agents is mostly not possible. Therefore events is represented in proprietary classes, which are seri- the usage of standardized communication languages is alized for transfer. As a result, agent communication is lim- highly desirable. ited to an a priori defined domain of information chunks to which collaborating agents have to be tailored. In order to • Agent systems need many resources for reasoning and achieve wider inter-operability we are currently evaluating interference mechanisms. In a more and more global the combination of two techniques. For standardized com- and connected world with constantly upcoming smaller munication between heterogeneous agents we will use the computers (e.g. wearables, handhelds and cellulars) FIPA Agent Communication Language (ACL). The DARPA being an ideal environment for personal agent sys- Agent Communication Language / Ontology Inference Layer tems, the need for small-sized agent systems support- (DAML+OIL1 ) will serve as content language for the ACL. ing these heavyweight mechanisms is obvious. This architecture seems to be a promising combination be- cause of two reasons. First, agents of this kind are able • Using isolated and pre-defined terminologies for inter- to collaborate with other heterogenous agents in an ad hoc agent communication leads to agents which are tai- manner because of the standardized FIPA communication lored to specific vocabularies. In a global world where interface. Second, they do not have to be tailored to pro- even general terms (e. g. time, location) are represented prietary content vocabularies in advance, because they can in different manners, mechanisms for identification of use ontology-based Semantic Web techniques as a mecha- shared meanings are needed. Otherwise, totally unac- nism for identification of the meaning of the terms they quainted agents will mostly not be able to communi- communicate. However, it has to be figured out if these cate successfully. two techniques fit seamlessly into a given agent architecture like CIA. Potential problems include the semantic compat- 2. COMBINING NEW TECHNOLOGIES ibility of DAML and ACL for example. These and other In order to redress these restrictions we suggest to com- questions have to be considered with respect to the highly bine three technologies addressing different levels of agent dynamic infrastructure of a CIA system. communication. We will describe them shortly and argue their reasonable combination. Keywords 2.1 The CIA Multi-Agent Framework Ontologies and communication languages, DAML, OIL, Ap- CIA is a multi-agent framework for personal assistance [5] plications of ontologies, Ontology translation/mapping, Agent in highly dynamic environments. The main idea is to pro- communication, FIPA ACL, Lightweight agent systems vide an infrastructure for personal agents with various basic services like hardware- and location-independent communi- 1. LIMITATIONS OF CURRENT AGENT AR- cation, persistence, security, mobility and user interaction. CHITECTURES By using these services, an agent programmer is able to con- centrate on implementing the actual application logic. Per- It is commonly believed that personalized software agents sonal agents of one user are combined in a so-called agent will make daily work more efficient by collaboration with cluster which may be split over several physical hosts. CIA other available agents relying on distributed information on allows the integration of agents from external agent clusters the Web. Achieving this will require a flexible agent archi- and even from different agent systems. Agents communi- tecture, a standardized communication protocol and an ad- cate with events via topic-based channels using the asyn- equate link-up with knowledge representation technologies chronous publish/subscriber event model [6]. The underly- 1 In the following we will use the term DAML for short. ing implementation of the communication infrastructure is exchangeable in order to support any physical network con- perfectly together in order to build a framework for commu- nection and therefore enables the integration of CIA systems nication between heterogeneous agents (other related work in highly dynamic networks and in special on small devices. relies on similar assumptions, see [2] for example). CIA pro- Event types are pre-defined serializable Java classes consist- vides a robust architecture for multi-agent communication ing of a type, a set of pre-defined headers and a body. The in highly dynamic environments. FIPA ACL is well suited pre-defined headers are internally used for technical infor- to serve as conversational logic on top of the technical frame- mation (e. g. routing or quality-of-service) in order to deliver work CIA. An ontology language like DAML enables agents events accurately and reliably. The body contains instances to negotiate about arbitrary domains without narrowing on of any serializable Java class. It is used in combination with specific terms in advance. the event type to define application specific information or queries. Communication partners must have the same infor- 3. PROBLEMS TO BE SOLVED mation about these class definitions and consequently, later As explained above these three technologies are well suit- extensions are impossible. Moreover, for every application able for sophisticated inter-agent communication systems. the conversation procedure must be implemented again. However, we still have to cope with the problem of combin- ing them in a reasonable and efficient way. Integration of 2.2 The FIPA Agent Communication Language FIPA ACL within CIA can take place by embedding ACL The intention of FIPA ACL [3] is to provide conversational messages as text-based XML into CIA event bodies. Involv- logic to agents, thus raising the semantic level of agent com- ing the complete FIPA standard requires additional exami- munication to a higher level than existing technologies, e. g. nations since the user-centered agent cluster in CIA differs event-based communication in CIA. In order to achieve this, from FIPA’s view of an agent platform. The combination each of the FIPA ACL communication primitives, called of FIPA ACL with DAML leads to some interesting open Communicative Acts, is given a precise semantics by provid- questions we are currently working on: ing pre- and postconditions expressed in a first order modal • Does the CL have to represent modal operators, ob- logic. With this semantics, the agent is able to express his jects, proposition or actions of the FIPA ACL seman- personal attitude (e. g. belief, uncertainty, choice, intention) tics? If so, which is the minimal set of elements a CL towards his achieved knowledge rather than the semantics has to cover, esp. w.r.t. the desired expressiveness for of the knowledge itself. Based on this underlying semantic CIA applications? model, the agent can compile sensible options for his next action. An alternative approach for setting up intelligent • Do the semantics of both languages have to be disjoint conversations is to identify certain repeatedly used conver- in any case? sation patterns called Interaction Protocols (IPs) by examin- • What are the additional requirements for CLs? Does ing typical agent application areas, e. g. an auction. Agents every CL have to provide a possibility for querying in can communicate by agreeing to an FIPA IP and engage in order to be a candidate for a FIPA CL? a meaningful conversation simply by following a path within this IP. FIPA ACL merely places few constraints on the con- The subject of our investigation will be how DAML can meet tent language (CL) itself (how the content of a message is the requirements that FIPA imposes on CLs. expressed). It only provides the conversation envelope for Concerning CIA and DAML the challenge is to provide heavy the actual information being exchanged. weight reasoning services for agents running on small portable devices using remote reasoning service invocation. We in- 2.3 Ontology Languages and the Semantic Web tend to make this service available through a special reason- The vision of the Semantic Web is to bring meaning and ing agent running on a powerful host within the pertaining structure into information stored in Web sites or exchanged agent cluster. between Web-enabled agents [1]. Information on the Web will then migrate to knowledge due to annotated or inherent 4. REFERENCES [1] T. Berners-Lee, J. Hendler, and O. Lassila. The terms which are interconnected with terms of other sites, Semantic Web. Scientific American, May 2001. plus a set of explicit assumptions expressing the intended meaning of those terms. In the Semantic Web terminology [2] R. S. Cost et al. ITtalks: A Case Study in the Semantic such a set of knowledge terms is commonly called an on- Web and DAML+OIL. IEEE Intelligent Systems, tology. A promising candidate for a Web-based ontology 17(1):40 – 46, 2002. modeling language is DAML [4]. DAML provides a shared [3] Foundation for Intelligent Physical Agents. FIPA vocabulary of terms with formal semantics and is there- Communicative Act Library Specification, Version H. fore suitable for inter-agent knowledge exchange. A nec- http://www.fipa.org/specs/fipa00037/. essary precondition is that agents relate the terms of their [4] J. Hendler and D. McGuinness. The DARPA Agent own knowledge base with corresponding terms of ontologies Markup Language. IEEE Intelligent Systems, 15(5):34 somewhere in the Semantic Web. Mutual understanding of – 43, May 2000. agents can be accomplished if a semantic interconnection be- [5] F. Kargl, T. Illmann, and M. Weber. CIA - a tween the terminologies of the communicating agents can be Collaboration and Coordination Infrastructure for found. DAML-enabled inference engines will help to make Personal Agents. In Proc. of DAIS’99. Kluwer these semantic relationships automatically explicit. Appro- Academic Publishers, 1999. priate reasoning systems for this purpose are currently de- [6] F. Kargl, T. Illmann, and M. Weber. Evaluation of veloped utilizing research results from the established field Java Messaging Middleware as a Platform for Software of knowledge representation. Agent Communication. In Proc. of JIT 99. Springer, We believe that the previously mentioned technologies fit 1999.