DAI-List Digest Friday, 9 April 1993 Issue Number 115 Topics: Free Trial of AI/IS/CS News Service Query on Sharing of Knowledge Bases Query on Cooperative Engineering Design Query on DAI Re: Query on DAI Description of C.A.D.D.I.E. Administrivia: Please send submissions to DAI-List@mcc.com. Send other requests, such as changes in your e-mail address, to DAI-List-Request@mcc.com. ---------------------------------------------------------------------- From: Ken Laws Subject: Free Trial of AI/IS/CS News Service Date: Thu 8 Apr 93 16:53:02-PDT Greetings! I'm the editor of the Computists' Communique, an AI/IS/CS email news service of Computists International. Send me email saying where you saw this announcement for a *free* two-month get-acquainted subscription. You'll get job ads, journal calls, NSF announcements, grant and research news, online resources, business tips, analysis, and commentary. The Communique is about 32KB (8 pages) per week, with a high signal-to-noise ratio. Eclectic, but with special focus on AI research, information technology, software applications, and entrepreneurship. Time-saving, informative, insightful, concise, timely, useful, and no risk -- now or ever. Dr. Kenneth I. Laws Computists International laws@ai.sri.com ------------------------------ From: ntayar@imag.imag.fr (Nina Tayar) Subject: sharing of knowledge bases Date: Fri, 2 Apr 1993 14:15:55 GMT Hello Colleagues, I am a Ph.D. student at the university of GRENOBLE-FRANCE. I want to know if there are other ppeople in this group who are interested in the management of shared knowledge bases. In order to build a large scale knowledge base, we find necessary the development of a basic knowledge base which can be shared by all applications. My main occupation consists of supporting and managing shared knowledge bases at a high level (i.e., the knowledge base level and not the knowledge level). In other words, I am trying to find a design nodel for knowledge bases which will help in supporting their sharing. By a design model I mean structure and function of such knowledge bases. I just found a reference which explains how to represent and to treat a knowledge base. It is "KRYPTON: A FUNCTIONAL APPROACH TO KNOWLEDGE REPRESENTATION" written by R.J. BRACHMAN, where BRACHMAN, in order to construct his knowledge base called Krypton, treated a knowled ge base like an abstract data type characterized by a set of operations. All interactions between a user and a Krypton knowledge base are mediated by these operations. I need to know if there are other references which study the problem of representation structures for knowledge bases (and not for knowledge). In addition, I look forword to discuss my work and my ideas concerning modeling of knowledge bases or sharing of them, with anyone who feels involved in this domain of research. It will be really such a good thing if we can deal with knowledge bases as typed structures and if we can associate to these structures a set of oprerations to manipulate them. Thank you in advance for any help. Sincerely, Nina TAYAR LIFIA-HITELLA INRIA RHONES ALPES 46, ave. FELIX VIALLET 38031 GRENOBLE FRANCE Tel : (33) 76-57-47-99 email : Nina.Tayar@imag.fr [[I suggest the KIF and KQML efforts (AI Magazine, Fall 1991) - Huhns]] ------------------------------ From: lakeman@cs.utwente.nl (Siem Lakeman) Subject: Cooperative Engineering Design Date: Mon, 5 Apr 1993 15:07:06 GMT Keywords: Cooperative Engineering Design, DAI, CSCW Could someone inform me about literature and/or projects on the field of cooperative engineering design. Thanking you in advance, Simon. University of Twente, Dept. of Computer Science ------------------------------ From: rsheldon@news.dcs.warwick.ac.uk (Richard A Sheldon) Subject: Help Date: 5 Apr 93 22:46:41 GMT I'm looking for some advice about articles and books (articles preferably) that would introduce Distributed Artificial Intelligence. In particular, I'm looking for information on Speech Acts and their use in DAI communication. Any help would be gratefully achieved. Richard A Sheldon Department of Computer Science, Warwick University, England ------------------------------ From: rapaport@cs.buffalo.edu (William J. Rapaport) Subject: Re: Help Date: 9 Apr 93 16:25:05 GMT Organization: State University of New York at Buffalo/Comp Sci rsheldon@news.dcs.warwick.ac.uk (Richard A Sheldon) writes: > >I'm looking for some advice about articles and books (articles >preferably) that would introduce Distributed Artificial Intelligence. As always, the first place to check is: Stuart C. Shapiro (ed.), 1992, Encyclopedia of Artificial Intelligence, 2nd edition (New York: John Wiley). ------------------------------ From: ch_f507@titan.kingston.ac.uk Subject: distributed artificial intelligence Date: 8 Apr 93 13:02:24 GMT C.A.D.D.I.E. (Control and Direction of Distributed Intelligent Agents) Dowty Command and Control Systems would like to announce the completion of a 3 year study into Distributed Artificial Intelligence (DAI). Funded by the Department of Trade and Industry, this project has been a 3 part collaboration between ourselves, Logica, and the University of Essex. The goals of this project were to achieve a greater understanding of DAI, and to develop a testbed (CADDIE) which could illustrate the tools and techniques that potentially make DAI an extremely powerful modelling environment. CADDIE is a dynamic/stochastic modelling tool that expresses the behaviour of individuals (Agents) and the behaviour of groups of agents within an organisational unit called a functional unit (FUN). The tool allows the user to define an organisational structure and to populate that structure with intelligent agents. FUN structures can be created by defining a hierarchy of authority and responsibility levels. Individual agents within CADDIE have a set of generic capabilites. They can communicate with each other, store information about the environment they live in, and act upon this information even when it is uncertain. Any agent can set a goal to achieve and generate a plan to reach that goal and if required can consult other agents to request help, i.e. multiagent planning. Inference techniques are used to decide which task to execute next, depending upon the level and type of knowledge that an agent has at that time. To illustrate the potential of this new generic toolset two scenarios were developed. The major one is the Emergency Services scenario, which was developed with the collaboration of TO20 of New Scotland Yard. This was taken from an actual incident that occurred in the Thames Valley Police Area, where a tanker laden with chemicals crashed and deposited its load in a busy urban area. CADDIE allows users to model the events that occurred that day, and demonstrate whether the procedures that were used to clear up the spill were effective. It allows you to ask 'what if ?' questions by changing the level and type of knowledge that agents have, or corrupting the messages that pass between agents so as to feed incorrect knowledge to those agents that have to make decisions. The second scenario demonstrates that CADDIE can be used to model incidents at any planning level. The Emergency Services scenario is at the tactical level whilst the Air Traffic Management (ATM) scenario is aimed at the strategic level. This scenario demonstrates a model of flow management units (FMU) that control the flow of aircraft around Europe's skies. It provides a simple demonstration to show if distributed planning is more effective than central planning of these aircraft. During the study it became apparent that DAI is about modelling groups of humans. To model an incident such as the Emergency Services described above, we had to describe human behaviour. This meant that we had to develop models to deal with uncertainty, the interaction of humans within an organisation, the setting and achieving of goals, and the use of resources to carry out tasks. This lead to the development of the C2 (squared) shell which lies above the CADDIE toolset and provides a generic set of functions for any typical scenario. The ability to ask multiple 'what if?' questions requires some sort of measurement for judging whether one execution of an incident is more effective than another. Performance metrics have been developed by the team which look at all the areas covered by a typical scenario such as computational, C2, and application specific metrics. These are easily configurable and allow the user to analyse different results specific to their requirements. CADDIE is a hypothesis tester, that allows you to programme in procedures that you would normally use in a real world situation. These procedures can then be tested to the extreme by setting different initial conditions, and firing off random events designed to create impossible situations. Sensitivity analysis can be performed to analyse the effect on results from changes in pre-conditions. Thus users can adjust the procedures in the model to produce the results that they require. The ultimate aim will be to extract these new rules from the model and apply them to real life situations to see if they are more effective. The toolset that has been developed fully supports an Open Systems environment. A client/server architecture has been implemented that allows multiple users to access the simulation. This means that different users can access different parts of the toolset and observe results on multiple screens. CADDIE sits on a UNIX base and was designed using the Booch Object Oriented Methodology. Several programming languages are used such as C++, ROCK (Representation of Corporate Knowledge), Prolog, and our own CADDIE syntax that provides a simple method of describing Agents, FUNs, and the knowledge that they have. Thus the potential for CADDIE is massive. If it can be proved that new procedures first developed on CADDIE are more effective then this will bring major benefits to such organisations as the police, fire brigade, and the ambulance service. Vast sums of money can be saved by using CADDIE to verify techniques that would otherwise be very costly to prove in the real world. Of course training exercises in real life are still important, but CADDIE will allow you to develop results from exercises that have taken place, and prepare you for ones that you are going to carry out. In summary CADDIE can be used as a cost effective tool for running multiple scenarios under different conditions, to verify procedures that are being used in real life situations. CADDIE is also an extremely powerful tool for research into organisational theory and DAI. Further details of CADDIE can be obtained by contacting John Proffitt (or any member of the CADDIE team) at the address below. Demonstrations can also be arranged at Dowty, Feltham, by contacting us and booking a place. The CADDIE team is made up of the following members: John Proffitt Chris Breeze Patrick Cheesman Robert Hope Mari-Ann Buckman * Dowty Command & Control Systems Ltd, * Tel: 081-894-5511 * Gresham House, * Int'l: +44-81-894-5511 * Twickenham Road, * * Feltham, * Fax: 081-894-1916 * Middlesex, * Int'l: +44-81-894-1916 * TW13 6HA, * * UNITED KINGDOM * ch_f507@tamara.king.ac.uk ------------------------------ End of DAI-List Digest Issue #115 *********************************