Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!wuarchive!udel!burdvax!PRC.Unisys.COM!finin From: finin@PRC.Unisys.COM (Tim Finin) Newsgroups: comp.ai Subject: Re: Configuration/Constraints info wanted Keywords: Beacon Configurator Message-ID: <15965@burdvax.PRC.Unisys.COM> Date: 20 Dec 90 16:09:07 GMT References: <1990Dec20.093442.21947@portia.Stanford.EDU> Sender: news@PRC.Unisys.COM Reply-To: finin@prc.unisys.com Organization: Unisys Center for Advanced Information Technology Lines: 63 In article <1990Dec20.093442.21947@portia.Stanford.EDU>, liemandt@lindy.stanford.edu (Joe Liemandt) writes: |> |> I am interested in learning more about this area. Any pointers to |> articles or books? I have read all about XCON. |> ... |> Are there any other large configuration systems implemented besides |> XCON? |> ... We developed a configuration system for Unisys equipment here at the Paoli Research Center (now known as the Unisys Center for Advanced Information Technology). Here is a brief description. More information can be found in: David B. Searls and Lewis M. Norton. Logic-based configuration with a semantic network. Journal of Logic Programming, 1988. BEACON is an expert system developed by CAIT to aid in designing configurations of computers, or any other complex assemblages. In particular, BEACON attempts to ensure that such configurations are: (1) correct and complete at the time of creation; (2) translated into correct and complete sets of available components; and (3) based on timely engineering "product" data and in agreement with system configuration rules. The configuration design problem domain is complex, is highly product dependent, and may change with each product release. For large or rapidly changing domains such as this, system developers find it difficult to cope with the complexity and rate of change of the domain dependent logic that must be housed in their programs. With traditional software, developers simply can't match the pace of change of the knowledge; the software maintenance effort is too great and proof of correctness is difficult. BEACON implements a highly structured KB using the KNET knowledg representation system. The aggregation hierarchy here parallels the physical breakdown of systems into configurable subsystems and orderable components. Significant component attributes are also modeled in KNET on the aggregation hierarchy; the common (more general) aspects of similar systems, subsystems, and components are recorded at successively higher levels, while unique (more specialized) aspects are recorded at successively lower levels. KNET also gives modelers the power to define higher level restrictions on a configuration, through the constraint mechanism. Constraints can also provide directions to specialized application programs which work on a KNET structure, such as KNET's Database Loader which incorporates catalog-level product information (instance-level facts) into the KB from an external DB. The Database Loader permits a clean separation of volatile data from more stable structural knowledge. Modeling of KBs for BEACON is performed by experts using the KNET Browser/Editor. This tool is used for the construction, navigation, and editing of the KB, and includes a custom window-based interface. The actual configurator, used at the time a system is ordered, is an application program which traverses the KNET model of that system. It interactively prompts the user for input as it walks the network, enforcing all KNET constraints as it proceeds, screening out irrelevant and invalid information, and presenting only relevant selections at appropriate times. The configurator thus tailors its interaction with the salesperson for each order proposal, minimizing each session. It too makes use of a specialized window-based interface.