Generalized configurator using a declaratively constructed two-level bi-partite graph as a knowledge representation

• US 5,630,025 A
• Filed: 07/13/1994
• Issued: 05/13/1997
• Est. Priority Date: 07/13/1994
• Status: Expired due to Term

• Current Assignee: Unisys Corporation
• Original Assignee: Unisys Corporation
• Inventors: Dolby, Nigel I., Nagle, Timothy E., Goessling, Thomas R.
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/274,618
• Time in Patent Office: 1,035 Days
• Field of Search: 395/50, 395/51, 395/54, 395/919, 395/10-13, 395/60-61, 395/76
• US Class Current: 706/46
• CPC Class Codes: G06F 2111/04 Constraint-based CAD
• CPC Class Codes: G06F 30/18 Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling circuit design at the physical level G06F30/39; network planning tools for wireless communication networks H04W16/18
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

A generalized configuration expert system for generating a complete, legal, and near-optimal configuration for any complex system consisting of multiple components is disclosed. The present invention allows a developer to specify a configurator framework for solving a particular configuration problem. A user then operates the customized configurator to generate a configuration solution based on the user'"'"'s requests and the system'"'"'s requirements and constraints. The generalized configurator uses declaratively constructed graphs and multiple interacting packing engines. A two-level, bi-partite, spreading activation graph is used as the knowledge representation of the components to be configured and their associated relationships. The invention dynamically manages the interaction of the multiple packer engines to select the appropriate piece of the total configuration problem to work on at any point in time, while still taking into account the other packing problems. The invention provides the ability to declaratively define the constraints used by the packing engines to assure correct configuration results.

Claims

1. An expert system using control logic based on spreading activation and graphs as a knowledge representation for generating a value configuration of connected components, said expert system comprising:

a priori net means for storing component definitions declaratively specified by a configurator developer, wherein said component definitions and the implied requirements of their use are represented as first nodes in a first spreading activation bi-partite graph;

instance net means for storing instances of components defined in said a priori net means interactively selected by a configurator user, wherein said instances and the interconnections between said instances are represented as second nodes in a second bi-partite graph; and

processing means coupled to said a priori net means and said instance net means for accepting requests from the configurator user to configure selected components, matching said configurator user requests to said component definitions, automatically propagating logical implications of said configurator user requests across as many ones of said first nodes of said first spreading activation bi-partite graph and said second nodes of said second bi-partite graph as are required to build a complete set of connected components fulfilling said configurator user requests by creating and connecting selected instances of said selected components only if creation and connection of said selected instances are valid based on said component definitions in said a priori net means and prior configurator user requests, and reporting the configuration resulting from said configurator user requests to the configurator user.

2. The expert system as in claim 1, further comprising loader means coupled to said a prior net means for converting said component definitions specified by the configuration developer from a text-based external format to a bi-partite graph-based internal format, and for loading said component definitions represented in said bi-partite graph-based internal format into said a priori net means.

3. The expert system as in claim 2, further comprising interface means coupled to said processing means for accepting said requests to configure selected components from the configurator user, for forwarding said requests to said processing means, and for displaying the configuration received from said processing means to the configurator user.

4. The expert system as in claim 3, further comprising a net definition knowledge base means referenced by said loader means and said processing means, said net definition knowledge base means having at least one piece of knowledge relating to each type of objects that may represent said component definitions.

5. The expert system as in claim 4, wherein said processing means determines the validity of said requests to connect said selected components by analyzing the relationships between said first nodes in said first spreading activation bi-partite graph and said second nodes in said second bi-partite graph, said first and second nodes being selected by automatically performing spreading activation logic on said first spreading activation bi-partite graph and said second bi-partite graph.

6. A non-rule-based configurator expert system expert using control logic based on spreading activation and graphs as a knowledge representation for generating a complete, valid, near-optimal configuration of connected components, the configurator expert system being customized for a given configuration domain by a configurator developer, and operated by a configurator user to generate a configuration solution based on user requests and predetermined component requirements and connection constraint expressions contained in component definitions, said non-rule-based configurator expert system comprising:

an a priori net containing component definitions declaratively specified by the configurator developer, wherein said component definitions and the implied requirements of their use are represented as first nodes in a first spreading activation bi-partite graph;

an instance net containing instances of components defined in said a priori net interactively selected by the configurator user, wherein said instances and the interconnection between said instances are represented as second nodes in a second bi-partite graph; and

a logic engine coupled to said a priori net and said instance net, wherein said logic engine accepts requests from the configurator user to connect selected components, matches the configurator user requests to said component definitions stored in said a priori net, automatically propagates logical implications of configurator user requests across as many ones of said first nodes of said first spreading activation bi-partite graph and said second nodes of said second bi-partite graph as are required to build a complete set of fully interconnected components fulfilling said configurator user requests by creating and connecting selected instances of configurator user.

7. The non-rule-based configurator expert system as in claim 6, further comprising a loader module coupled to said a priori net to convert said component definitions specified by the configurator developer from a text-based external format to a graph-based internal format, and load said component definitions represents in said graph-said internal format into said a priori net.

8. The non-rule-based configurator expert system as in claim 7, further comprising an interface module coupled to said loader module and said logic engine, wherein said interface module receives said component definitions from the configurator developer, forwards said component definitions to said loader module, accepts said requests to connect selected components from the configurator user, forwards said requests to said logic engine, and displays the configuration received from said logic engine to the configurator user.

9. The non-rule-based configurator expert system as in claim 8, further comprising a net definition knowledge base referenced by said loader module and said logic engine, said net definition knowledge base having at least one piece of knowledge relating to the type of objects that may represents said component definitions.

10. The expert system as in claim 9, wherein said logic engine determines the validity of said requests to connect said selected components by analyzing the relationships between nodes contained in said first spreading activation bi-partite graph and said second bi-partite graph, said first and said second nodes being selected by automatically performing spreading activation logic on said first spreading activation bi-partite graph and said second bi-partite graph.

11. A non-rule-based configurator expert system using control logic based on spreading activation and graphs as a knowledge representation for generating a valued configuration of components connected to form a computer system, said non-rule-based configurator expert system comprising:

a priori net means for storing computer system component definitions declaratively specified by a configurator developer, wherein said component definitions and the implied requirements of their use are represented as first nodes in a first spreading activation bi-partite graph;

instance net means for storing instances of computer system components defined in said a priori net means interactively selected by a configurator user, wherein said instances and the interconnections between said instances are represented as second nodes in a second bi-partite graph; and

processing means coupled to said a priori net means and said instance net means for accepting requests from the configurator user to configure selected computer system components, matching said configurator user requests to said computer system component definitions stored by said a priori database means, automatically propagating logical implications of configurator user requests across as many ones of said first nodes of said first spreading activation bi-partite graph and said second nodes of said second bi-partite graph as are required to build a complete set of connected computer system components fulfilling said configurator user requests by creating and connecting selected instances of said selected computer system components in said instance database means only if creation and connection of said selected instances are valid based on said computer system component definitions in said a priori net means and prior configurator user requests, and reporting the configuration resulting from said configurator user requests to the configurator user.

12. The non-rule-based configurator expert system as in claim 11, further comprising loader means coupled to said a priori net means for converting said computer system component definitions specified by the configuration developer from a text-based external format to a graph-based internal format, and for loading said computer system component definitions represented in said graph-based internal format into said a priori net means.

13. The non-rule-based configurator expert system as in claim 12, further comprising interface means coupled to said processing means for accepting said requests to said processing means, and for displaying the configuration received from said processing means to the configurator user.

14. The non-rule-based configurator expert system as in claim 13, further comprising a net definition knowledge base means coupled to said loader means and said processing means, said net definition knowledge base means having at least one piece of knowledge relating to each type of objects that may represent said computer system component definitions and referenced by said loader means and said processing means.

15. The non-rule-based configurator expert system as in claim 14, wherein said processing means determines the validity of said requests to connect said selected computer system components by analyzing the relationships between said first nodes in said first spreading activation bi-partite graph and said second nodes in said second bi-partite graph, said first and second nodes being identified by automatically performing spreading activation logic on said first spreading activation bi-partite graph and said second bi-partite graph.

16. A computer-based, non-rule-based configurator expert system using control logic based on spreading activation and graphs as a knowledge representation for generating a complete, valid, near-optimal configuration of connected components, the configurator expert system being customized for a given configuration domain by a configurator developer, and operated by a configurator user to generate a configuration solution based on configurator user requests, predetermined component requirements and connection constraints contained in component definitions, said computer-based, non-rule-based configurator expert system comprising:

an a priori net representing component definitions declaratively specified by the configurator developer, said component definitions being logically coupled to resources implied by the user of said component definitions to form a first bi-partite, spreading activation graph wherein an arc of said first bi-partite, spreading activation graph connects a node representing one of said component definitions to at least one node representing at least one of said resources;

an instance net representing created instances of components defined in said a priori net interactively selected by the configurator user, said instances being logically

coupled to instances of said resources to form a second bi-partite graph, wherein an arc of said second bi-partite graph connects a node representing one of said instances to at least one node representing at least one instance of said implied resources; and

a logic engine coupled to said a priori net and said instance net, wherein said logic engine accepts requests from the configurator user to connect selected components, matches the requests to said component definitions, automatically propagates logical implications of configurator user requests across as many nodes of said first bi-partite, spreading activation graph and said second bi-partite graph as are required to build a complete set of interconnected components fulfilling said configurator user requests, creates and connects said selected instances of said selected components in said second bi-partite graph only if creation and connection of said selected instances are valid based on said component definitions in said a priori net and previous configurator user requests, validity being determined by analyzing the relationships between nodes contained in said first bi-partite, spreading activation graph and said second bi-partite graph and connection constraints, and reports the configuration resulting from said requests to the configurator user.

17. The computer-based, non-ruled-based configurator expert system as in claim 16, further comprising a loader module coupled to said a priori net, wherein said loader module converts said component definitions declaratively specified by the configurator developer from a text-based external format to said fist bi-partite, spreading activation graph-based internal format, and loads said component definitions represented in said first bi-partite, spreading activation graph-based internal format into said a priori net.

18. The computer-based, non-rule-based configurator expert system as in claim 17, further comprising an interface module coupled to said loader module and said logic engine, where said interface module receives said component definitions from the configurator developer, forwards said component definitions to said loader module, accepts said requests to connect selected components from the configurator user via an input device, forwards said requests to said logic engine, and displays the configuration received from said logic engine on a display means to the configurator user.

19. The computer-based, non-rule-based configurator expert system as in claim 18, further comprising a net definition knowledge base referenced by said loader module and said logic engine, said net definition knowledge base having at least one piece of knowledge relating to the type of objects that may represent said component definitions.

20. A computer-based, non-rule-based configurator expert system using control logic based on spreading activation and graphs as a knowledge representation for generating a complete, valid, near-optimal configuration of components connected to form a computer system, the configurator expert system being customized for a given computer system configuration domain by a configurator developer, and operated by a configurator user to generate a configuration solution based on configurator user requests, predetermined computer system component requirements and connection constraints contained in computer system component definitions, said computer-based, non-rule-based configurator expert system comprising:

an a priori net representing computer system component definitions declaratively specified by the configurator developer, said computer system component definitions being logically coupled to resources implied by the use of said computer system component definitions to form a first bi-partite, spreading activation graph wherein an arc of said first bi-partite, spreading activation graph connects a node representing one of said component definitions to at least one node representing at lest one of said resources;

an instance net representing instances of computer system components defined in said a priori net interactively selected by the configurator user, said instances being logically coupled to instances of said resources to form a second bi-partite graph, wherein an arc of said second bi-partite graph connects a node representing one of said instances to at least one node representing at least one instance of said resources; and

a logic engine coupled to said a priori net and said instance net, wherein said logic engine accepts requests from the configurator user to connect selected computer system components, matches the requests to said computer system component definitions, automatically propagates logical implications of configurator user requests across as many nodes of said first bi-partite, spreading activation graph and said second bi-partite graph as are required to build a complete set of interconnected computer system components fulfilling said configurator user requests, creates and connects selected instances of said selected computer system components in said second bi-partite graph only if creation and connection of said selected instances are valid based on said computer system component definitions and previous configurator user requests, validity being determined by analyzing the relationships between nodes inherent in said first bi-partite, spreading activation graph and said second bi-partite graph, and connection constraints, and reports the configuration resulting from said requests to the configurator user.

21. The computer-based, non-rule-based configurator expert system as in claim 20, further comprising a loader module coupled to said a priori net, wherein said loader module converts said computer system component definitions declaratively specified by the configurator developer from a text-based external format to said first bi-partite, spreading activation graph-based internal format, and loads computer system component definitions represented in said first bi-partite, spreading activation graph-based internal format into said a priori net.

22. The computer-based, non-rule-based configurator expert system as in claim 21, further comprising an interface module coupled to said loader module and said logic engine, wherein said interface module accepts said requests to connect selected computer system components from the configurator user via an input device, forwards said requests to said logic engine, and displays the configuration received from said logic engine on a display means to the configurator user.

23. The computer-based, non-rule-based configurator expert system as in claim 22, further comprising a net definition knowledge base referenced by said loader module and said logic engine, said net definition knowledge base having at least one piece of knowledge relating to each type of objects that may represent said computer system component definitions.

24. The computer-based, non-rule-based configurator expert system as in claim 23, wherein said interface module further comprises a graphic user interface for accepting inputs from and displaying icons, menus, and data entry windows to the configurator user.

25. The computer-based, non-rule-based configurator expert system as in claim 24, wherein said expert system utilizes a system architecture based on the derivation of configuration knowledge by automatically propagating logical implications of configurator user requests across nodes of said first bi-partite, spreading activation graph and creating instances to fulfill said configurator user requests in said second bi-partite graph.

26. The computer-based, non-rule-based configurator expert system as in claim 25, wherein said a priori net can be defined, modified, updated, or maintained declaratively by the configurator developer via a text editor and said loader module, and said expert system may be modified without re-compiling any included computer programs.

27. The computer-based, non-rule-based configurator expert system as in claim 26, wherein said instance net can be modified, updated and maintained by the configurator user via the interface module.

28. A computer-implemented method for generating a complete, valid near-optimal configuration of connected components, wherein each component can be described by a component definition, said computer-implemented method comprising the steps of:

(a) declaratively defining components to be configured;

(b) representing the component definitions and the resources implied by their use as nodes in a first bi-partite, spreading activation graph, wherein an arc of said first bi-partite, spreading activation graph connects a first selected node representing one of said component definitions to at least one second selected node representing said resources;

(c) accepting a request to configure selected components from a configurator user;

(d) automatically propagating according to spreading activation the logical implications of implementing said configurator user request across multiple nodes of said fist bi-partite, spreading activation graph to build a complete set of interconnected components fulfilling said configurator user request by creating and connecting instances of said selected components and said resources defined in said first bi-partite, spreading activation graph as nodes in a second bi-partite graph in response to said configurator user requests;

(e) determining if said creation and connection of said instances is valid based on the component definitions, previous configurator user requests, and the current state of said second bi-partite graph;

(f) removing said instances from said second bi-partite graph if said creation and connection of said instances was determined to be invalid in the previous step; and

(g) reporting the configuration represented by said second bi-partite graph which resulted from said configurator user request to the configurator user.

29. The computer-implemented method of claim 28, wherein said representing step comprises the steps of:

(a) obtaining a definition of a current component from a declaratively specified text-based file of component definitions supplied by the configurator developer;

(b) creating an item node in said first bi-partite, spreading activation graph for said current component;

(c) loading said item node with information describing the requirements and constraints on connecting said current component to other components in said first bi-partite, spreading activation graph;

(d) creating one or more resource nodes representing resources implied by the user of said current component;

(e) connecting said item node with said resource nodes and any pre-existing item nodes supplying resources consumed by said item node; and

(f) repeating steps (a) through (e) for all components defined by the configurator developer.

30. The computer-implemented method of claim 28, wherein said accepting step comprises the steps of:

(a) displaying menus, icons, and data entry windows as part of a graphical user interface to the configurator user; and

(b) accepting configurator user requests to configure components by responding to commands and data entered via said menus, icons, and data entry windows.

31. The computer-implemented method of claim 28, wherein said reporting step comprises the steps of:

(a) converting the nodes of said second bi-partite graph into display objects capable of being displayed by a graphical user interface; and

(b) displaying said display objects on a display for viewing by the configurator user.

32. A computer-implemented method for generating a complete, valid near-optimal configuration of components connected to form a computer system, wherein each computer system component can be described by a computer system component definition, comprising the steps of:

(a) declaratively defining computer system components to be configured;

(b) representing the computer system component definitions and the resources implied by their use as nodes in a first bi-partite, spreading activation graph;

(c) accepting a request to configure selected computer system components from a configurator user;

(d) automatically propagating according to spreading activation the logical implications of implementing said configurator user request across multiple nodes of said first bi-partite, spreading activation graph to build a complete set of interconnected computer system components fulfilling said configurator user request by creating and connecting instances of said selected computer system components and said resources defined in said first bi-partite, spreading activation graph as nodes in a second bi-partite graph in response to said configurator user request;

(e) determining if said automatic spreading activation propagation and creation and connection of said instances is valid based on the computer system component definitions, previous configurator user requests, and the current state of said second bi-partite graph;

(f) removing said instances from said second bi-partite graph if said creation and connection of said instances was determined to be invalid in the previous step; and

(g) reporting the computer system configuration represented by said second bi-partite graph which resulted from said request to the configurator user.

33. The computer-implemented method of claim 32, wherein said representing step comprises the steps of:

(a) obtaining a definition of a current computer system component from a declaratively specified, text-based file of computer system component definitions supplied by the configurator developer;

(b) creating an item node in said first bi-partite, spreading activation graph for said current computer system component;

(c) loading said item node with information describing the requirements and constraints on connecting said current computer system component to other computer system components in said first bi-partite, spreading activation graph;

(d) creating one or more resource nodes representing resources implied by the use of said current computer system component;

(e) connecting said item node with said resource nodes and any pre-existing item nodes supplying resources consumed by said item node; and

(f) repeating steps (a) through (e) for all computer system components defined by the configurator developer.

34. The computer-implemented method of claim 32, wherein said accepting step comprises the steps of:

(a) displaying menus, icons, and data entry windows as part of a graphical user interface to the configurator user; and

(b) accepting configurator user requests to connect computer system components by responding to commands and data entered via said menus, icons, and data entry windows.

35. The computer-implemented method of claim 32, wherein said reporting step comprises the steps of:

(a) converting the nodes of said second bi-partite graph into display objects capable of being displayed by a graphical user interface; and

(b) displaying said display objects on a display for viewing by the configurator user.