Expert system design

• For theoretical introduction and other information, read the following article: ARD
• The ARD design is supported by tools:
  • VARDA
  • HJed
• During this lab classes the VARDA tool will be used.

• The VARDA-based design requires a knowledge of Prolog language.
• The most convenient way of design is to prepare a separate file, which contains an ARD model. Then such a file can be loaded into VARDA. The simple example of VARDA file can be found here: hekate_case_thermostat-mdl.pl
• VARDA provides a dedicated interface (predicates) that facilitate the ARD design:
  • Adding a new attribute:

    ard_att_add('Attribute_name')

    Creates an attribute with the specified name. In order to use an attribute, it must be previously defined with the help of this predicate.

  • Adding a new property:

    ard_property_add(['Attribute_name'])

    Creates the property containing one attribute called Attribute_name. This predicate is usually used on the first stage of designing for creation of the first, most general model of the system.

  • Model finalization:

    ard_finalize(property,[list_of_attributes])

    Finalization allows for creation a more detailed description of the system. In practice, it replaces a property containing one conceptual attribute with other property containing a several attributes, which more precise describe the system than the conceptual one i.e.:

    ard_finalize(['Time'],['Date', 'Hour', season, operation])

    In this example the property containing the attribute Time is replaced with the property containing attributes Date, Hour, season and operation. These attributes more precise define the concept of Time. The finalization causes the following changes on a ARD diagram.
    Before:
    
    and after:
    

  • Split operation allows for division of one property containing the several attributes into several properties. During the split operation, the dependencies between new properites must be defined:

    ard_split(complex_property,
                   list_of_properties,
                   list_of_dependencies_between_properties)

    • complex_property - is in the form [att1, att2, att3]
    • list_of_properties - list of new properites:
      • [ [att1], [att2], [at3] ]
      • [ [att1, att2], [at3] ]
      • [ [att1], [att2, at3] ]
    • list_of_dependencies_between_properties a list of pairs that defines a list of dependencies between properties. In example:

      [[[att1], [att2]], [[att1], [att3]]]

      defines the following dependencies:

      • property [att2] depends on property [ att1]
      • property [att3] depends on property [att1]

    • [[[att1, att2], [att3]]]

      defines the following dependencies:

      • property [att3] depends on property[att1, att2]

    • [[[att1], [att2, att3]], [[att2,att3], [att4]]]

      defines the following dependencies:

      • property [att2, att3] depends on property[att1]
      • existing property [att4] depends on property [att2, att3]
    • The complete code can be as follows:

      ard_split([att1, att2, att3],
                     [[att1], [att2, at3]],
                     [
                         [[att1], [att2, att3]], 
                         [[att2,att3], [att4]]
                     ])

      For the above example, the split operation in the form:

      ard_split(['Date','Hour',season,operation],
      		  [['Date','Hour'],[season,operation]],
      		  [[['Date','Hour'],[season,operation]],[[season,operation],['Temperature']]]),

      causes the following changes on a ARD diagram.
      Before:
      
      and after:
      

• In this exercise the model of the Cashpoint system will be designed. The design starts with the highest ARD level and TPH diagram:
  
• The details can be found here: ard_design
• The first step is to create a VARDA file, which describes the above system.
• Then the file can be loaded into VARDA:

  ./varda filename.pl

• All the errors should be fixed.
• Use the following template:

  varda_model_desc :- write('Cashpoint FirstName LastName').
  
  varda_model_cashpoint :-
    % model definition
  :- varda_model_cashpoint.

• Po poprawnym wczytaniu modelu mamy do dyspozycji następujące polecenia:
• When the model is loaded successfully, the following commands can be used:
  • shl. start interactive shell,
  • sar. show the ARD diagram, can be used sar(File),
  • shi. show the ARD history (TPH), can be used shi(File),
  • sha. show the TPH and ARD combined, with can be used sha(File),
  • hic. ARD history collapse, one level up,
  • tic. TPH/ARD history collapse, one level up, can be used tic(File),
  • gax. generate XTT,
  • sxt. show the XTT diagram, can be used sxt(File),
  • pxt. print the XTT diagram,
  • xop. optimize XTT,
  • kid. kill all displays,
  • axg. write ARD model in XML (HML2), can be used axg(File),
  • xxg. write XTT model in XML (MHL2), can be used xxg(File),
  • arp. write ARD model in Prolog, can be used arp(File), The file can be subsequently consulted to recreate the model,
  • pur. purge the entire model.
  • pux. purge the XTT model.
  • gen. run the experimental ARD generator.
  • hlp. get help,
  • hlt. go „bye, bye!”.
• Using these commands the ARD and TPH diagrams of the more general levels can be displayed.
• The last step is to create a hml file containing a XTT model schema. This file is a starting point for the next lab classes.

Z braku lepszego miejsca tutaj studenci wpisują komentarze natury ogólnej do tego lab.