GRASS logo

NAME

g.infer - An rule-based expert system shell based on CLIPS.

KEYWORDS

general, expert system, inference, CLIPS, rule engine, rete

SYNOPSIS

g.g.infer
g.infer help
g.inferv [-bflmrsvx] [raster=Raster layers] [raster3d=Volume layers] [vector=Vector layers] [facts=Fact file] [payload=Rule base] [columns=Output attributes] [output=Output layer] [strategy=Inference strategy] [salience=Salience evaluation] [module=Rule module(s) for inference focus] [limit=Upper limit for inference runs] [--verbose] [--quiet]

Flags:

-d
Activate CLIPS dribble mode. Content will be written to file: outputlayer.dribble
-l
Don't use raster layer labels
-s
Stop execution before inference run
-t
Traceback of CLIPS error messages
-v
Don't delete intermediate vector layer for raster output
--verbose
Verbose module output
--quiet
Quiet module output

Parameters:

raster=Raster Layer(s)
GRASS raster layers to convert into CLIPS facts
raster3d=Raster3d Layer(s)
GRASS raster3d layers to convert into CLIPS facts
vector=Vector Layer(s)
GRASS vector layers to convert into CLIPS facts
output=Layer Name
Name of output layer (a layer used for input ot new layer)
facts=Facts File
Filename for upload of CLIPS facts
rules=Rule Base File
File containing rules for upload (batchstar) into CLIPS facts. Allows the user to input rules to g.infer from an ASCII file. This rulesfile must exist in the user's current working directory or be given by its full pathname. File rules statements take the same form as those given on the CLIPS command line.
columns=String
GRASS attributes for new output layer
strategy=Inference Strategy
Name of the inference strategy to be used
Options: depth,breadth,lex,mea,complexity,simplicity,random
Default: depth
salience=Salience Evaluation
Name of the salience evaluation approach used for the rulebase
Options: when_defined,when_activated,every_cycle
Default: when_defined
module=Rule Module Name(s)
Name of one or more module which will be put on the focus stack before the inference run
limit=Layer Name
Upper limit of inference runs
save=CLIPS entities
Selection of CLIP entities to be written to disk - add binary to override ASCII output
Options: constructs,facts,instances,binary
Default: ASCII files
watch=CLIPS entities
Selection of CLIP entities to be watched during inference run
Options: activations,compilations,facts,functions,genericfunctions,globals,methods,messagehandlers,messages,rules,slots,statistics,all
Default: none
print=CLIPS entities
Selection of CLIP entities to be printed out before the inference run
Options: agenda,breakpoints,classes,deffacts,definstances,facts,functions,generics,globals,instances,rules,subclassinstances,templates
Default: none
config=CLIPS Configuration
Configuration options for CLIPS engine
Options: autofloatdividend,dynamicconstraintchecking,factduplication,incrementalreset,resetglobals,sequenceoperatorrecognition,staticconstraintchecking
Default: none
classdefault=CLIPS Default Setting
Class default of CLIPS engine
Options: convenience,conservation
Default: convenience

DESCRIPTION (DRAFT)

The module g.infer provides an expert system shell for the GRASS GIS environment based on the C Language Integrated Production System (CLIPS). GRASS raster-, raster3d- and vector- layers and GRASS session metadata are transformed into logical facts to be acted on by CLIPS. The behaviour of the CLIPS inference engine is controlled by a set of user-defined production rules will be applied by the CLIPS inference engine. The production rules are typed into a file by the user using a system editor like vi, and then input to g.infer as the rules-file named on the command line. The inferred results are used to update one of the input layers or to create a new vector layer. In the latetr case the attribute structure is defined by the columns-option). Besides the mandatory rulebase, additional CLIPS-constructs (facts, procedural code, CLIPS Object Oriented Language (COOL) objects) can be provided prior to the inference run via the rules-file or the facts-option. All CLIPS constructs can be printed (print-option) or saved to file (save-option) and the inference run can be logged (dribble-option) to file.

The results are used to generate a new raster map layer named infer in the user's current mapset. If run without arguments, g.infer starts with a GUI or the command line interface. Currently no CLIPS command line prompt for interactive use is provided. The program will be run non-interactively if the user specifies the name of a rules file and any desired flags on the command line, using the form (simple raster processing):

g.infer [-lt] raster=rasterlayer rules=rulebase output=rasterlayer
where rulebase is the name of an ASCII file containing valid production rules to g.infer, and the (optional) flags -l and -t have the meanings described in the PARAMETERS section.

Alternately, the user can simply type g.infer on the command line, without program arguments. In this case, the user will be prompted for the needed parameter value and flag settings using the standard GRASS parser interface.

NOTES (DRAFT)

The current version of g.infer lacks a spatial index of all facts derived from the GRASS input layers. Rule constructs making searches or changes to large numbers of facts can become quickly computation intensive.

Documentation for CLIPS and pyCLIPS is recommended for in depth srtudy on how to formulate rules in CLIPS or how to use COOL (CLIPS Object Oriented Language) instances.

Interfacing with othe rule-based systems such as Protege2000 can be achieved by using the -x flag to create output files containing the facts, rules and instances known to CLIPS before the actual inference run. This is also a good starting point for debugging.

It is not recommended to use g.infer as a substitute for r.mapcalc, which provides far superior computation performance. If such tasks nevertheless must be achieved in g.infer, it is advised to use declarative programming (see CLIPS manual FIIFAAFOOO) instead of inference rules.

tbd: Only point vector support is provided

tbd: rule-driven issuing of GRASS commands

tbd: g.infer functions to turn GIS spatial layers into CLIPS-facts.

CLIPS Documentation (DRAFT)

The usage of CLIPS is beyond the scope of this module description. Detailed information about the usage and programming of CLIPS is available here.

A recommended starting point is the CLIPS users guide.

A introduction book about expert systems with an emphasis on CLIPS is Expert Systems by Giaratano and Riley.

SEE ALSO

r.infer

See also r.infer

GRASS Debugging

tbd

pyCLIPS Debugging

tbd

CLIPS Debugging

tbd

AUTHOR

Peter Löwe, GISIX.com


Special recognition goes to:

James Westervelt, U.S.Army Construction Engineering Research Laboratory

Mr. Westervelt released the original GRASS inference module, r.infer, which was maintained until versions GRASS 5.x.

Robert Inder, Interactive Information Limited

Mr. Inder developed Clips And Perl with Extensions (CAPE), available here, which was used instead of pyclips to implement a successor module GRASSCAPE .

Gary Riley

For documentation effort and maintenance of CLIPS. Mr. Riley develops and maintains the CLIPS module

Francesco Garosi/JKS

Mr. Garosi develops and maintains the pyCLIPS module, which fully embeds the CLIPS engine, with COOL (Clips Object Oriented Language) and environments support.

Last changed: $Date: 2012-02-10 19:00:00 +0100 (Fri, 10. Feb 2012) $


Main index - general index - Full index

© 2012 GRASS Development Team