Notes from:

This page describes one approach, developed in February 2000, for design of a bdi-based multiagent approach to the swan DSS. A second, alternative approach can be found here in graphical outline form. The original "Draft Discussion Document" on using BDI Agents is here.

Comments on both approaches are welcome.

Applying a BDI Agent Architecture to the Swan Management DSS:
Regarding the swan management DSS, the following agents might be considered:

each national wildlife refuge (NWR) or other locale (e.g., Bear River Migratory Bird Refuge, Grays Lake NWR, Grand Teton National Park, Harrimann State Park, National Elk Refuge, Red Rock Lakes NWR, Yellowstone NP)
each zone as delineated in the RMP Plan
the Tri-State Area
the Rocky Mountain Population (RMP) of trumpeter swans

It will require some additional domain investigation to determine if individual zone agents are needed. Their primary purpose appears to be to provide population segment analysis at a medium geographic scale. The Tri-State Area and RMP agents may, in fact, be able to be handled by a single agent, as the domain knowledge appears to be the same. My initial thought is that each of the expert systems being developed can be considered to be an agent, however each are not structured with a BDI architecture. Currently there are four such expert systems planned or implemented using a commercial shell and web implementation product (MultiLogic):

breeding habitat needs
management of montane palustrine wetlands
migration paths and chronologies
principles of flyway management:

contribution to breeding goals
contribution to wintering redistribution goals

REFUGE SPECIFIC AGENT (generic)

Beliefs: # adults, last year # cygnets, last year cygnet:adult ratio, last year # adults, 5 year trend # cygnets, 5 year trend cygnet:adult ratio, 5 year trend quality of breeding habitat assessment: size depth ice-free pre-laying nest sites flooding brood habitat acres of habitat areas connected by migration current water levels	Belief Revision Function: would run breeding habitat and migration expert systems again as needed; would access trends DB/KB [Should this be three agents (or at least three brf's) --trends, breeding habitat quality, and migration?]
Options: Spring planned water level Summer planned water level Fall planned water level Winter planned water level	Option Generation Function: would run montane wetland management expert system as needed [May not require all intentions to do this. Is this a flaw in applying bdi logic?]
Intentions: increase, maintain, or decrease breeding numbers increase, maintain, or decrease wintering numbers meet wildlife objective and associated habitat type (from montane wetland management expert system)	Filter: calls RMP Agent to determine intentions on numbers. [This is external to the Refuge Specific Agent. Is this a flaw in applying bdi logic?] Running the ogf would also ensure that wildlife/ habitat intentions are met.
Actions: recommended water levels for the current season	Action Selection Function: ??? [Write an algorithm to determine if revised options are attainable/ realistic and if they will contribute to reaching revised intentions.]

Belief Revision Function: would run breeding habitat and migration expert systems again as needed; would access trends DB/KB

[Should this be three agents (or at least three brf's) --trends, breeding habitat quality, and migration?]

RMP Agent

Beliefs: # cygnets, last year cygnet:adult ratio, last year # adults, 5 year trend # cygnets, 5 year trend cygnet:adult ratio, 5 year trend what local areas comprise population segments/ zones breeding zone designations wintering zone designations wetland complex description rating as breeding habitat health/ survival risks	Belief Revision Function: would access trends DB/KB; would query user
Options: maintain breeding area establish new breeding area maintain wintering area establish new wintering area reduce existing wintering area Spring migration areas?[?] Fall migration areas?[?]	Option Generation Function: The set of options is static. [Write an algorithm for eliminating any options obviously infeasible for individual refuge/ time.???]
Intentions: local area should or should not try to contribute to RMP wintering numbers local area should or should not try to contribute to RMP wintering numbers and distribution	Filter: would run flyway management expert system
Actions: advise Refuge Specific Agent about population goals related to seasons/ breeding cycle [Should this be two separate agents, one for breeding and one for wintering?]	Action Selection Function: [Write algorithm with rules to prioritize seasonal goals.]

Return to the original "Draft Discussion Document" on using BDI Agents.

Return to the Swan Management Decision Support System homepage.

USDI - Geological Survey
Biological Resources Division
Northern Rocky Mountain Science Center
Maintainer: Rick Sojda (sojda@swan.msu.montana.edu)