The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #92125

Title:Using Art to Model Indigenous Knowledge
Authors:Keedy,GD
Date: 1992
Presented:
Keywords:
Abstract:There were two main objectives for this MSc project. The first was to produce a computer based simulation of an agroforestry system, based on the knowledge provided by the native farmers of the ecosystem. This data is called indigenous knowledge. The second objective was to evaluate the performance of the ART expert system as the implementation tool for the simulation. The agroforestry system selected to be the simulation domain was the Kandyan Forest Gardens of Sri Lanka. This ecosystem involves growing cashcrops, fruit and vegetables in a small area of land which surrounds the farmer's home. The simulation implemented was a forward chaining system which created a network of all the possible garden states over a set number of months. Each of these states was the result of a particular management decision hypothesised by the system. The range of management decisions were taken from the indigenous knowledge. The simulation was temporal in nature and each successive state represented the garden one month later than its parent. Qualitative variables were used to represent each garden state and the influences which affected those states. The simulation generated all the possible states because the domain model was not comprehensive enough to make the system more deterministic. Some of the garden states in the network were pruned by rules, so that the combinatorial explosion of states did not bring the simulation to a grinding halt. The network was also examined by a set of constraint rules which flagged a set of management decisions as good or bad. The system then displayed all the good paths for the user to examine. The project proved that ART could be used to implement an ecological simulation. The indigenous knowledge can be used to represent the domain, but requires more detail in areas specific to simulation, for example the causal factors which determine management decisions. The simulation technique of proposing all possible states from a domain model is
Download:NO ONLINE COPY

[Search These Pages] [DAI Home Page] [Comment]