Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


4.3. A PROPOSED DOMAIN SPECIFIC METHODOLOGY

The methodology proposed here is based on our experience during the last 7 years in developing expert systems in agriculture. It is not intended to be a comprehensive methodology but it will give some thoughts on domain-specific task architectures in agriculture. Generally, we can classify agricultural operations into precultivation operations and during cultivation operations. Precultivation operations are interested in preparing the field for growing a particular crop. During cultivation operations can be further classified into diagnosis operations (which concern finding a reasonable explanation for undesired phenomenon), and scheduling operations (which include irrigation, fertilization, and treatment operations). We will concentrate in this section on scheduling in the agriculture domain as an example.

Our methodology is based on two principles that have emerged during the last decade of AI research:

  • Knowledge-level modeling: Knowledge should be modeled on a higher level than that of exploited knowledge representation formalisms, to avoid premature design decisions and to facilitate communication with domain experts.
  • Reusability of task and domain knowledge: This principle implies that the complexity of KBS development can be relieved by the construction of reusable components libraries, just as any other engineering activity.

Following the model-driven approach, the KBS development process is composed of two phases. In the first phase, an epistemological model is constructed, representing the conceptual model for the application problem. In the second phase, this model is transformed into a computational model, implemented on computer, and preserving the structure of the epistemological model.

An epistemological model in CommonKADS consists of three main components, namely: task knowledge, inference knowledge, and domain knowledge. Generating one generic model for each task (irrigation, fertilization, and treatment), will increase the model reusability among many different crop types, which minimizes the effort and the cost of knowledge-based systems development. The following subsection describes the inference model of irrigation as an example of this methodology.

The main goal of this task is to design an irrigation schedule for a particular crop in a particular farm. The output schedule is simply a plan of water quantities to be applied and the time of application, according to the requirements of the plant, and the affecting factors like soil type, climate, source of water, etc.

Figure 1 shows the constructed generic inference structure for the irrigation task. As illustrated in Figure 1, the irrigation task starts with acquiring a case description, which includes parameters that affect the irrigation process, such as soil, water, and climate data. The expand inference step uses these initial parameters to derive other parameters needed in the subsequent inference steps. The knowledge required for this inference step is the derivation model that can be represented as object hierarchies, rules, facts, and mathematical functions, according to the nature of knowledge. The compute inference step, calculates the irrigation interval using the input and derived parameters. The knowledge required by this step is represented in the form of mathematical functions. The propose step generates a preliminary irrigation schedule depending on a set of fixed mathematical functions regardless of any additional environmental constraints, which includes all circumstances that are not covered by the irrigation model. The check violation step evaluates the proposed irrigation schedule with regard to some environmental constraints (e.g., the existence of certain disorders in the farm) and reports these violations to be fixed by the fix inference step, to produce the final, acceptable irrigation schedule.


FIGURE 1 Irrigation inference structure.

5. RESEARCH ISSUES

The research issues to be addressed in this section are: the integration of other software components with agricultural expert systems, agricultural knowledge sharing and reuse, intelligent retrieval of agricultural data, and automatic knowledge acquisition.

5.1. INTEGRATION OF SOFTWARE COMPONENTS WITH AGRICULTURAL EXPERT SYSTEMS

Some existing expert systems are integrated with other software components such as crop simulation models (Kamel et al., 1994), GIS (Loh et al., 1994), and multimedia (Rafea et al., 1995). An agent-based approach can be used to integrate distributed heterogeneous systems and hence solve the integration problems in addition to other problems related to distributed components in different environments. However, more research is needed, as described in the following subsections.

5.1.1. Crop Simulation Models and Expert Systems

The integration of numerical simulation models into the crop management expert system is receiving widespread attention. The reason is largely due to the perceived "naturalness" of an interaction between "experience" to quickly center on a part of a large search space, and numerical methods to select the correct exact solution from the narrowed possibilities. This mode of interaction leverages strengths of both expert systems and simulation models, using the expert system to quickly limit the search space, then using simulation-based methods to find the best candidate within the current focus. For example, the simulation model CERES wheat takes as input boundary conditions, such as planting date and irrigation regime, and then predicts (among other items) grain output at harvest. Although quite accurate in its output, one difficulty in the production use of CERES is the level of expertise that must be employed to set the initial input parameters. A scheme has been given by Kamel et al. (1994), that uses a compiled-level problem solver to determine the initial planting parameters (seed variety, planting date, planting and land preparation methods). These values are used to parameterize the CERES wheat model. The model is then used to predict iteratively the water and nutrient needs of the crop.

The example given has proved to be very useful in integrating numerical and symbolic methods to solve specific problems. This may encourage the integration of other types of simulation models such as pests infestation and economic models with expert systems. There remains further research to be done on the best ways of integration and interfaces between the expert system and the simulation model.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.