Knowledge engineering for modelling reasoning in a diagnosis task: application to search and rescue

• Published date: 01 September 2007
• Author: Stéphane Schvartz,Irène Abi‐Zeid,Nicole Tourigny
• DOI: http://doi.org/10.1002/cjas.24
• Date: 01 September 2007

Can J Adm Sci

Copyright © 2007 ASAC. Published by John Wiley & Sons, Ltd. 196 24(3), 196–211

Knowledge Engineering for Modelling

Reasoning in a Diagnosis Task:

Application to Search and Rescue*

Stéphane Schvartz

Laboratoire ERICAE

Irène Abi-Zeid**

Laboratoire MONADE

Nicole Tourigny

Laboratoire ERICAE

Université Laval

Canadian Journal of Administrative Sciences

Revue canadienne des sciences de l’administration

24: 196–211 (2007)

Published online in Wiley Interscience (www.interscience.wiley.com). DOI: 10.1002/CJAS.24

This project was funded by a bourse en milieu pratique (FCAR),1 the National Search and Rescue Secretariat New Initiatives Fund, and by the NSERC.

We would like to thank all the military off‌i cers who participated in this project, especially the expert/instructor for his time and generosity. We would

also like to thank the anonymous reviewers for their constructive comments and suggestions. We are very grateful to Oscar Nilo for his technical

assistance.

* Please note that this manuscript was originally submitted and reviewed in French and translated into English for publication in the print and electronic

edition of CJAS. The French version of the manuscript will be accessible electronically from Wiley Interscience (www.interscience.wiley.com/journal/cjas).

** Please address correspondence to Irène Abi-Zeid, Laboratoire MONADE, Département d’opérations et systèmes de décision, Faculté des Sciences

de l’Administration, Université Laval, Québec, (Québec), G1K 7P4, Canada. Email:irene.abi-zeid@osd.ulaval.ca

Abstract

This paper pertains to the application of knowledge engi-

neering methods to aeronautical search and rescue in

Canada. We modelled at the knowledge level the reason-

ing process of a search mission coordinator when con-

ducting a diagnosis task to determine the routes likely

followed by an aircraft missing overland. Knowledge

engineering allowed us to develop a reasoning model

based on documents and on interviews with domain

experts. Our study was conducted in the CommonKADS

knowledge modelling framework. The proposed model

was validated by domain experts and implemented in

a rule-based prototype. Copyright © 2007 ASAC. Pub-

lished by John Wiley & Sons, Ltd.

JEL Classif‌i cations: D83, Information

and Knowledge

Keywords: Knowledge engineering, knowledge

modelling, reasoning, search and rescue, diagnosis

Résumé

Cet article porte sur l’application de méthodes

d’ingénierie des connaissances au domaine de la recher-

che et sauvetage aéronautique au Canada. Nous avons

modélisé au niveau des connaissances le raisonnement

du coordonnateur de recherche lorsqu’il doit effectuer

un diagnostic pour déterminer les routes éventuellement

suivies par un aéronef porté disparu. L’ingénierie des

connaissances nous a permis d’expliciter un modèle de

raisonnement avec l’aide de documents et d’experts du

domaine. Notre démarche s’est effectuée dans le cadre

de modélisation des connaissances CommonKADS. Le

modèle proposé a été validé par des experts du domaine

et opérationnalisé dans un prototype de système à base

de règles. Copyright © 2007 ASAC. Published by John

Wiley & Sons, Ltd.

Mots-clés : ingénierie des connaissances; modélisation

des connaissances; raisonnement; recherche et

sauvetage; diagnostic

KNOWLEDGE ENGINEERING FOR MODELLING REASONING SCHVARTZ ET AL.

Can J Adm Sci

Copyright © 2007 ASAC. Published by John Wiley & Sons, Ltd. 197 24(3), 196–211

Search and Rescue (SAR) is one of the greatest

humanitarian activities in Canada where thousands of

aeronautical, maritime, and ground incidents occur each

year. While the great majority of these incidents are

quickly resolved and people in distress promptly rescued,

dozens of people still die or are never found following

considerable search and rescue efforts. In 2004 for

example, according to the National Search and Rescue

Secretariat,2 there were 683 aeronautical incidents and

5,682 maritime incidents in Canada, resulting in 112

fatalities. Three Joint Rescue Coordination Centres

(JRCC) located in Halifax (Nova Scotia), Trenton

(Ontario), and Victoria (British Columbia) as well as two

maritime rescue subcentres located in St-Jean (New-

foundland) and Quebec (Quebec) are responsible for

coordinating aeronautical and maritime search and rescue

activities in Canada. Civilian employees (from the

Department of Fisheries and Oceans) and military off‌i -

cers (from the Canadian Forces) act as search mission

coordinators in these rescue centres. Coordinators are

responsible for resolving aeronautical or maritime

incidents and for planning subsequent operations to

locate and assist people in distress. Consequently, one of

the responsibilities of the JRCC is to estimate the where-

abouts of a missing aircraft by determining the possibility

area: the area most likely to contain the search object

and where search operations will be conducted (IAMSAR,3

1999). Incidents generally relate to small aircraft in a

recreational aviation context and not to major commer-

cial aviation. In this project, we focused on aircraft

missing overland.

Determining the possibility area for missing aircraft

is a complex task that requires a high level of expertise.

At this time, there are no clearly established, documented

procedures describing the reasoning involved in this task.

The IAMSAR manual is relatively vague on this subject

in situations where the distress position and/or distress

time are unknown. The Canadian Search Area Determi-

nation (CSAD)4 method (Saunders, 1987) def‌i nes a rect-

angular area of 20 nautical miles wide centered on the

route initially planned. This method does not explicitly

take into account plausible routes that the pilot might

have followed. Nonetheless, these routes are a crucial

element for def‌i ning the possibility area. As a matter of

fact, coordinators almost always modify the CSAD

area based on available information. By doing so,

they implicitly take into account a set of plausible

routes.

In the absence of a standardized reasoning method,

coordinators employ heuristic approaches based on their

intuition and experiences. Formulating a structured

method to determine a set of routes possibly followed by

the missing aircraft is highly desirable as it would help

the less experienced coordinators by providing for a

codif‌i cation and sharing of expert knowledge. Further-

more, a well codif‌i ed method is a prerequisite for the

development of a knowledge-based system (KBS) that is

easy to maintain, f‌l exible, and helpful to the coordinator.

Yet, how does one def‌i ne such a structured method when

current practices rely so heavily on heuristics and when

there is little detailed documentation on the subject?

How does one construct a reasoning model so that it can

be used in a software tool? How does one def‌i ne

the knowledge required for the application of this

model?

To answer these questions, we drew upon two dis-

ciplines that focus on knowledge, an essential asset in

any organization: knowledge management and knowl-

edge engineering. Indeed, over the last few years, knowl-

edge management has become an important stake in

companies, gaining signif‌i cant strength (Ermine, 2004).

The general objectives of knowledge management are

based on three key points: knowledge capitalization,

sharing, and creation. Its aim is to improve processes and

performances by using the knowledge of individuals who

carry out these processes (Roy, 2005). Knowledge engi-

neering is a discipline based on “the study of concepts,

methods and techniques used to model and/or acquire

knowledge for systems that perform or assist humans in

carrying out tasks with little or no a priori formalization”

(Charlet, Zacklad, Kassel, & Bourigault, 2000, p. 2).

Knowledge modelling at a conceptual level is useful to

structure and formalize the knowledge necessary for

complex reasoning problems. This knowledge can later

be made widely available in a knowledge-based system.

In general, knowledge- intensive human activities can be

modelled by a limited number of task types: classif‌i ca-

tion, assessment, diagnosis, monitoring, planning, design,

scheduling, prediction, and so forth. As a matter of fact,

a major f‌i nding from our knowledge acquisition phase is

that a possibility area may be explicitly def‌i ned based on

routes presumably followed by the aircraft, where a pre-

sumed route is the result of a multihypothesis diagnosis

pertaining to the sequence of events around the distress

incident.

Our study, conducted in a qualitative research frame-

work, has two objectives. The f‌i rst is to construct a rea-

soning model for a diagnosis task in a context where

there are several types of hypotheses that cannot be easily

corroborated. The second objective is to apply this model

by developing a structured approach to the determination

of routes possibly followed by a missing aircraft. Our

research was conducted using the CommonKADS knowl-

edge engineering method (Schreiber et al., 2000) and was

grounded in knowledge sources including published pro-

cedures, case reports, and domain experts.