Une approche pour supporter l'analyse qualitative des suites d'actions dans un environnement géographique virtuel et dynamique : l'analyse " What-if " comme exemple

Authors: Haddad, Hedi
Advisor: Moulin, Bernard
Abstract: We propose an approach and a tool based on multi-agent geosimulation techniques in order to support courses of action’s (COAs) “What if” analysis in the context of dynamic geographical environments. We present the characteristics of “What if” thinking as a three-step mental simulation process based on qualitative causal reasoning. We stress humans’ cognition limits of such a process in dynamic geographical contexts and we introduce our research motivations. Then we present our multi-agent geosimulation-based approach and we identify its characteristics. We address next three main problems. The first problem concerns modeling of dynamic geographical phenomena. We stress the limits of existing models and we present our model which is based on the concept of spatio-temporal situations. Particularly, we explain how we define our spatio-temporal situations based on the concept of cognitive archetypes proposed by the linguist J-P. Desclés. The second problem consists in transforming the results of multi-agent geosimulations into a qualitative representation expressed in terms of spatio-temporal situations and represented using the conceptual graphs formalism. We present the different steps required for such a transformation. The third problem concerns causal reasoning about spatio-temporal situations. In order to address this problem, we were inspired by works of causal reasoning research community to identify the constraints that must hold to identify causal relationships between spatio-temporal situations. These constraints are 1) knowledge about causality, 2) temporal causal constraints and 3) spatial causal constraints. These constraints are used to infer causal relationships among the results of multi-agent geosimulations. Finally, we present MAGS-COA, a proof on concept that we implemented in order to evaluate the suitability of our approach as a support to real problem solving. The main contributions of this thesis are: 1- An approach based on multi-agent geosimulation to support COA’s “What if” analysis in the context of virtual geographic environments. 2- The application of a model proposed in the linguistic research community to a problem of interest to spatial reasoning research community. 3- A qualitative model based on cognitive archetypes to model spatio-temporal situations. 4- MAGS-COA, a platform of simulation and qualitative analysis of spatio-temporal situations. 5- An algorithm to identify causal relationships between spatio-temporal situations.
Document Type: Thèse de doctorat
Issue Date: 2009
Open Access Date: 16 April 2018
Permalink: http://hdl.handle.net/20.500.11794/20815
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
SizeFormat 
26160.pdf6.81 MBAdobe PDFView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.