Project : Web Portal for environmental health specialists

The creation of web portal in specialized domains (vertical portals) is becoming a necessity for users who want to efficiently access to the informational resources which are available in various domains of interest on the Internet and the corporative intranets, as well as specialized document bases or data bases. We aim at developing a web portal for environmental health professionals. This web portal will offer a variety of functionalities such as indexing, search and summarization of web page contents. This web portal will offer to the professional a unique web page in which he or she will easily find all the useful links for his or her activities. The web portal will be developed using Copernic's Aggregator and Summarizer products integrated with an ontology developed for the domain of environmental health.

Projects/Projets

Project : Literature survey on distributed continual planning

Distributed Continual Planning (DCP) emerged in the few latest years and since that, it is appealing to the interest of many researchers.
DCP implies two main planning characteristics: the distribution of the environment which implies the involvement of several participants and the continuity of the planning process which should remedy to the changes that can occur during the planning or the execution of the plan. The present literature survey aims at clarifying the notion of the DCP as well as at investigating the possibilities to use this DCP to solve certain military problems.

We first present the characteristics and the application of the DCP and clarify the notion of the planning process and particularly in a military context. Because DCP is usually supported by software agents and military people are often brought to organize and reorganize particularly in joint and coalition operations, we explored the characteristics of organizations and the organization within a society of agents. Because we believe that mobile agents can perform planning efficiently within dynamic environments, we present some research efforts based on this idea. We also believe that DCP's progress can take advantage of several recent technological advances. Finally, we present several possible applications of tDCP in the military domain and describe two fictive scenarios which were designed by the military community in order to experiment the efficiency of software agents in solving coalition problems. 

Projects/Projets

Project : Optimization of Models and Graphical Processing of 3D Data in Urban Environments

A research project has been launched at CRDV in 2000 in order to model and visualize urban environments. This project aims at proposing a method and a set of tools which will enable users to manipulate and exploit a complete set of three-dimensional data representing urban environments. Within this project we work on various techniques to optimize models and graphical processing of 3D data representing urban environments (currently the central part of Quebec city). This will be used in a 3D virtual world in which various kinds of agents can move. We use the software Creator and the API Vega of the company MultiGen-Paradigm.

Projects/Projets

Project: User-Oriented Explanation and Argumentation Systems

We conducted an extensive literature review on the characteristics of explanation systems. We also investigate how argument management can be integrated in an explanation system. We study different techniques that can be used to develop an argumentation support system in relation to research in multi-agent systems. Several applications might benefit from this research: decision support system, computer-mediated communication, collaborative reasoning, design rationale.

Projects/Projets

Project: Mobile agents for Command and Control Systems

The objective is to study how using mobile agents may help supporting activities taking place in Command & Control Information
Systems (CCISs). We conducted an exploratory study on the use of stationary and mobile agents in the design of planning systems within
distributed and dynamic environments. The proposed approach provides practical solutions to problems which classical planning
techniques failed to solve efficiently. The approach was illustrated with an example of the aeromedical evacuation domain. The agents
have been developed on top of the multiagent framework Beegent.

We also developed an approach to compare different architectures for interoperable environments. We worked on four different
architectures: three of them use stationary agents and well-known negotiation organizations (use of a broker agent, use of the contract net
protocol), the fourth one uses mobile agents and a meeting infrastructure. We proposed a comparison function based on three main
factors (message type, message size and risk). We also developed a method and a softwre which is used to compare the architectures
considering various scenarios that influence the interactions between the different agents involved in the negotiation process supported
by the various architectures. 

Projects/Projets

Project: Discourse Modeling

The objective is to develop models and tools to understand the pragmatic and semantic content of discourses. One possible outcome is the simulation of conversations between agents (human and/or artificial agents). We are currently studying argumentative texts (such as movie critics) in order to model the argumentative structure of the discourse in relation to a representation of the locutor's domain knowledge and the structure of discursive goals. We want to extend this approach to the analysis of conversations in which agents exchange arguments (political debates, etc.). We are also exploring how misunderstandings may be modeled using a model of the conversation developed by Moulin, Rousseau and Bouzouba during the past five years.

Projects/Projets

Project: Generalization Project

Map generalization is the process of obtaining a map at a given scale (such as 1:100000) using the information contained in a map at a larger scale (such as 1:20000). The idea behind such an approach is to keep the spatial information at a unique scale (such as 1:20000) and derive from it maps at smaller scales. Hence, only one data base is updated. Several projects have been conducted in this research area, most of them providing various sets of specialized algorithms to help cartographers during the generalization process. However, there is no universal approach for map generalization. The objective of the present project is to propose a method and a support tool (in the form of an assistant expert system) that will help cartographers in choosing the appropriate tools for each task of the generalization process.
The proposed method will put the emphasis on the characterization of map users' needs and map producers' objectives. A cognitive study is conducted in order to get appropriate knowledge about the know-how and practices of expert cartographers. We also investigate how techniques based on multiple representations of geographic information can be combined with pure map generalization methods. Starting from a geographic data base in which features are recorded at a given scale, we consider that an initial selection of features has been done based on the identification of the semantic properties required from features to be displayed at the target scale. Applying this selection process, we get a new data base, called intermediate data base, that will need to be transformed in order to generate the map at the target scale.The proposed approach is based on the identification of anomalies (we call them symptoms) which arise when a map is produced
from the intermediate data. The transformation process aims at eliminating these anomalies and at producing a map which respect acceptability criteria.

We developed a methodological framework for map production, including the steps for map generalization based on the formalization of various constraints applying on map design (topology, semantic, legibility, aesthetics). In order to automate part of the symptoms detection process, we propose an algorithm which aims at grouping small neighboring objects on a map using a neural network approach, and more specifically a variation of Kohonen Maps. This algorithm takes into account the sizes of the objects to be regrouped and the fixed polygonal subdivisions of an area. This algorithm could be used to solve certain problems of feature extraction. For example, one immediate application would be to identify high density areas where houses are represented using small individual rectangles and to
replace them by a polygonal area representing the corresponding urban area.

We also investigate the influence of the context on the decisions taken by cartographers while generalizing different categories of data.
The objective is to develop a system that will learn and abstract sequences of generalizations operations from the cartographer's actions in chosen contexts. Usually, map generalization algorithms are uniformely applied on the data set corresponding to the map to be generalized.
This is fine if the different regions of the map have similar characteristics (urban areas, agricultural areas, isolated areas, etc.). However this
is not always the case, and a parameter setting which gives good results in a given area could give a bad cartographic solution in other areas. Based on this observation, we showed the importance of adapting generalization operations to local contexts in the map. We call a local context an area with distinctive characteristics (presence of certain types of entities with a certain density, presence of certain topological properties between those entities, etc.). It is easy for a cartographer to identify local contexts (dense urban areas, sparse urban areas, rural areas, woods and lakes, islands and archipelagos, etc.). We are developing a system which will enable a cartographer to locate on a map displayed on screen certain local contexts (also called prototypocal areas) that s/he finds typical. The program explores the data set and identifies similar areas on the map. Then, we ask the cartographer to perform generalization operations on the prototypocal areas,
setting appropriate parameters for the various generalization algorithms. Then, the system applies the same parameters and algorithms to other areas in the map which are similar to the cartographer's prototypical areas. 

This project is integrated in a larger GEOIDE project "Development of Automated Techniques to Extract, Generalize, and Access Geospatial Information from Hyperspatial Remotely Sensed Data" for which you can download a PDF file, copy of a presentation given by B. Moulin about the Project (Sept 99)

Projects/Projets

Project: Cartographic Interface for the Multi-Dimensional Exploration of Environmental Health Indicators on the World Wide Web

To reduce and monitor health risks of environmental origin requires easy and rapid access to high quality statistics which must be analysed in a useful manner to support decisions and interventions. Several public health organizations wish to acquire a GIS application permitting them to georeference their data. However, GIS are limited for several kinds of tasks such as multi-scale analysis (e.g. local vs. regional analysis), spatial statistics, and handling temporal indices (historical and predictive). Those are usual tasks for an environmental health specialist. Simultaneous with the development of GIS tools, the computer world witnessed the appearance of (1) Data Marts (sort of simplified Data Warehouse, presenting strongly aggregated statistics in a multilevel and multi-theme structure called multidimensional), (2) statistical analysis tools (e.g. SPSS, SAS), (3) tools for data exploration (OLAP or On-Line Analytical Processing), and (4) automatic Data Mining. Each of these tools provide solutions, although in a separate and partial manner, of a part of the above-mentioned problems and only for non spatial information. The main objective of this project is to construct an Intranet-based system to describe and analyse complex phenomena in environmental health, using a simple and rapid cartographic interface based upon the multidimensional approach of Data Marts and OLAP tools.In addition we want to facilitate the temporal and multi-scale management of spatial data by exploiting the multi-dimensional approach used in Data Marts, to navigate through these data rapidly and easily by way of a SOLAP cartographic user interface.

In this project W. Aly and H. Irandoust work on the creation of a tool and a knowledge base in order to index and retrieve documents useful to environmental halth professionals (texts, images, videos) trying to take advantage of a formalization of the semantic content of those documents.

Web-site:http://sirs.scg.ulaval.ca/icemse/