Toward Self-Generalizing Objects and On-the-Fly Map Generalization Mamane Nouri Sabo, Yvan Be ´dard, Bernard Moulin, and Eveline Bernier Centre for Research in Geomatics / Universite´Laval / Que´bec / QC / Canada Abstract Map generalization is a complex task that sometimes requires human intervention. In order to support such a process on the fly, we propose a generalization approach based on self-generalizing objects (SGOs) that encapsulate geometric patterns (forms common to several cartographic features), generalization algorithms, and spatial integrity constraints. During a database enrichment process, an SGO is created and associated with a cartographic feature or a group of features. Each SGO created is then transformed into a software agent (SGO agent) in a multi-agent on-the-fly map- generalization system. SGO agents are equipped with behaviours that enable them to coordinate the generalization process. This article presents the concept of the SGO and two prototypes developed to support this approach: a prototype for the creation of SGOs and another for the on-the-fly map generalization (which uses the created SGOs). Keywords: self-generalizing object (SGO), on-the-fly map generalization, geometric pattern, database enrichment Re ´sume ´ La ge ´ne ´ralisation cartographique est un processus complexe qui demande parfois l’intervention humaine. Afin de supporter un tel processus a ` la vole ´e, nous proposons une approche de ge ´ne ´ralisation qui se base sur les SGO (objets auto- ge ´ne ´ralisants ou self-generalizing objects) qui encapsulent a ` la fois des patrons ge ´ome ´triques (qui sont des formes communes a ` plusieurs objets cartographiques), des algorithmes de ge ´ne ´ralisation et des contraintes d’inte ´grite ´ spatiales. Lors d’un processus d’enrichissement de la base de donne ´es, un SGO est cre ´e ´ et associe ´a ` chaque objet ou groupe d’objets de la carte. Les SGO cre ´e ´s sont ensuite automatiquement transforme ´s en agents logiciels (agents SGO) dans un syste `me multi-agent de ge ´ne ´ralisation a ` la vole ´e. Les agents SGO sont dote ´s de comportements qui leurs permettent de coordonner le processus de ge ´ne ´ralisation. Dans cet article, nous pre ´senterons le concept des SGO et les prototypes (un prototype pour la cre ´ation et l’enrichissement des SGO et un autre pour la ge ´ne ´ralisation a ` la vole ´e, utilisant les SGO cre ´e ´s) de ´veloppe ´s pour supporter cette approche. Mots cle ´s : objets auto-ge ´ne ´ralisants (SGO), la ge ´ne ´ralisation cartographique a ` la vole ´e, patron ge ´ome ´trique, enrichissement des bases de donne ´es 1. Introduction In the past, map production was the cartographer’s responsibility, because of the specific skills required to create maps. Users received maps created in advance and generally produced in a series. However, the technological developments of the last decade have generated a new type of medium for spatial data dissemination, the World Wide Web. In addition, there are new applications such as Web mapping and spatial database querying. With the emergence of these new applications, cartographic data has become more acces- sible to the general public and can be better adapted to users’ needs. cartographica (volume 43, issue 3), pp. 155–173 doi:10.3138/carto.43.3.155 155