Representation of Interactions in a Multi-Level Multi-Agent Model for Cartography Constraint Solving

Abstract

International audienceThe objective of cartographic generalisation is to simplify geographic data in order to create legible maps when scale decreases. It often requires to reason at different levels of abstraction (e.g. a building, a city). To automate this process, Multi-Agent approaches have been used since several years, by modelling map objects (e.g. buildings) as autonomous entities trying to solve constraints through appropriate transformations. Yet, those approaches are not able to deal with all situations that appear between cartographic objects in a map. Indeed, though a map intrinsically involves objects that belong to several description, scale or organisation levels, there is no explicit multi-level representation in agent-based cartographic models. Thus we assume that the use of a multi-level multi-agent model would improve the automated generalisation process. Especially, the PADAWAN model is a multi-agent model offering multi-level capabilities which match quite well the requirements for the multi-level organisation of cartographic objects. In this paper, we expose how we use this model on the one hand, to reify multi-level relations between cartographic agents, and on the other hand, to represent the constraints and the actions proposed to solve them, as interactions between the agents

An Agent-Based, Multilevel Welfare Assessment Model for Encompassing Assignment and Matching Problems

“…The multilevel agent-based meta-models [13,16,6,17] took a growing importance within MAS in recent years. Indeed, from their origin in the field of simulation, they have spread to other areas, such as distributed constraint solving [12]. Thus, we think that they can also benefit to the types of problems we have discussed above, especially by introducing a capability to represent intermediary stages between "atomic" individuals and the overall system.…”

“…In order to implement these features, the formalism we chose is a multilevel agentbased simulation meta-model called "PADAWAN" [16]. It has already been successfully adapted for handling distributed constraint solving in the context of cartographic generalisation [12]. Besides, in the context of allocation or matching problems, some simplifying assumptions can be made.…”

Multi-agent Multi-level Cartographic Generalisation in CartAGen

“…An agent will evaluate its constraints and choose an interaction to execute. More details on the agent behaviour are given in [5].…”

“…[1;2;3;4]. A work in progress [5] aims at enhancing multi-level representation in agent-oriented in order to deal with more complex relation between objects. The proposed model is based on the PADAWAN model [6], initially designed for multi-level simulation, which has been adapted to perform constraint problem solving dedicated to the specific cartographic generalisation application case.…”