The COGIT laboratory of the French National Mapping Agency made research on generalisation to automate map making from geo data base. The solving method is based on Multi-agent paradigm and owns a large set of spatial analytical tools. These results should allow to characterise urban space at different levels and to study their morphometric dynamic. To reach this goal we should adapt and implement the vector agent principles and tools on our Open Source platform named GeOxygène and to enrich it with various methods conceived at the LIV and LSIIT laboratories during the FoDoMust project (ACI 'Masse de données' 200'-2007). In order to characterise urban space from remote sensing data, both laboratories have formalized geographical entities at various scales by means of 1/ a dictionary containing samples of identified objects and 2/ one ontology that describes the identifies phenomena. A set of segmentation and identification algorithms based on low level information (radiometry, texture) complete the proposed urban formalization. The experiments on the urban objects formalization at several scales would be reused in this project and adapted to the context of vector databases. The operational objective of this research project is to propose an Opensource platform for the analysis and the simulation of urban growth. This simulation can help to better describe the urban dynamic at a local level. More precisely, this project aims to ensure the transfer and the integration of analysis tools of the COGIT and the Strasbourg university laboratories into an open source GIS. This GIS will be based on the open source GeOxygene platform deposited by the COGIT laboratory in 2005. The project also aims to create a set of urban characterisation and simulation modules in vector mode which incorporate constraints related to the spatial environment and the public policy. At least the platform will be enriched by a friendly learning module adapted to the specific case of urban data. The learning tools will be used to enrich the ontology, the urban characterisation and to detect specific urban growth mechanism thanks to the existence of multi-temporal urban data sets. The project is composed of 4 scientific tasks and one organisational one (T0). Task 0 : Project management AIm : to organise technical meetings, to ensure the furniture of the different deliverables (report and implementation code) in time, to send the required information to the ANR, to share the results between partners and to organise an international workshop at the end of the project. Task 1: Develop GeOpenSim Aim : to conceive an Open Source urban simulated module on top of GeOxygène. This module will be based on vector agents that represent urban information depicted at different levels of details. The module will integrate analytical methods conceived in task 2 and a friendly learning module conceived in task 3. The simulated module will be evaluated and improved thanks to the results given by task 4. The modelling will include the representation of different kinds of agents, their aims, their behaviours and their interactions. Task 2: Formalisation and construction of urban objects. Aim: to propose an ontology of urban entities as well as a set of methods necessary for their detection, their characterisation and the identification of their spatial dynamic. The urban objects will be described at the micro and meso level (group level). The main detection and characterisation methods will be coded, tested and improved Task 3: Ontology improvement and conception of a learning module to characterise urban objects. Aim: Adapting classical learning methods for the detection and characterisation urban objects and to facilitate the use of such methods for non expert users. Then improving the ontology proposed in task 2 by means of the proposed learning tools applied on data base specifications and on the data themselves. Task 4: Validation by application study Aim: to test the module GeOpenSim by means of specific cases of urban growth analysis. This study will be done with urban research laboratories. This application will enlighten 1/ the appropriateness of the proposed agent modelling 2/ the completeness of the urban analytical tools, 3/ the flexibility of the system to integrate new and specific urban growth mechanism.