PADI-Simul: an agent-based geosimulation software supporting the design of geographic spaces

Abstract In the PADI Project we explore how geosimulation techniques based on agent technology can support designers when creating geographic spaces. As a case study we work on the design of parks. We propose to simulate the way a geographic space could be used by its future users, mainly in terms of navigation and occupation of space, and to display the usage patterns emerging from the simulations. To this end, we developed a system composed of two main modules: PADI-Design and PADI-Simul . In this paper we present an overview of PADI-Design which implements some important functionalities of a CAAD tool for park design. Then, we present PADI-Simul , a multi-agent system which can be used to model and simulate the behaviors of hundreds of agents of various types moving in a geographic space. Agents are modeled using three main categories of knowledge structures: navigation knowledge (space apprehension map, spatial orientation map, an agents' location map), self knowledge (stable states, dynamic states and the associated functions) and behavioral knowledge (behavior hierarchy). We discuss these knowledge structures, describe the agent's navigation capabilities and present a trace function which is used to assess the simulation results. We also show how some of PADI-Simul 's functionalities have been integrated in a module which automatically creates a first sketch of paths in the park and suggests them to the designer at the end of the design phase. Finally, we compare our system to some other geosimulation systems ( RBSim2, PEDFLOW , STREETS ) and provide an overview of future extensions of the PADI Project.