MAGIC - Multi-Agent Geographic simulations based on Interoperable Components
Oct 2002 - Jun 2006
Multi-agent simulations enable the representation of situations in which individuals have complex and different behaviours, to study the global consequences that arise as emergent structures resulting from the interaction processes. Agent simulations
involve interaction and communication between the actors in the process and facilitate negotiation processes among them. Moreover, individuals in the simulation can be considered as hierarchical structures, which contemplate fine-grain entities belonging to lower level structures.
There are several advantages to using this type of approach while studying spatially-mapped phenomena and it is clear that the implementation of simulation models using multi-agent systems would benefit from the access to common operators available in Geographic Information Systems (GIS) packages.
Recent developments into commercial GIS packages are characterised by the use of object-oriented (OO) technology. New GIS applications can thus represent and manipulate spatial features as objects, whose state and behaviour can be customised, extended or completely reformulated. The characteristics of agents are also primarily OO. This means that it is currently possible to develop OO architectures involving entities endowed with proactive behaviour ("agentified") as well as spatially-aware. Integration needs felt by scientists trying to use their spatial data as input for their computational models can now be handled transparently.
In spite of these developments, it is common practice in agent simulation architectures to have an import facility that reads data from one specific GIS format.
It is still not clear what is the best approach in integrating these two types of technology, with the aim of creating Spatial agent simulations.
This project aims to study the possibilities for including spatial analysis tools (characteristic of GIS) in agent-based simulations. The implementation of necessary GIS operators in existing successful agent architectures will be considered. The limitations of this approach will be analysed in comparison with that of including agent facilities in a GIS package (either by implementation or through interoperability).
The results of the project will be a comparison of the two methodologies and a proposal for a framework to build a spatial agent-based simulation tool.
The resulting structure will serve to implement a simulation of environmental impact problems for Gestilamas, the Portuguese professional association of natural stone industry.
The primary aim of the project is the development of tools that will enable the creation of adaptive
agent-based simulations where each agent can improve its performance through learning and embedded spatial analysis capabilities. Current work in this area has already demonstrated the added-value given by agent-based systems in simulating complex systems where the spatial dimension is of major importance.
To achieve this aim it will be necessary to develop a methodology for using the multi-agents systems approach while endowing agents with a geographic analysis capacity. This objective should be achieved through interoperability between agent implementation platforms and geographic information systems.
The final aim is the implementation of a framework, which will follow the developed methodology and will enable the creation of agent simulation where agents can use spatial analysis operations to determine their future actions in the simulation.
The implementation of the case-study tests the potential of the tools developed and constitutes one of the sub objectives of the project .
