An Extensible Agent-Based Framework for Exploring Climate Change Adaptation
ARC DP1093290



The project is supported by an ARC Discovery Grant, and additional funding is being sought.


Project Description:
The goal of this project is to facilitate exploration of possible climate change adaptation (CCA) strategies, using an interactive platform incorporating multiple agent-based simulation modules. The aim is to build up complex simulations by incrementally adding new agent-based models created by members of a large distributed community, interested in the application area. Each module will capture a different aspect of the situation, and could potentially be created independently by people with expertise relating only to that aspect. We are thus developing an extensible, open-source framework which allows individual modules, possibly pre-existing and implemented under different paradigms, to be integrated in a common environment.
Agent-based modelling (ABM), is particularly suited to addressing the complexity of CCA because it allows emergent phenomena to be captured at the macro-level, by modelling autonomous, decision-making entities at the micro-level, then simulating their interaction with each other and the environment. This ABM approach will be enhanced by our proposed inclusion of entities based on the Belief Desire Intention (BDI) agent architecture, which facilitates more complex reasoning agents than are commonly used in ABM modelling. These entities may include complex social organisations or groups as well as individuals. The platform we are using to integrate the modules, and provide an interface for user interaction, is Civilization IV, an existing game engine with a substantial user base. Thus there is the potential for broad uptake and contribution from many diverse sources, which will increase the impact the work is able to have. The extension of ABM to an interactive platform is an innovative idea which builds on the technology developed for games, using it in a manner similar that described as serious games.
The project is a cross-disciplinary one, involving both technical and social science challenges.

The specific aims are:
  • Identification of key concepts and their semantics, for exploring CCA questions via (interactive) agent based simulation.
  • Creation of an extensible framework for modelling and implementing CCA problems at the operational level, using a modular approach.
  • Development of an understanding on how to manage the balance between control and emergence in the simulation.
  • Establishment of a methodology and preliminary toolset for building a new application or extending an existing one.
  • Production of one or more prototypes focussing on a particular CCA question/area

Technical challenges include:
  • Integration of multiple modules affecting (possibly the same aspects of) the situation in different ways.
  • Techniques to support building applications from a library of expert modules.
  • Issues regarding widely differing temporal granularity of modules
  • Scalability of approaches

Participants: Lin Padgham, Fabio Zambetta, Colin Fudge, Alexis Drogoul, Dave Scerri, Sarah Hickmott
Collaborators: Peter Hayes, Darryn McEvoy
Students involved: Yizhaq Yehuda, Ferdinand Gouw

Student Projects

Documentation

Links:

Internal Wiki