Collective Adaptive Behaviour in Space
Many collective adaptive systems consist of a large number of spatially distributed heterogeneous entities, and hence space must be considered in any formal approach to designing and managing such systems. The aim of this work package is to consider representations of space, design process algebra constructs to express space and finally to consider space in finding model parameters. The work package has three main tasks:
- The first task covers Scalable Representations of Space. It considers different representations of space from the literature, and the relationships between these representations, and it will produce a formal framework for quantitatively relating representations in terms of abstraction and approximation. For example, continuous movement in space can be abstracted to a set of patches, each representing a portion of space. Since efficient evaluation is important, a goal is to understand which representations are appropriate for different CAS and this process will be driven by the case studies of the project.
- The second task focuses on Stochastic process algebras for spatial aspects, where process algebraic features will be designed for the representations of space identified in the first task, using the framework to ensure new constructs treat space in a uniform fashion, as far as this is possible. These new constructs will be given appropriate stochastic and fluid semantics, exploiting the various approximation techniques already developed. Furthermore, abstraction based on approximation between the new constructs will be investigated. Different dialects of the basic language may be designed for CAS scenarios with differing spatial requirements.
- In the last task, Parameter and model fitting from spatial data, the role of space when fitting parameters of models will be considered. Explicit representations of space both add complexity to the process and constrain the parameter space. Spatial information may allow local fitting of parameters followed by integration of parameters from different localities, and novel procedures will be developed to take this into account.
Contact: Vashti Galpin – School of Informatics, University of Edinburgh