A Simulation Environment for Emergent Properties Heather R. Turner, Susan Stepney, and Fiona A. C. Polack Department of Computer Science, University of York, Heslington, York, YO10 5DD, UK {turner,susan,fiona}@cs.york.ac.uk This research is funded by the EPSRC and BAE Systems as part of a CASE Studentship. Abstract: We propose a multi-layer architecture for simulating emergent properties. This is implemented as a form of cellular automaton at the lowest layer, with mobile processes to represent objects at multiple up- per layers. This architecture supports multiple levels of emergence. 1 Introduction Many definitions of emergence share the theme of existence of distinct levels, between which the emergence occurs. They might correspond to, for example, a change of spatial or temporal scale, be used only to simplify a description of behaviour, or they may be essential in identifying emergent properties. Regardless of the form of the emer- gence, levels always seem to be present, and can lead to a hierarchical structure of emergence. Researchers tend to focus on the generation of emergence in a particular environment, or on sim- ulating a particular natural behaviour. We provide a broader simulation architecture to study emer- gence, and emergence hierarchies as phenomena in general, whilst also permitting specific, con- strained, simulations. 2 Overview of Architecture The architecture is based on a series of commu- nicating layers of abstraction. The lowest layer is implemented as a variant cellular automata (VCA), that need not be strictly finite, regular, and/or de- terministic. This layer encodes environmental in- formation, and captures absolute spatial reference. It forms the substrate of the emergent architecture. In keeping with similar work by Capcarrere [1], we call this the environmental layer. The next layer is constructed from mobile pro- cesses, communicating with other processes in the same layer, the layer below, and the layer above, via mobile channels. We implement these mo- bile processes using features of the occam-π lan- guage [2]. This primary mobile layer uses rules without explicit spatial reference; it relies on rela- tive spatial information transmitted from the lower layer. Additional layers of mobile processes can be added, either corresponding to levels in the emer- gence hierarchy, or purely for implementation con- venience. In their most basic form, the mobile processes simply tag and track emergent features at the lower levels, allowing the emergent features to be treated as objects in their own right at the higher level. But they may additonally encompass some of the logic of the simulation, providing information and instruction to lower level processes. This use of higher level structuring allows us to engineer specific emergent behaviour. The system is designed with higher level objects, then rules are systematically migrated down to lower level objects. By migrating rules to the lowest VCA layer, we can remove the higher layers and achieve a primitive mono-layer system for implementation. 3 Case Studies We are now using our occam-π simulation envi- ronment to study some simple models: a basic 1D model of platelet flow and blood clot formation [3]; emergent properties in a 2D environment, incor- porating rule combination and migration [4]; un- bounded growth of diffusion limited aggregation. References [1] Capcarrere MS: An evolving ontogenetic cellu- lar system for better adaptiveness. BioSystems 76:177–189, 2004. [2] Barnes FRM, Welch PH: Communicating mobile processes. Communicating Process Architectures. IOS Press, 2004. [3] Polack FAC et al.: An architecture for modelling emergence in CA-like systems. In ECAL 2005, LNCS. Springer, 2005. [4] Turner HR, Stepney S: Rule Migration: Exploring a design framework for modelling emergence in CA-like systems. In Workshop on Unconventional Computing, ECAL 2005.