|InterJournal Complex Systems, 1673
|Manuscript Number: |
Submission Date: 2006
|Complex Stellar Dynamics and Galactic Patterns Formation|
In cosmology, confrontation between numerical simulation and data resulting from observation shows to main problems : + results given by the simulation models highly depend on number of elements used. As a realistic number of elements should be around 10^41 for a typical spiral galaxy, there is a problem of computing power. + a shift exists between observation and numerical simulation on spheroidal galactic formation : angular velocity of peripheral stars get by numerical simulation does not match with observation. This paper defines a new approach for cosmological simulation based on complex systems theroy : we use a hierarchical multi-agents system to study the complexity of stellar dynamics. Most cosmological simulations use point-mass particles and global laws. Our approach is different : agents and their interactions are aggregated to differents higher-level agents. At each level of the model the behavior of these higher-level agents emerge from the behavior of individual lower-level agents. Our model uses physicaly-based laws and agent interactions to present stellar structures has the result of self-organisation. By parameterizing these laws and interactions, we obtain a space of possible universes. This space is qualitatively and quantitatively classified in complexity classes. We try to establish a link with Wolfram's classes of complexity (ordered, choatic, complex) and characterize what leads to complex patterns. Nevertheless a strong bond with cosmology is kept by showing the capacity of our model to exibit structures close to thoses of the observable universe. By replacing the universe "as-we-know-it" in a larger picture of universe "as-it-could-be" we try to understand patterns formation and dynamics evolution. In this framework, this paper focus will be given to the hierarchical multi-agents model definition and its validation process.
|Submit referee report/comment|