Parallel simulation for the analysis of population dynamics in high-performance environments

The aim of this project is to design and develop a simulator that studies the demographic evolution of population. The simulator reproduces the internal interactions among individuals. Due to the nature of this kind of social processes the chosen methodology has been agent-based modeling. Applying agent-based modeling has produced interesting results in some research areas such as economics, archeology, psychology or sociology.

There is a long-term aim of this research, which is to create a tool that will help in making decisions on family planning policies. Due to the inherent complexity of social dynamics this project takes advantage of parallel programming and high performance computing in order to deal with simulation of large-scale population. This is why other institutions have collaborated in this project: Lancaster University, where a prototype of micro-based simulation for population dynamics was developed and Barcelona Supercomputing Center, focused on exploring the use of Marenostrum supercomputer and other high performance facilities to contribute to social sciences research.

Specific objectives:

• To design and implement a scalable parallel simulation to run individual-level demographic models for policy analysis.
• To design and implement a graphical modeling language tool that acts as a user interface between model developers (non-computer specialists) and the simulator.
• To evaluate the effectiveness of the graphical specification language tool using a set of real case studies by real users.
• To evaluate the performance of the simulator using a set of artificial population with different characteristics and different cognitive and social behaviors-

C++, MPI (Message Passing Interface), and the libraries of parallel and distributed simulation μsik and libsynk were used for the developing of this project.