Section: Overall Objectives

Develop a strategy of software production

Although the software production is not the main objective of the team, we aim to assemble models within virtual ecosystems (with the objective that it can replace or guide real experiments) in the four coming years. Co-developments of simulation software will be required within INRIA or outside. Our working plan is in three steps:

1. In a first step, we shall develop a Scilab or Matlab toolbox within the team.

2. Simulations of IBM or PDE require significant efforts, in terms of computer implementation and numerical methods.

   The main problem in the implementation of IBM's is the size of the population, particularly for applications in microbiology where the size of populations of bacteria can be very large. In agent-based models (ABM), population sizes are smaller and each individual features sophisticated behavior, while in IBM, population sizes are usually larger but with limited individual activities. The population size impacts both the execution time and the memory size, but the main bottleneck is the execution time because of the communication between the individuals. The idea will be to propose, through an object-oriented approach, data structures that will limit this communication. From the hardware point of view, grid computing could lower execution time but only on a limited range. New techniques exploiting the architecture of graphics processing units (GPU), that are specifically adapted to IBM simulations, seems very promising. These activities will have to be developed in association with specialists for our most ambitious developments: MOMAT already cited but also possibly the EPI OASIS and researchers from the LIRMM.

3. For the design and development of user-friendly and graphical interfaces that need to be easily accessible by biologists and bioprocess engineers, we shall look for the help of service companies specialized in agronomy and biotechnology applications (such as ITK Company).