Thinking GRID for Electromagnetic Simulation of oversized structures
Leaders: Hervé Aubert (MINC), Thierry Monteil (MRS), Patricia Stolf (ASTRE)
One of the challenges facing Information Technology industries today is that of creat- ing smaller and smaller embedded systems that include more and more intelligence. Such smart systems carry complex software architectures as well as hardware platforms that are connected to sophisticated communication infrastructures. Designing these systems need robust methodologies to reduce their development cycle and prototyping phase.
In this context, it is primordial to optimize the concept of the communication hard- ware. The complexity of such systems makes them difficult to optimize because of the increasing number of unknown parameters. Contemporary tools and methods will soon become obsolete to handle future issues.
Furthermore, communication objects will be more and more integrated in environ- ments filled with various metallic and dielectric structures of different sizes with respect to the carrying wavelengths.
The communication item will require a realistic dimension to establish proper propa- gation within multiple obstacles and realize good connections. For example, transmission of signals from airborne sensors or wheel antennas emitting data like tire pressure to a collector located inside a vehicle will suffer interference from the overwhelming metallic structure. These perturbations must be taken into account to correctly predict power yield from antenna to receptor.
Broadly speaking, the challenge concerns the use of algorithms and electromagnetic theory to solve difficulties caused by electromagnetic diffraction by very large objects in proximity of antennae. A rigorous approach requires numerical resolution of large systems with super calculators and grids.
The main objective is to develop a new 2D and 3D approach to simulate the electro- magnetic behaviour of large structures (planes, cars, buildings, etc).
A new method will be proposed combining TLM and SCT. This global electromag- netic hybrid simulation of large objects is to be validated theoretically by creating an efficient algorithm. Multithreading and MPI will be implemented. Two approaches will be used: MPI at grid level (using mpich-G2, gridMPI) and/or multi-parametric execution to explore more solutions. Autonomous deployment and intelligent use of administration software will benefit from the specificities of grid computing.
Dynamic response to exterior or interior events (such as breakdown or performance loss of an array of machines) will be sought by automatic execution of new simulations in self adapting network set-ups.
This challenge will address coupling between grid computing and electromagnetic simulation with autonomic policies, a subject little investigated worldwide up to now.