Massively parallel processing for dynamic airspace configuration

The design of airspace and traffic flows through airspace is currently conducted through manual processes. In order to achieve the vision of the Next Generation Air Transportation System (NextGen), airspace resources will be allocated to accommodate traffic demand dynamically. The computational loads required for wide-area dynamic airspace configuration (DAC) [1] processing is a significant impediment for transitioning this concept into a pertinent and usable technology. Customized computing hardware could be applied to address some of these issues, but they are costly and more than often the software does not scale graciously as new technologies become available. In this research effort, we address an algorithm and computing architecture for the automated partitioning of airspace into sectors that is based on the technology of Graphic Processing Units (GPU) and the Compute Unified Device Architecture (CUDA). In this paper we provide a parallel implementation of a selection of Dynamic Density (DD) factors, elements of the objective function that is used in DAC optimization. We will show that our parallel approach provides a significant overall performance gain compared with the original sequential implementation of the selected DD factors.