Friday 27 January 2012

The NVIDIA Blog

The NVIDIA Blog

Link to NVIDIA

Using GPUs to Help Modernize U.S. Air Traffic System

Posted: 27 Jan 2012 08:53 AM PST

cover2011_final _V2

Air traffic control is getting an upgrade, and GPU technology is helping the U.S. Federal Aviation Administration's (FAA's) NextGen initiative reach new heights.

Built shortly after World War II, the U.S. air traffic control system is based on a network of radar towers. This means that 50,000 to 60,000 flights a day must fly indirect routes to stay within range of the towers, adding to time spent in the air and wasting jet fuel.

And, as many of us have experienced firsthand, the current system is often not equipped to deal with bad weather or airport closures that can quickly cascade throughout the country.

The CUDA-DAC team at Mosaic ATM
(Left to Right: Chris Brinton, Bill Guo, Bart Gallet)

To modernize U.S. air traffic control, the federal government has launched an ambitious initiative called NextGen. NextGen represents an evolution from a ground-based system of air traffic control to a satellite-based system of air traffic management.

The FAA explains that NextGen “will let pilots know the precise locations of other airplanes around them. That allows more planes in the sky while enhancing the safety of travel.”

Last week, we wrote about Optimal Synthesis and the NextGen-related work the company is doing with NASA in the domain of air traffic flow management. This week, we take a look at Mosaic ATM of Leesburg, Virg.

Bart Gallet, a principal analyst at Mosaic ATM, describes NextGen as “a substantial and long-term change in the management and operation of the national airspace system.”

Gallet and his team are focused on an area known as Dynamic Airspace Configuration (DAC), which will enable air traffic controllers to adapt to user demand in real-time while managing such variables as weather and traffic congestion.

He says, “The design of airspace and traffic flows is currently conducted through manual processes. In order to achieve the vision of NextGen, airspace resources will be allocated to accommodate traffic demand dynamically.”

The challenge is that analyzing thousands of flight paths through hundreds of sectors and calculating optimal airspace configurations with varying weather conditions requires enormous computational resources.

Enter the GPU.

Gallet and his team are experimenting with NVIDIA GPUs, CUDA and parallel computing to accelerate “dynamic airspace decomposition” of the national airspace.

They presented their findings recently at the Digital Avionics Systems Conference, concluding "the parallel implementation for large data sets provides a significant performance increase compared with the original sequentially coded version.”

Does this mean that computers will replace air traffic controllers? Definitely not. While NextGen will leverage the latest GPU technology, Gallet says that humans will always be a critical part of the equation.

“Because of the need for system monitoring and anomaly detection and handling, the human being as a decision maker will always remain a core component of air traffic management,” he says.

“The most significant change in air traffic management in the future will be the integration of the computer with the human decision-making process. Today, computer-based tools are provided to controllers and air traffic managers, but those tools have very little 'intelligence' and what intelligence does exist in these tools is provided to the human in a minimally interactive manner.”

No comments:

Post a Comment