DARPA wants better forecasts for near-Earth orbit space missions
And it'll use machine learning to create the necessary prediction models.
DARPA, the government's mad science wing, wants "weather" forecasts for outer space like we've got here on terra firma. The idea is that as we're increasingly launching satellites into low-Earth orbit -- not to mention prepping for commercial spaceflight -- we'll need the ability to predict smaller space weather events as well as big ones.
Currently, we can predict things like sun spots, solar winds or coronal mass ejections, which play hell with high-orbit assets. "But we can't make one-hour to 72-hour predictions of local, smaller space environment disturbances close to the Earth, from magnetic substorms to auroral-E, which can interfere with a host of ground and space-based assets," according to Air Force Major Charlton Lewis.
The initiative is called the Space Environment Exploitation, or SEE for short. It'll take some doing to get the project started, though. The military will first have to develop a "holistic space environment physics model," and then dramatically upscale the amount of measurement samples taken from low-Earth orbit.
To address the former, DARPA said it'll use new GPUs and tensor processing units to create new, higher resolution physics models. For the latter, the agency hopes to machine learning as a means to "extend the life of satellite measurements virtually and ... increase the number of satellites virtually" to boost sample sizes. It will also work to upgrade existing ground-based sensors.
For tactical reasons, commanders know what the weather will be like in the theater of operations for the three days surrounding a ground-based mission. SEE's ultimate goal is to have as accurate of prediction models for space missions as we do for terrestrial ones.
Think you have a good idea for DARPA? The agency is hosting a Proposers Day on July 31st, seeking out advice from magnetospeheric, ionospheric and thermospheric physicists and chemists in addition to experts on machine learning and tensor processing. There is a catch, though: Members of the general public (and media) aren't invited to attend.
