AI-based Data Compression Techniques to Help Unclog Space Communications

Article By : George Leopold

Lossless data compression and optical communications could create bigger pipes for sending and receiving satellite imagery.

Data compression algorithms and optical communication links could help unclog bandwidth bottlenecks for transmitting ultra-high-resolution imagery along with big data from the growing number of probes exploring our solar system.

Among the efforts are a NASA project launching this summer that will demonstrate emerging laser communications capabilities. On the receiving end, commercial space startup Metaspectral is advancing AI-based data compression techniques that promise to reduce file sizes to as little as 30 percent of the originals. While shrinking the size of downlinks, the lossless compression technique preserves key image data that can be used later by analysts. Those huge data sets can also be used to advance training of AI models.

The data management startup based in Vancouver, BC, specializes in real-time analytics applied to hyperspectral imagery, that is, visible to infrared. This week, Metaspectral announced its acceptance into a Los Angeles-based technology accelerator. The incubator is operated by the Starburst Aerospace Accelerator and the University of California at Los Angeles.

Participation in the accelerator has led to contacts with key technology agencies such as the Defense Advanced Research Projects Agency, said Metaspectral CEO Francis Doumet.

For current terrestrial applications, devices are overwhelming bandwidth capacity. Hence, “bandwidth is always playing catchup,” Doumet, noted in an interview.

The situation is similar for space applications, complicated by the reality that most satellites have a roughly eight-hour connectivity window each week to transmit data back to Earth.

Metespectral’s claims its lossless data compression approach is able reduce file sizes while consuming about 10 percent of available computing resources. The result is a halving of data transmission costs while bringing data down to Earth faster and preserving high-resolution data for analysis on the ground using neural networks and far greater computing resources.

That capability addresses the limits encountered in attempting to process image data at the ultimate edge: space.

“Edge processing is only good as long as you know what you are looking for at the time the image is captured,” Doumet said. That makes edge processing suitable for agriculture and weather applications, but unsuited for intelligence and surveillance applications like detecting the plume of a missile launch.

Metaspectral touts its approach as delivering high-resolution archival data to ground stations, enabling intelligence analysts to find the proverbial needle in a haystack.

Elsewhere, efforts are underway to demonstrate optical communications techniques that could provide a much bigger pipe for sending and receiving satellite data.

In late June, NASA will launch its Laser Communications Relay Demonstration (LCRD) payload as part of a Defense Department satellite. Once in geosynchronous orbit, LCRD will test laser communications capabilities with ground stations in California and Hawaii.

Click image to enlarge.

Later tests will seek to relay data from the International Space Station. An onboard terminal to be delivered to ISS next year will receive science data from ISS experiments and instruments, then transfer data to LCRD at 1.2 gigabits per second. LCRD will then transmit data files to ground stations at the same rate, NASA said.

A key hurdle for optical communications is weather conditions on Earth. Doumet of Metaspectral correctly points out that laser communications are vulnerable to weather. “If there’s a storm, it won’t work,” he said. “It has to be perfect conditions.”

In response, LCRD’s chief engineer notes that NASA is deploying atmospheric monitoring equipment to a pair of optical ground stations and is also developing algorithms to help NASA predict atmospheric conditions for optical transmissions between spacecraft and ground. “One of the reasons NASA refers to LCRD as a technology demonstration is we still have a lot to learn with regards to atmospheric prediction,” Bernard Edwards, LCRD co-investigator, noted in an email.

The space agency notes that laser communications will enable up to 100 times more data transmitted back to Earth than current radio frequency systems.

LCRD was initially demonstrated from lunar orbit in 2013, downlinking data via a laser signal at a rate of 622 Mb/s

If and when bandwidth is increased for future space communications, “We’re going to find more applications to stuff that pipe with data,” Doumet added.

This article was originally published on EE Times.

George Leopold has written about science and technology from Washington, D.C., since 1986. Besides EE Times, Leopold’s work has appeared in The New York Times, New Scientist, and other publications. He resides in Reston, Va.

Leave a comment