The ‘living globe’ that can help drones fly without GPS
Before the advent of GPS, especially at sea, navigation meant finding your position by looking up at the stars. Today, when the Global Positioning System isn’t working—or gets jammed by electronic warfare—drones are learning to do something similar, orienting themselves by looking down at the Earth instead.
This concept underpins a growing wave of efforts to use cameras, sensors, and AI to keep drones “aware” of their surroundings, allowing them to complete their missions or pursue their targets without the use of GPS, or even any communication signal. This week satellite giant Maxar entered the fray with Raptor, a suite of software that can determine a drone’s position on the planet by matching its camera feed with the company’s giant collection of 3D Earth data.
The system began taking shape about a year ago, says Peter Wilczynski, chief product officer at Maxar Intelligence. By then, he says, “it reached a fever pitch of people in defense realizing that all of a sudden, all of these systems that had been designed with GPS would not work in a modern battle space.”
An array of technologies for autonomous navigation and alternative positioning have emerged in recent years, driven by the rise of cheaper and more powerful drones, cameras, sensors, and chips. But even as other companies have tapped 2D and 3D imagery for GPS-free navigation, they weren’t leveraging the abundance of satellite data that Maxar had.
“As one of the leaders in this kind of 3D mapping, especially globally, you can either wait for customers to do that development or you can do it yourself,” says Wilczynski.
Compared to the basemaps used by other vision-based positioning systems, Wilczynski says Raptor’s 3D foundation offers advantages particularly at low altitudes, and during nighttime operations, especially with infrared cameras, and in complex terrain, with an accuracy of around three meters. Maxar’s 3D globe, totaling 125 petabytes in size, now covers 90-million-plus square kilometers, with about three million square kilometers added every month.
The imagery comes from Maxar’s space unit, which operates a constellation of 10 earth observation satellites that provide imagery to defense and intelligence, government, and industry customers. Last year, it added WorldView Legion, a fleet of six high-performing satellites that expanded its ability to more quickly revisit the most rapidly changing areas on Earth.
Acquired in 2022 by private equity firm Advent International for $6.2 billion, the company is a giant in a fast-growing commercial satellite industry, which has been propelled by cheaper flights to low earth orbit on SpaceX rockets, as well as a surge in interest following Russia’s invasion of Ukraine. Some of the earliest public reporting of the Russian military buildup along its border in 2022 came from nonclassified Maxar imagery, and its satellites have provided key intelligence for the country’s offensive and defensive operations.
The importance of that imagery became even clearer earlier this month, when the Trump administration temporarily suspended Ukraine’s access to the Defense Dept.-funded, Maxar-run cloud platform that U.S. agencies and allies use to task satellites and access images. The suspension cut off free imagery access to Ukraine, though it did not apply to paid access to Maxar’s imagery within Ukraine and did not impact the provision of imagery to U.S. allies including France and the U.K.
The war “emphasized the criticality of these technologies,” says Wilczynski, “and especially as Europe starts rearming and starts to think about what does a European continental defense system look like, it really does push a lot of those countries to think much more seriously about space-based reconnaissance technology.”
As satellites criss-cross the Earth far above Ukraine, drones have filled its skies, along with the electronic attacks intended to disable their navigation or communication systems. Some Russian counter-satellite systems have been so powerful that they have degraded the encrypted “M-Code signals” used by the U.S.’s constellation of GPS satellites. To survive the jamming and spoofing, Russian and Ukrainian military units have taken to flying drones attached to their controllers by miles-long fiber-optic cables, controlling them by wire rather than over the airwaves.
Defense officials around the globe are taking notice. The U.S. has put an emphasis on unmanned and autonomous systems, especially fleets and swarms of attritable drones, designed for one-way trips. Ret. Army General Mark Milley, former chairman of the Joint Chiefs of Staff, has said that within the next 10 to 15 years, up to a third of the U.S. military could consist of robotic systems. (U.S. policy still requires a human to pull the trigger.)
Maxar’s GPS-free positioning system is designed to work with existing front-facing drone cameras and on-board inertial navigation units. Performing the real-time image matching on the fly necessitates a GPU, either installed on a computer at a ground station or embedded in the drone itself, which would allow it to operate offline.
In addition to helping provide positioning for vehicles, remote operators working alongside drone controllers can use the software on commercial laptops to extract real-time target ground coordinates from the drones’ full-motion aerial video feeds. As with Maxar’s imagery data, the system is designed to integrate with workflows in map-based command and control systems, including software from Esri, Palantir, Lattice, and SitaWare.
Raptor represents a larger shift for Maxar, from 2D to 3D mapping, and a new business strategy that expands beyond overhead imagery and data. Wilczynski declined to name any Raptor customers, but said the company has already begun conducting operational tests with drone manufacturers “in conflict areas,” and plans to work with defense primes in Europe and the U.S. to upgrade existing non-GPS resilient technologies.
“You can upgrade the existing technology in situ without having to do a whole hardware refresh,” he says.
A push for autonomy without GPS (or maps)
In Ukraine, where the domestic drone industry could produce three million drones this year, autonomy has become a major focus. Government officials have said that more autonomous drones with AI targeting will arrive on the battlefield this year, potentially making way for “real drone swarm uses.”
American startups are also developing software and sensors to help drones fly autonomously, without communication or GPS signals, and find and attack ground targets and other drones. U.S. defense tech hotshot Anduril recently debuted an autonomous drone it tested in Ukraine, while drone maker Red Cat announced a partnership with Palantir to integrate visual navigation software into its Black Widow drones. YCombinator startup Theseus builds a small sensor unit for GPS-denied navigation that has attracted the interest of Special Operations Command.
Four U.S. autonomy specialists—CX2, Swan, Auterion, and KEF Robotics—are also working with Ukrainian drone makers to develop their systems, with endorsement from the Pentagon. Earlier this month, Pittsburgh-based Swan and Arlington, Va.-based Auterion were chosen to compete in a Defense Innovation Unit project called Artemis, in which they will partner with Ukrainian attack-drone makers to develop prototype units that can fly in GPS-denied skies by the end of fiscal year 2025.
Another Pittsburgh startup, KEF Robotics, has formed a joint venture with Kiev-based augmented reality company Sensorama Lab to build its autonomous-navigation systems for drones. The new venture, Blue Arrow, has already won a $50,000 award in a DIU hackathon, and is now preparing to test its plug-and-play software and hardware on the front lines in Ukraine.
Blue Arrow takes two appraoches to navigating and finding targets in GPS-denied environments. Terrain Relative Navigation (TRN), like Maxar’s Raptor, matches ground features with onboard maps, using machine learning to correct for outdated mapping data. Another approach, ideal for shorter distances, called visual inertial odometry (VIO), works without maps: instead, the system relies on the drone’s inertial data and computer vision techniques like optical flow estimation and simultaneous localization and mapping.
“We’ve constrained our algorithms and our software to a point where it doesn’t even need a full processor, so we can use just a very small part of the brain of an airplane,” says Olga Pogoda, a co-founder of Blue Arrow.
Researchers are developing more outlandish ways of giving drones geospatial awareness. Last year, NILEQ, a subsidiary of British missile-maker MBDA, introduced a low-power visual navigation system that relies on neuromorphic cameras, which, rather than processing a whole image, can operate more efficiently by measuring the differences between pixels in a moving image.
The stars can still be useful, too. Last year, engineers at the University of South Australia demonstrated a new, low-cost prototype system that merges vision-based algorithmic computing with celestial triangulation for UAVs flying at night. Accuracy remains a challenge.
Maxar is now developing methods to use 2D satellite imagery and drone video to more quickly update its 3D reference map, enabling more up-to-date mapping, especially in areas of rapid change. Improving accuracy is also a priority, especially at lower altitudes; eventually, Raptor could help replace GPS for ground-based autonomous vehicles, too.
Beyond positioning and navigation, Raptor represents a broader vision at Maxar for a shared, dynamic “living globe,” says Wilczynski, one capable of connecting disparate systems and sensor data in near real-time. That project has been in the works since at least 2015, when Maxar predecessor DigitalGlobe entered into a joint venture with defense manufacturer Saab to produce photorealistic 3D digital elevation models for enterprise and government use. In 2020, Maxar acquired Saab’s stake in that company, Vricon, and it began developing Raptor last year, after it hired Wilczynski, a former executive at data fusion giant Palantir.
“This general GPS-denied problem space really gave us a natural way to fuse together 2D data from drones, 2D data from satellites into a 3D world,” says Wilczynski.
Among its biggest customers for the 3D Earth data is the U.S. Army, which pays Maxar to power its One World Terrain (OWT) program, a global simulation environment for use in military training. The company won the $95 million OWT contract in June 2019, with its most recent extension awarded last year.
“We have this belief that mapping over the next 10 years is going to shift from predominantly 2D to predominantly 3D,” he says.
