Signals from SpaceX Starlink broadband satellites can be used to pinpoint locations on Earth to within 8 meters of accuracy, engineering researchers reported in a new peer-reviewed paper. Their report is part of a growing body of research into using signals from low Earth orbit (LEO) satellites for navigation, similar to how GPS works.
This technology won’t replace your smartphone’s map application any time soon, and this initial experiment apparently required 13 minutes of tracking six Starlink satellites to pinpoint a location on Earth. But researchers were able to achieve the locational feat without any help from SpaceX, and they say the test proves the method could be used for navigation.
“The researchers did not need assistance from SpaceX to use the satellite signals, and they emphasized that they had no access to the actual data being sent through the satellites—only to information related to the satellite’s location and movement,” an Ohio State News article said.
“We eavesdropped on the signal, and then we designed sophisticated algorithms to pinpoint our location, and we showed that it works with great accuracy,” Zak Kassas, the director of CARMEN (Center for Automated Vehicle Research with Multimodal AssurEd Navigation), a US Department of Transportation-funded center at Ohio State University, said in the article. “And even though Starlink wasn’t designed for navigation purposes, we showed that it was possible to learn parts of the system well enough to use it for navigation.”
The research was conducted by Kassas along with Joe Khalife (a postdoctoral fellow at the University of California, Irvine) and Mohammad Neinavaie (a PhD student at UC-Irvine). Kassas is also a UC-Irvine professor and director of the Autonomous Systems Perception, Intelligence, and Navigation (ASPIN) Laboratory, while Khalife and Neinavaie are members of the lab. Their experiment was conducted using an antenna on the UC Irvine campus.
Kassas said his “team has used similar techniques with other low-Earth orbit satellite constellations, but with less accuracy, pinpointing locations within about 23 meters,” according to the Ohio State News article. “The team has also been working with the US Air Force to pinpoint locations of high-altitude aircraft; they were able to come within 5 meters using land-based cellular signals,” Kassas said in the article. GPS provides signals with average errors of less than 1 meter.
The paper is titled The First Carrier Phase Tracking and Positioning Results with Starlink LEO Satellite Signals and was published last week in the journal IEEE Transactions on Aerospace and Electronic Systems. Researchers also presented their findings at an Institute of Navigation conference. Their work was funded by grants from the US Office of Naval Research, the National Science Foundation, and the Department of Transportation.
“Signals of opportunity”
The researchers’ paper said that “various theoretical and experimental studies” have considered the possibility of using “signals of opportunity” from LEO broadband satellites for navigation.
“With SpaceX having launched more than a thousand space vehicles (SVs) into LEO, a renaissance in LEO-based navigation has started,” they wrote. “Signals from LEO SVs are received with higher power compared to medium Earth orbit (MEO) where GNSS [Global Navigation Satellite System] SVs reside. Moreover, LEO SVs are more abundant than GNSS SVs to make up for the reduced footprint, and their signals are spatially and spectrally diverse.”
Another advantage of LEO satellites is that “they do not require additional, costly services or infrastructure from the broadband provider.” But that doesn’t mean the task for researchers was easy. “However, broadband providers do not usually disclose the transmitted signal structure to protect their intellectual property. As such, one would have to dissect LEO SV signals to draw navigation observables,” they wrote.
The summary of the researchers’ conference presentation noted that broadband providers could change their protocols to support navigation. But the researchers argue that their own third-party approach is more viable despite requiring “more sophisticated receiver architectures.”
“[T]ailoring the existing protocols to support navigation capabilities require significant changes to existing infrastructure, the cost of which private companies such as OneWeb, SpaceX, Boeing, and others, which are planning to launch tens of thousands of broadband Internet satellites into LEO, may not be willing to pay,” they wrote. “Moreover, if these companies agree to that additional cost, there will be no guarantees that they would not charge the users for extra navigation services. Under these circumstances, exploiting broadband LEO satellite signals opportunistically, becomes a more viable approach.”
A new algorithm
The researchers previously considered a “cognitive approach to tracking the Doppler frequency of unknown LEO SV signals” but said in their most recent paper that this method “cannot estimate the carrier phase, nor can it be adopted here since it requires knowledge of the period of the beacon within the transmitted signal, which is unknown in the case of Starlink LEO SVs.” To overcome that barrier, they “develop[ed] a carrier phase tracking algorithm for Starlink signals without prior knowledge of their structure.”
The paper said:
Little is known about Starlink downlink signals or their air interface in general, except for the channel frequencies and bandwidths. One cannot readily design a receiver to track Starlink signals with the aforementioned information only as a deeper understanding of the signals is needed. Software-deﬁned radios (SDRs) come in handy in such situations, since they allow one to sample bands of the radio frequency spectrum. However, there are two main challenges for sampling Starlink signals: (i) the signals are transmitted in Ku/Ka-bands, which is beyond the carrier frequencies that most commercial SDRs can support, and (ii) the downlink channel bandwidths can be up to 240 MHz, which also surpasses the capabilities of current commercial SDRs. The first challenge can be resolved by using a mixer/downconverter between the antenna and the SDR. However, the sampling bandwidth can only be as high as the SDR allows. In general, opportunistic navigation frameworks do not require much information from the communication/navigation source (e.g., decoding telemetry or ephemeris data or synchronizing to a certain preamble). Therefore, the aim of the receiver is to exploit enough of the downlink signal to be able [to] produce raw navigation observables (e.g., Doppler and carrier phase).
Tracking six satellites for 800 seconds
During the experiment, “a stationary National Instruments (NI) universal software radio peripheral (USRP) 2945R was equipped with a consumer-grade Ku antenna and low-noise block downconverter (LNB) to receive Starlink signals in the Ku-band,” they wrote. “The sampling bandwidth was set to 2.5 MHz and the carrier frequency was set to 11.325 GHz, which is one of the Starlink downlink frequencies.”
The researchers recorded Starlink signals for 800 seconds, or about 13.3 minutes. “During this period, a total of six Starlink SVs transmitting at 11.325 GHz passed over the receiver, one at a time,” they wrote. Researchers stored samples of the Ku signals “for off-line processing.”
The receiver’s position was estimated using a weighted nonlinear least-squares (WNLS) estimator. The result was 25.9 meters off the real location, but the error dropped to less than 8 meters upon “equipping the receiver with an altimeter (to know its altitude).”
The paper’s conclusion said:
This letter showed the first carrier phase tracking and positioning results with real Starlink LEO SV signals. A model of a Starlink SV’s transmitted signal was formulated, and an adaptive KF (Kalman filter]-based carrier phase tracking algorithm was developed to track the Starlink signal. Experimental results showed carrier phase tracking of six Starlink LEO SVs over a period of approximately 800 seconds. The resulting positioning performance was: 7.7 m 2–D error when the receiver’s altitude is known, and 25.9 m 2–D error and 33.5 m 3–D error when the receiver’s altitude is unknown.
SpaceX has launched over 1,700 satellites but plans to eventually launch tens of thousands in order to expand the broadband service’s capacity and availability. Those additional satellites would presumably also make it easier to build navigational systems of the type envisioned in the new research.
We contacted the researchers today to ask about the prospects of using Starlink satellites to get location results in something closer to real time and about how they envision LEO-based systems being used for navigation when the methods and technology are more advanced. We’ll update this article if we get a response.
Update: Kassas responded to us and provided more information on the experiment, noting in an email that “we waited 800 seconds in order to get signals from six satellites, since we’re not able yet to see six satellites above us simultaneously.” That will change as SpaceX launches more satellites. “We’re preparing another experiment for real-time position estimation in which we’ll use four Starlink satellite signals simultaneously above us,” Kassas said.
Eventually, SpaceX’s plan to launch tens of thousands of satellites “will allow for both real-time navigation and a much higher level of accuracy than what we achieved so far,” he said. In the long run, Kassas sees this as “a standalone navigation system” instead of one that merely complements GPS. “GPS signals are readily jammable and spoofable and are not reliable/usable in many environments (e.g., deep urban canyons, indoors, and under tree canopy),” he said. “We’re hoping that LEO satellites will provide an alternative resilient and accurate navigation system if/when GPS signals aren’t available or are compromised.”