The GPS Directorate wants industry to more quickly develop innovative user equipment that integrates both the modernized GPS signals and signals from international constellations like Galileo.

“In the future,” said the organization’s new director, Col. Steve Whitney, “it’s going to be important that our industry partners and the Directorate investigate ways to pull in these new signals — and that includes some of the non-GPS signals — into our user equipment.”

The GPS Directorate wants industry to more quickly develop innovative user equipment that integrates both the modernized GPS signals and signals from international constellations like Galileo.

“In the future,” said the organization’s new director, Col. Steve Whitney, “it’s going to be important that our industry partners and the Directorate investigate ways to pull in these new signals — and that includes some of the non-GPS signals — into our user equipment.”

Whitney’s interest in incorporating more GNSS signals comes from his experience as senior materiel leader in the Directorate’s User Equipment Division where he worked with the military’s largest Foreign Military Sales (FMS) program.

“We have over 57 partner nations that buy GPS user equipment,” Whitney told Inside GNSS, “and so we have lots of discussions with them about that. That’s really where I think the international engagement in terms of incorporating new signals comes in.”

Some of those discussions likely take place in an 11-nation research development test and evaluation (RDT&E) group formed by the United States and its allies. The group, said Michael Sanjume, deputy chief of the User Equipment Division, works together on advanced technology positioning, navigation, and timing (PNT) applications.

“One of the exploratory project arrangements is looking at the feasibility of open signal GNSS receivers,” Sanjume told Inside GNSS. They are “looking at the possibilities of how we could do it, what level of integration it would be done at —things like that,” he said, although Sanjume stressed that any new application would not appear until later generations of Military GPS User Equipment (MGUE).

Whitney said it made sense to incorporate non-GPS signals. “I don’t see why we shouldn’t go that way and couldn’t go that way,” he said. “I think that’s the smart way to go about it. I think that in the future, though, it’s not necessarily going to be something where the government mandates how to do things. The government is going to turn to industry and ask for their best and brightest solutions. That’s kind of what we’re trying to establish in the commercial market strategy in MGUE.”

Industry Option
That approach, which aims to increase competition within the MGUE program, has been under development for some time.

“Rather than go through complicated contract competition and then downselect to a few vendors, the Directorate is going to establish criteria for MGUE security certification, cryptography, and signal protection and compatibility,” said Whitney, and then leave it industry to do the integration. He said there would be conversations with companies along the way to let them know what the directorate thinks is important.

“I have got the requirements, the specifics that have to be delivered — but we can also talk to them about ‘Hey, these type of areas should be looked at, and you should find a way to do that.’ I think you’ll find that industry, as we discovered in MGUE, is more than capable of responding on a faster timeframe than if we go through a formal program. That’s really the heart of that commercial market- based strategy —invigorating industry and getting them trained and going.”

The GPS Directorate took a concrete step in that direction when it posted a notice August 3 on the Fed Biz Ops website <fbo.gov> announcing an industry meeting on September 1–2. The meeting, about the development of a next-generation cryptographic processor for MGUE, was a directorate initiative specifically aimed at attracting more companies, including small businesses, to compete in the military GPS market.

Whitney said he believes competition has already inspired the MGUE program’s current contractors — Rockwell, L3/Interstate Electronics Corporation, and Raytheon — to begin working to integrate international signals into their devices. “I know from personal discussions all three of them have had talks about it,” he said.

Although he couldn’t say exactly what the current state of play was, “I know that they are making strategic partnerships in how they go about doing that”

“This is now going to be about market share,” Whitney told Inside GNSS. “You’re trying to bring the industry forces to bear. And that’s really what it is. They’re trying to become the first one to get there with the greatest

The post Military Receivers: Air Force Wants Foreign GNSS Capability appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

1996 soccer game in the Midwest, (Rick Dikeman image)

Nouméa ground station after the flood

A pencil and a coffee cup show the size of NASA’s teeny tiny PhoneSat

Bonus Hotspot: Naro Tartaruga AUV

Pacific lamprey spawning (photo by Jeremy Monroe, Fresh Waters Illustrated)

“Return of the Bucentaurn to the Molo on Ascension Day”, by (Giovanni Antonio Canal) Canaletto

The U.S. Naval Observatory Alternate Master Clock at 2nd Space Operations Squadron, Schriever AFB in Colorado. This photo was taken in January, 2006 during the addition of a leap second. The USNO master clocks control GPS timing. They are accurate to within one second every 20 million years (Satellites are so picky! Humans, on the other hand, just want to know if we’re too late for lunch) USAF photo by A1C Jason Ridder.

Detail of Compass/ BeiDou2 system diagram

Hotspot 6: Beluga A300 600ST

1. I LOVE MY JOB BUT…
Bakersfield, California USA

1. I LOVE MY JOB BUT…
Bakersfield, California USA
√ A California sales manager for a wire transfer firm downloaded a required app called “Xora,” to her smart phone. It used GPS to monitor an employee’s location, arrival and departure time and said its software “captures a tremendous amount of powerful data, which can be pulled on demand with detailed reports or automated and emailed to you.” The plaintiff was fine with that — during work hours. As it turned out, the monitoring never stopped. Employees were required to leave their phones on 24/7 in case a client called. In her complaint, the sales manager said her supervisor “bragged that he knew how fast she had been driving at specific moments ever since she had installed the app.” The plaintiff said this was illegal and removed the app. Shortly thereafter, she got fired. She filed suit in the Bakersfield Superior Court for invasion of privacy and labor code violations, among other things. The case is still pending. Xora, now known as Click Software, is a mobile workforce management company with many clients. We hope its powerful capabilities include an automatic kill switch — or at least some lullabies after 10 p.m.

2. TO THE RESCUE
On the Mediterranean
√ Refugees from civic breakdown in Syria, Eritrea, Nigeria, Somalia, Sudan and elsewhere are pouring across the Mediterranean towards asylum in Europe and 2,500 have died this year so far. A Louisiana millionaire couple who lost their house to Hurricane Katrina spent US$ 8 million to save as many refugees as possible from a watery death. Christophe Catrambone is the founder of the Migrant Offshore Aid Station (MOAS), a charitable foundation. He and a 20-member crew use two GPS-equipped Schiebel Camcopter S-100 unmanned aerial vehicles lent by the company, a base ship, and rigid hulled inflatable boats to search for migrants in trouble and assist them. Doctors Without Borders provides medical help. The unmanned air vehicles launch from the ship and send back imagery, 24 hours a day, even in bad weather. The foundation website claims 10,000 lives have been saved since the project began one year ago.

3. TREMORS
Hurricane Sandy’s path, 2012
√ Seismologists know that hurricanes shake the ground, not just the air, creating mini-quakes that are hard to place. In a study using the North American Earthscope project seismic monitoring site where GPS reference receivers provide complementary measurement systems for resolving strain-rate, researchers found they could locate these minor tremors more precisely. The scientists followed Hurricane Sandy for six days during its 2012 assault on the east coast of the United States. They ran through 117,370 different combinations of 485 seismic sensors for each hour of data and found out the signals came from the eye of the hurricane. The scientists said this might become a practical way to predict increases in storm intensity using the seismic network to remotely monitor air pressure changes inside the storm. Their study appeared in Solid Earth, the journal of geophysical research, this August.

4. BEWARE OF HUMAN
Ontario, Canada
√ HitchBOT, the traveling robot that depended on humans to give it a ride, successfully crossed Canada and explored Belgium and Germany before being vandalized in the USA on August 1. Designed by two researchers from Ryerson and McMasters universities in Ontario, Canada, HitchBOT used GPS to locate itself, sophisticated voice recognition and patterning integrated with social media to communicate, and a playful body made up of odds and ends to inspire affection. Its creators wanted to know how humans might relate to the fast-arriving future when “smart” machines will appear to be sentient. Thousands of humans treated HitchBOT as if it were alive — for good or ill. “Sometimes bad things happen to good robots,” HitchBOT said in its last message. Next time, Hitch, hire bodyguards.

The post GNSS Hotspots | September 2015 appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Return to main article: John Raquet: A Family Affair

COMPASS POINTS

Engineering specialties

Navigation; navigation by signals of opportunity; sensor fusion/integrated avionics; GPS; navigation warfare; software development; stochastic estimation.

Favorite equation

e^iπ = –1

Return to main article: John Raquet: A Family Affair

COMPASS POINTS

Engineering specialties

Navigation; navigation by signals of opportunity; sensor fusion/integrated avionics; GPS; navigation warfare; software development; stochastic estimation.

Favorite equation

e^iπ = –1

(i is an imaginary number representing the square root of negative 1)

Raquet says this special case of Euler’s formula is interesting because it involves many of the basic concepts of mathematics: an exponential function involving e, the imaginary number i, the value of π, a negative sign, and the value of 1. “And yes,” he adds, “this equation has been on our dining room whiteboard from time to time!”

GNSS event that most signifies that GNSS has ‘arrived’

“About 10 years ago,” says Raquet, “I was checking out at Wal-Mart, and the Wal-Mart cashier started explaining to me some of the details of how GPS works, including how GPS sends a data stream to the receiver so it can calculate satellite positions. He had no idea that I knew anything about GPS and was just making small talk. At that point, I knew that GNSS had arrived.”

What popular notions about GNSS most annoy you?

Raquet says he’s annoyed by the popular notion that if you have a GPS receiver, then the Air Force can track you. “I used to give an annual lecture on GPS to a group of high school students. At the beginning of this lecture, I would ask them whether or not the government can track them if they have a GPS receiver on, and usually about two thirds would say that it can. I would then attempt to explain that a GPS receiver is only a receiver, and that there is no way for the GPS system to know anything about who is using the system or where they are. After this long explanation, about half of the students still thought that the government was tracking you if you had a GPS receiver!”

As a consumer, what GNSS product, application, or engineering innovation would you most like to see?

Raquet says he’d love to see a GNSS-controlled Frisbee that he could throw and that would always come right back to him. “I think that knowing where the Frisbee is, using GNSS, is the easy part of this problem,” he says. “The bigger challenge is figuring out how to automatically steer a Frisbee in flight!”

Mentors

His first mentor was his father, Charles, a NASA engineer who “inspired within me a love for figuring out how things work,” and his second, Raquet’s Ph.D. advisor at the University of Calgary, Dr. Gérard Lachapelle.

Patents

Haker, M., and J. Raquet, “Global Navigation Satellite System Signal Decomposition and Parameterization Algorithm,” US Patent no. 9,025,640. Issued May 5, 2015 (A method to break down a GNSS signal into a set of direct and multipath components)

Morrison, J., and J. Raquet, and M. Veth, “Coded Aperture Aided Navigation and Geolocation System,” US Patent no. 8,577,538, Issued November 5, 2013

Martin, R., and J. Velotta, and J. Raquet, “Navigation and Position Determination with a Multicarrier Modulation Signal of Opportunity,” US Patent no. 8,072,383, Issued December 6, 2011

The post John Raquet’s Compass Points appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Two Galileo satellites were carried into space today (September 10, 2015) with a successful launch at 10:08 p.m. (EDT) ((02:08 GMT on September 11) from the Soyuz Launch Complex at the European spaceport in French Guiana.

The satellites, named Alba and Oriana after Spanish and Russian children, respectively, who won a European Commission Galileo drawing competition, are the fifth and sixth full operational capability (FOC) spacecraft to be launched by the European GNSS program. Developed and built by OHB System AG, the spacecraft were launched on board a Russian Soyuz ST-B rocket.

All the Soyuz stages performed as planned, with the Fregat upper stage releasing the satellites in opposite directions in their target orbit at roughly 23,500 kilometers in altitude, about 3 hours and 48 minutes after liftoff.

“The deployment of Europe’s Galileo system is rapidly gathering pace,” said Jan Woerner, director general of the European Space Agency (ESA).

Two further Galileo satellites are still scheduled for launch by end of this year. These satellites have completed testing at ESA’s ESTEC technical center in Noordwijk, the Netherlands, with another two satellites now undergoing their own test campaigns.

Next year the deployment of the Galileo system will be able to be accelerated by the arrival of a new line of specially customized Ariane 5 rockets that will double, from two to four, the number of satellites that can be launched at one time.

The post ESA Launches Two Galileo Satellites appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

But some of the limelight needs to fall a little further from center stage — out there where the Indian Regional Navigation Satellite System (IRNSS) and Japan’s Quasi-Zenith Satellite System (QZSS) are not waiting idly in the wings.

When China joined the GNSS club in 2007, it turned a satnav triumvirate into a quartet.

But some of the limelight needs to fall a little further from center stage — out there where the Indian Regional Navigation Satellite System (IRNSS) and Japan’s Quasi-Zenith Satellite System (QZSS) are not waiting idly in the wings.

If all goes as scheduled — and both India and Japan are moving ahead with increased confidence — the regional systems will substantially enrich the world’s positioning, navigation, and timing (PNT) resources. IRNSS should be completed this year, with seven satellites on orbit. QZSS will have four satellites up by 2018 and seven by the end of 2023.

It’s all too easy to conflate regional systems with GNSS augmentation systems, such as WAAS, EGNOS, or Russia’s SDCM — systems that largely replicate and improve the quality of the underlying GNSS service features. (Inside GNSS probably contributes to this lack of differentiation between the two types of systems by having a single “Augmentation/Region” button on its website.)

But India and Japan have their own augmentation systems in GAGAN (GPS-Aided Geo-Augmented Navigation) and MSAS (MTSAT Satellite-based Augmentation System), respectively. QZSS and IRNSS, in contrast, are essentially autonomous systems that literally operate outside the orbits and frequencies of the world’s four GNSS systems.

These are not trivial considerations.

As last year’s disappearance of Flight MH370 reminded us, the Indian Ocean region in which IRNSS operates is large and largely unchartered. So, too, the Western Pacific surveyed by QZSS. Having expanded GNSS resources available there will bring benefits to nations well beyond those providing the regional services.

Broadly speaking, regional systems do augment GNSS services, it’s true. But they also move beyond them, bringing their own distinctive characteristics and technological opportunities.

QZSS’s L-band Experimental (LEX) signal — to be renamed L6 when QZSS becomes fully operational — is designed to transmit correction messages that enable positioning, navigation, and timing applications that require centimeter-level accuracy.

The L6 signal (at 1278.75 MHz) will incorporate Geo++ State Space Representation (SSR) technology in which relevant ionospheric effects are estimated in real time using reference network observations, making these suitable for single-frequency GNSS receivers.

Similarly, IRNSS transmits a unique hybrid data structure at L5 (1164.45–1188.45 MHz) that provides grid-based ionospheric corrections for single-frequency users, as well as a ranging signal on S-band (2483.5–2500 MHz) — a relative novelty for GNSS systems.

The QZSS program also attempts to address another challenge: the need for indoor PNT capabilities. The ground-based IMES or the Indoor Messaging System uses pseudorandom noise codes (PRNs), operates in the L1 frequency, and has message formats compatible with QZSS and GPS.

Beyond their technical distinctiveness, regional systems ensure a relatively greater control over strategic national assets and the critical infrastructures and applications that rely upon them. This, in turn, strengthens the sovereignty of nations that possess and operate them.

Both Japan and India are full members of the International Committee on GNSS Providers Forum, along with the “Big Four”: the United States, China, Russia, and the European Union.

They have earned their places at the table.

The post Let Us Now Praise appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

At one time, GPS was expected to supplant a wide range of navigation technologies in the world’s positioning, navigation, and timing (PNT) portfolio. But an unexpected thing happened along the way.

As GPS — and more recently, GNSS — moved from concept, to development, to reality, its vulnerabilities became more apparent, along with its remarkable qualities of accessibility, accuracy, and affordability. Interference to low-power GNSS signals, fear of spoofing attacks and intentional jamming, diminished performance in some operational environments — these and other factors have led policy makers and some user communities to reconsider their expectation of GPS universality and, instead, to seek alternative PNT (APNT) resources. In the United States, a 2004 presidential directive mandated creation of a backup system for GPS to ensure the uninterrupted provision of PNT services.

We called on Sherman Lo help us understand what is at stake in the search for APNT. Lo is a senior research engineer in the GPS Laboratory at Stanford University, where he earned his Ph.D. in aeronautics and astronautics. For the last several years, he has served as an investigator for the Federal Aviation Administration’s evaluation of APNT alternatives.

IGM: What are the leading functions/values/features being looked for in alternative PNT systems?

LO: This topic is a source of much discussion in the APNT community as there is not general agreement in several areas. I think the features needed from APNT are generally agreed upon: robustness, integrity/authenticity, and accuracy (timing or positioning).

However, a debate exists about what those things actually entail. For example, while it is clear that APNT must handle a GNSS outage, no consensus has emerged regarding the length of time and extent in area for which an outage must be managed. I think the differences of opinion come from several factors. First, there have been few major incidents of GNSS denial or spoofing. Second, the uses of and threats to PNT are evolving.

I think that the various stakeholders have different time horizons in mind for APNT. The PNT targets and threats in 2035 will be different than those today or 2025. The challenge with PNT (APNT or otherwise) infrastructure is that it takes time to build out, equip, and modify. It is not like consumer devices or software where we can count on rapid turnover or updates. We must think and plan for the future, sometimes far into the future, and get consensus on what is needed.

For example, FAA APNT currently needs to only provide about one-nautical-mile position accuracy. Hence, distance measuring equipment (DME) — either DME/DME or DME/DME with inertial reference unit (IRU) avionics — is sufficient. However, in the future airspace (2025 and later) as envisioned under the FAA Next Generation Air Transportation System (NextGen), reliance on GNSS and GNSS level performance will be greater and some APNT capabilities will need to improve — to perhaps 0.3 to 0.5 nautical mile accuracy along with better low-altitude coverage.

I suspect that similar issues face other industries such as telecommunications, where one-microsecond timing is sufficient for today but we may be talking about 100 nanoseconds or better in the near future.

IGM: The Federal Aviation Administration (FAA) has emerged as a leader in the search for APNT. Are other regulated carriers —rail, maritime, commercial transport, etc. — substantially engaged in this activity and, if so, how?

LO: The FAA has shown great leadership in seeking robust and redundant navigation capability. As for other regulated carriers, the work with which I am most familiar is that of the General Lighthouse Authorities of the United Kingdom and Ireland (GLA) with maritime navigation. The GLA has been at the forefront developing eLoran for maritime APNT and are leading international standardization efforts on eLoran for maritime applications.

However, I feel that other civil agencies are becoming increasingly cognizant of the need for APNT. The Department of Homeland Security (DHS) has engaged in numerous activities to scope out the need for APNT and examining alternatives. It has held interference exercises to help users understand these threats and develop mitigations awareness and supported development of enhanced Loran (eLoran) through its cooperative agreements.

IGM: What are the leading candidate technologies for providing APNT service?

LO: Right now, it seems as if each agency has its own leading candidates. For FAA, DME [distance measuring equipment] will be a basis for APNT in the near term. DHS is looking into eLoran, DARPA is developing chip scale inertials, and so on.

This diversity of solutions is due to many reasons — each group has systems to meet specific special needs of their mission. On one hand, heterogeneous solutions are a good thing, as they make it more challenging for attackers. But on the other hand, from a cost and a security perspective, I feel that it is not good to have too many different solutions. Having fewer solutions will allow us to focus on the security of each solution rather than just counting on security through diversity.

I believe that, as different APNT technologies mature, there will be a convergence of solutions with a handful of trusted systems. We track the development of robust alternatives by other groups, as they may offer significant benefits to our FAA APNT efforts. eLoran, should it further develop in the United States, may provide a source of robust time synchronization for APNT. The availability of low-cost, high-accuracy inertial would be very useful for APNT and reduce the requirements on ground infrastructure.

IGM: What is the state of play for APNT in other nations and what kinds of relations/interactions are there between their efforts and those in the United States?

LO: Several other nations and groups have expressed interest in US FAA APNT activities. Eurocontrol has talked to us about their APNT efforts and plans to investigated it within the next set of Single European Sky ATM Research (SESAR) activities. SESAR is the European airspace modernization plan similar to the US NextGen. We have and continue to work with the German Aerospace Center (DLR) and National Cheng Kung University in Taiwan on various APNT research within their airspaces. We interact regularly with these groups. DLR is a participant with US and European APNT activities. I think FAA leadership on APNT has led many other groups to look more closely at this issue.

Beyond aviation, I think the other major area of APNT is the ongoing eLoran effort in South Korea, the United Kingdom, and the United States. Korea is planning to modernize their existing stations to eLoran and build three new stations. This eLoran is based on the know-how developed in US/UK efforts.

IGM: Is APNT a significant concern for consumers — the people with smartphones, PNDs, in-vehicle navigation systems, etc. — and, if so, how should those concerns be addressed?

LO: I absolutely think that APNT will be a significant concern for consumers for all those items that you mentioned. Positioning is cheap and I think in the future devices will integrate GNSS chips for even minor benefits (for example, GNSS in cameras and laptops).

As for how to address the need for APNT, it depends on the product category and it characteristics. For me, two characteristics come to the forefront. The first is the likelihood and effect on safety of losing PNT. In this dimension, I also think about the likelihood of loss — are there any incentives for jamming or spoofing the GNSS signal in this application. The second characteristic is how quickly would we add new PNT systems should the need arise.

Smartphones seem to be on one end of the spectrum in terms of these two characteristics. First, for these devices the GNSS degradation (interference, spoofing, or otherwise) is generally a nuisance but not a safety event. Second, and most important, smartphones are technology items for which people refresh the technology very quickly, and so new technology can quickly be fielded to address PNT problems if needed.

Other consumer devices (unmanned aerial vehicles, robot lawn mowers) may be more challenging.

The post Alternative PNT appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

John Raquet’s Compass Points

With an imposing 6’2” physique and a disarming grin, John Raquet rises above the crowd. To colleagues he’s a top-flight engineer and university professor, and director of the Air Force Institute of Technology (AFIT) Autonomy and Navigation Technology Center. But he is also a former all-star basketball player, a preacher, sometime soccer coach, former military officer, and, most definitely, a family man.

John Raquet’s Compass Points

With an imposing 6’2” physique and a disarming grin, John Raquet rises above the crowd. To colleagues he’s a top-flight engineer and university professor, and director of the Air Force Institute of Technology (AFIT) Autonomy and Navigation Technology Center. But he is also a former all-star basketball player, a preacher, sometime soccer coach, former military officer, and, most definitely, a family man.

And boy what a family: wife Cindy plus eight kids, who have managed, among other things, to keep him in wrestling shape.

“Thinking back to when I was a Ph.D. student at the University of Calgary over a three-year period, I remember at times coming home after a frustrating day in my student office and finding the children happy to see me and ready for a good friendly wrestle on the floor,” says Raquet.

“They really didn’t care if I was struggling in a class or couldn’t get a navigation computer program to compile. Playing with them on the floor after a difficult day was an incredible boost, helping me to keep in mind that some things are more important than others,” he continues. “So, rather than being a distraction from my career, I feel like my children have actually helped me to keep things in perspective.”

In the Beginning. . . .
Raquet grew up in the Cleveland, Ohio, area, in a town called North Olmsted. In high school, he worked at a garden center and was pretty serious about basketball, making an all-Cleveland all-star team his senior year.

His mother, Sharon, was a nurse. His father, Charles Raquet, spent his entire career as an engineer at NASA Lewis, now NASA Glenn, in Cleveland.

Charles Raquet completed a Ph.D. in physics from Carnegie Tech, now Carnegie Mellon, when John was about six months old.

Working for NASA, Charles Raquet spent much of his time developing antenna technology for communication satellites. One of the highlights of John Raquet’s undergraduate university days was having an internship at the same NASA facility where his father worked, testing out arcjet engines for spacecraft in 1988.

“Growing up, I had always liked visiting my father’s office and his lab, so it was very exciting to actually be working at the same NASA facility,” John recalls.

By 1989, Raquet had a fine undergraduate career at the U.S. Air Force Academy under his belt and a bachelor of science in astronautical engineering in his pocket, and he was active military. His next move was a master of science program in aeronautical/astronautical engineering at the Massachusetts Institute of Technology. There he worked to develop an innovative, fuel-efficient, collision-avoidance and docking system for the Space Shuttle.

But there was a major development in his personal life as well. While still in graduate school at MIT, he met Cindy, who was working as a recent Wellesley College graduate in a small high-tech company in Boston. She was one of only a few non-engineers.

Cindy has remained a true partner and devoted traveling companion to John throughout the years. “She has attended probably 15 Institute of Navigation [ION] GNSS conferences with me,” he says, “and even though she’s not a navigation engineer, she knows lots of people who attend, because each place that we’ve lived has involved organizations that are active in the ION.”

After receiving his master’s degree, Raquet was assigned to what is now the 746th Test Squadron at Holloman Air Force Base, near Alamogordo, New Mexico.

He was originally told that he was going to be a flight test engineer, and he was very excited about the opportunity to fly as part of his job, but when he arrived, Holloman leaders discovered that Raquet had a masters degree and decided to put him in the reference systems development element.

“This was one of several times in my life where I look back and am glad that I didn’t get what I thought I wanted,” Raquet says, because it put him in the middle of an emerging technology, the Global Positioning System, that was beginning to transform the world of positioning, navigation, and timing.

“In the reference systems element, we were starting to use GPS as an input to our flight reference systems, which were used to generate truth data when testing navigation systems, mostly INS [inertial navigation] at that time,” he says. “During the four years I was there, we went from not using GPS, to using GPS pseudoranges, to pseudorange DGPS [differential GPS], to incorporating a full integer-resolved, carrier-phase DGPS solution into our reference system.”

So, it turned out that working with reference systems allowed Raquet to “cut his teeth” on GNSS and ultimately set the direction for the rest of his career.

Unforgettable
Raquet says he “first got truly excited about GNSS” when Dr. Gérard Lachapelle came to Holloman AFB in the early 1990s to teach a course on carrier-phase integer ambiguity resolution.

“To this day, it still amazes me that, using carrier-phase DGPS, a receiver can tell you where you are within a centimeter, using signals from satellites 20,000 kilometers away and traveling at several kilometers/second.”

In the GPS class that he teaches today, Raquet has all of the students submit a work of “GPS Art” sketched by plotting the position of an GNSS antenna moved in whatever shape they want, using carrier-phase DGPS for positioning. “I suspect that this is a part of the course that they will never forget,” he says, “long after all of the details of the course have faded from memory.”

From New Mexico, it was on to Canada and back to school, this time as a doctoral student in geomatics engineering at the University of Calgary — Lachapelle’s stamping ground — from 1995 to 1998. There Raquet pioneered a new method for network-based precise differential GPS positioning that is currently deployed throughout the world.

And he remained busy on the home front: “Cindy and I started that program with two young children, and left with four — so, there are two Canadians in my family!”

Full Circle
Today, John Raquet is a professor of electrical engineering and Director of the Autonomy and Navigation Technology (ANT) Center at AFIT. Located at Wright Patterson AFB, near Dayton, Ohio, his office is not so very far from where he grew up playing basketball in North Olmsted.

Back in Ohio, Raquet appears to have achieved that much-sought but rarely found work-family balance.

“I absolutely love being a professor and running a research center,” he says. “One of the best parts of my job is getting to know each of my research students as I work with them over an extended period of time. I also have really enjoyed attempting to develop an engaging, fun culture among the ANT Center staff.”

And fun is something he knows about, keeping eight children, ages 8 through 23 — four boys and four girls — entertained and yet on track: traveling, playing whiffle ball in the back yard, hiking, camping, and on and on.

“We recently all took a trip to the Grand Canyon and back in our 15-passenger van, which was a blast!” says Dad.

One of the family highlights was the opportunity for them all to live in Finland for six months in 2010, when Raquet was AFIT’s first-ever Fulbright Scholar at the Tampere University of Technology.

Raquet found the Finnish posting both professionally and personally enriching. “Living in Europe enabled us to show our children quite a few historical sites that we never thought they would be able to see, at least while living in our own home.”

In the midst of this rich family life, however, Raquet has still managed to get a little work done.

In recent years, ANT has developed the Sensor Processing for Inertial Dynamics Error Reduction (SPIDER) filter framework to alleviate some of the issues inherent with specialized navigation post-processing software.

“We got tired having students struggling to reuse sensor integration software, to the point that they would often start over again from scratch,” Raquet explains. “To fix this, we developed the SPIDER filter frame-work, which has a number of nice features that enable the students to add in their own sensor model with-out affecting other sensor models that may already exist. This has significantly improved our ability to reuse this software.”

A familiar face at conferences and events, Raquet has many awards and honors and well more than 100 published articles and conference presentations to his credit as well as a long line of mentored students. And one still has the feeling that he’s just getting started.

“At the Center, we continue to do lots of work in navigation using non- GNSS means,” Raquet explains, “which is almost always more difficult and less accurate than GNSS, when it’s available.”

Recently, ANT has been investigating the use of magnetic field variations to determine absolute position, by comparing the measured variations to a map, demonstrating this capability indoors, in a ground vehicle, and in an aircraft.

“We have been spoiled by GNSS, and have come to depend on it,” Raquet says. “So, it is important to have alternative navigation signals to fall back on, if and when GNSS is not available.”

Imparting Perspective
Raquet has worked for the Air Force for his entire career, initially as an active duty military member and now as a government civilian. He was active duty from 1989 through 2003, which included his first five years as a professor, and then converted to civilian status.

Raquet decided to get out of the military at the 14-year point, six years before he would have reached full retirement, partly so that he could continue to be a professor, which would not have been possible if he had stayed in uniform.

And, as do many parents, he takes a part of his work home with him.

“Like my father, I like to talk with my children about how things work,” Raquet says. “After a number of drawings literally on the back of a napkin, we finally broke down and purchased a whiteboard which was installed in our dining room; so, now it’s much more convenient to make a quick drawing of something.”

Given that the couple — “mostly Cindy,” Raquet insists — have home-schooled all their children, a whiteboard seems a natural furnishing for their dining room.

And that choice seems to have paid off.

“We have three in university — two studying music, and one electrical engineering,” Raquet says. “We really don’t know yet what the other children will study, although my 10-year-old has let us know that he is just biding his time until he can build robots for a living!”

Music is another family activity, both listening and playing. Most of the children play one or two instruments, including, thus far, the harp, violin, two violas, French horn, flute, trombone, and piano. “

“We also enjoy breaking out into spontaneous song from time to time,” he adds.

John and Cindy enjoy music as well, although they spend more time teaching it and driving around to lessons than actually playing anything.

As a man of faith, Raquet is very active in the Arbor Church, a reformed Baptist congregation, as a pastor and, along with Cindy, as a counselor. Together they have provided premarital counseling for a number of couples, and, looking ahead to their 25th anniversary in January, the Raquets would seem to be good models for giving marital advice.

Raquet offers quite a bit of instruction for fellow church members and says he really enjoys preaching things other than engineering.

“It’s amazing how often engineering concepts provide good analogies for spiritual concepts,” he says, “although sometimes I have to pass on the best engineering analogies, because only engineers would really get it!”

The post John Raquet: A Family Affair appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

Ubiquitous location-aware mobile devices, mainly GPS-enabled smartphones, have led to a boom in location-based services (LBS), which have been revolutionizing businesses and lifestyles. Common uses of LBSs include asset tracking, location-based advertising, emergency roadside service, turn-by-turn navigation, and real-time traffic & road information sharing.

Ubiquitous location-aware mobile devices, mainly GPS-enabled smartphones, have led to a boom in location-based services (LBS), which have been revolutionizing businesses and lifestyles. Common uses of LBSs include asset tracking, location-based advertising, emergency roadside service, turn-by-turn navigation, and real-time traffic & road information sharing.

A major category in LBSs is proximity-based services, which allow users to search for friends or other points of interest around them. Examples of proximity-based mobile social networking include Apple’s “Find My Friends,” Facebook’s “Nearby Friends,” Tencent’s “WeChat,” Momo, and Nearby.

In Find My Friends, a user can see the locations of his friends and get notified when his friends are nearby. In WeChat, a user can find nearby users by the following two ways:

• Shake — A user shakes the phone, and the app will find other WeChat users who are also shaking at the moment locally and around the world. Then the user has an opportunity to message them and make new friends.
• Look Around — Look Around is like Shake without the shaking. The app simply finds other WeChat users who have been recently in the user’s vicinity.

To use an LBS, a user usually has to send his exact location to the service provider, sporadically or frequently. The contextual information attached to user locations may, however, also reveal the users’ habits, interests, activities, health status, and political and religious affiliations. The high level of intrusion and privacy threats associated has made many users reluctant to opt into LBSs. So, designing practical and effective privacy-preserving proximity-detection schemes would reassure users who have concerns about maintaining their privacy.

Past work on location privacy has explored quite a few approaches, such as anonymization, obfuscation, and adding dummies. A common concept underlying most of these approaches is to “degrade information in a controlled way before releasing it,” as summarized in the paper by B. Hoh and M. Gruteser listed in the Additional Resources section near the end of this article.

This article proposes a new approach to preserving location privacy in proximity-based services. Our approach makes use of the location information inside a GPS receiver. The key difference from previous work is that rather than obtaining accurate location information and then degrading it, we extract privacy-preserving location information directly from an intermediate step in GPS location estimation.

Our private proximity detection scheme presented in this article is designed for location-based “friending” applications in a global social network. A user shares his untagged range measurement, which is derived from the user’s GPS range measurements, with the server. The server can efficiently detect if any two users are within a threshold distance of each other. However, it is computationally intensive for the server to infer each user’s exact location from the untagged range measurement. Our approach can be used independently or together with other approaches, such as obfuscation, to provide a higher level of privacy protection.

This article describes how untagged range measurements preserve location privacy and a very efficient matching algorithm for proximity detection. It also evaluates the proximity detection performance through a theoretical analysis and field experiments. The evaluation results demonstrate the efficacy and robustness of our scheme.

Previous Work in the Field
When a large number of people are using a “friending” app, a centralized server that detects proximity between each pair of users greatly reduces the communication costs. In this scenario, the goal would be to enable the server to perform proximity detection without leaking user locations to the server. The server should be able to infer as little information about user location as possible from the data it receives.

A possible approach is privacy-preserving test described in the article by A. Narayanan et alia, which is based on the location tag initially studied by the publications by D. Qiu et alia listed in Additional Resources. With proper location tags, location proximity can be reduced to measuring the similarity between two sets of tags. Narayanan et alia suggested deriving tags from surrounding environment including WiFi traffic and access point identifiers, GSM signals, and GPS signals.

The untagged range measurement proposed in this article is closely related to the location tag. A major difference is that the location tag still requires certain types of cryptography to work. The ElGamal encryption suggested by A. Narayan et alia requires much more computational resources on the user end and server side than does our method proposed in this article.

Private Proximity Testing Using Untagged GPS Range Measurements
Traditionally, GPS range measurements are tagged with satellite pseudo-random noise (PRN) codes, in order to identify the specific GPS satellites from which the ranges are measured. The novelty of our approach is based on untagged range measurements, a vector of GPS range measurements without pseudo-random noise (PRN) code designations. Suppose user i shares his untagged range measurements

r_i = [r_i⁽¹⁾, r_i⁽²⁾, . . . , r_i^(K_i)]^T          (2)

where K_i is the number of satellites visible to this user. The range measurement made to the satellite k, r_i^(k) does not include the receiver clock bias, for all k = 1, . . . , K_i, as the receiver clock bias can be easily calculated and removed beforehand. Furthermore, we require r_i to be a sorted vector in ascending order, i.e.,

r_i⁽¹⁾ ≤ r_i⁽²⁾ ≤ . . . ≤ r_i^(K_i).

Location Privacy Protection. When the range measurements are not designated with PRN numbers, if the adversary has no knowledge of satellite orbits, the K range measurements can be seen as an ordered selection from the L satellites in the whole constellation. Therefore, the search space is K-per-mutations of L.

Usually, we have L ≈ 30 and K ≈ 10, and thus the size of search space is L!/(L – K)! ≈ 1.1 x 10¹⁴.
Even though a server may use the knowledge of satellite orbits to reduce the search space, a large number of permutations still remain to be searched. Intensive computation discourages an adversary from inferring a user’s actual location from untagged range measurements, especially in a large social network.

The untagged range measurements can also confuse an adversary. First, untagged range measurements seen at two or more distant locations may happen to be similar. These events are categorized as distant false alarms in a later section on “Proximity Detection Performance Analysis.” Second, a user can add dummy measurements to his sorted vector so that multiple locations exist at which the untagged range measurements will be a subset of the sorted vector. In this article, we assume no dummy measurements added to sorted vectors.

Furthermore, the untagged range measurements are ephemeral. The satellite-to-user range is decreasing when the satellite is approaching and is increasing when the satellite is leaving. The range change rate varies with satellite elevation. For satellites at the zenith, the range rate is close to zero. For low-elevation satellites, the range rate can be as high as ±930 meters per second. Therefore, untagged range measurements are valid for

Equation (3) (see inset photo, above right, for equations)

If we choose t = 10 kilometers, then untagged range measurements are valid for approximately 10 seconds.

Proximity Detection. Consider two users at locations x₁ and x₂, as shown in Figure 1. Without loss of generality, assume user 2 is to the north of user 1. Suppose a GPS satellite is visible to both users, and the elevation and azimuth of the GPS satellite seen by user 1 are α and β. When the two users are nearby, the distance between the two users is much shorter than the distance to the satellite. Thus, we have the approximation

|r₂ − r₁| ≈ ||x₂ − x₁||₂|cos α cos β|.          (4)

Define a threshold distance t > 0. The two users are deemed “nearby” if ||x₂ − x₁||₂ ≤ t. Therefore, a necessary condition for two users to be nearby is

|r₂ − r₁| ≤ t|cos α cos β| ≤ t.          (5)

In our scheme, the server has to do blind matching of range measurements because the users do not designate PRN numbers to range measurements in order to protect their privacy. We formulate this “blind matching” problem in an optimization framework and propose an algorithm to solve it.

Blind Matching As an Optimization Problem. Let ⊂ denote the subset relation between two sorted vectors. We write x ⊂ y if each element in x also belongs to y. Let card(x) denote the cardinality, i.e., number of elements, of a vector (or a set) x. The proximity detection problem can be formulated as the following optimization problem:

maximize c,
subject to c = card(q₁) = card(q₂),          (6)
          q₁ ∈ r₁,
q₂ ∈ r₂,
||q₁ − q₂||_∞ ≤ t,

where the infinity norm

||[u₁, . . . , u_n]^T||_∞ = max {|u1|, . . . , |u_n|}.

The optimization problem maximizes c, the number of matched range measurements. The decision whether the two users are nearby depends on c, card(r₁), and card(r₂). In this article, we use a very simple criterion: two users are decided to be nearby if the match ratio

Equation (7)

where the decision threshold ζ, 0 ≤ ζ ≤ 1, is selected to achieve certain detection error performance.

Efficient Blind Matching Algorithm. The optimization problem (6) is similar to the longest common subsequence (LCS) problem. Dynamic programming is often used to solve the LCS problem efficiently. Here we borrow this idea to solve the optimization problem.

Let r^(k) denote the kth element in the sorted vector r, and let r[n] denote the vector of the first n elements, i.e., r[n] = [r⁽¹⁾, r⁽²⁾, . . . , r⁽ⁿ⁾]^T. Let c(r₁[k₁], r₂[k₂]) denote the maximum number of matched range measurements between r₁[k₁] and r₂[k₂]. We then have the following recursive property:

Equation (8)

This algorithm achieves the worst-case time complexity of O(card(r₁) + card(r₂)) and the space complexity of O(1).
The algorithm only involves addition and comparison, two of the fastest operations on most CPUs. Therefore, our proximity detection algorithm is very efficient.

Once we obtain c using the previously described algorithm, we then use Equation (7) to determine if the two users are in proximity.

Proximity Detection Performance Analysis
As a statistical hypothesis test, private proximity detection has a probability of making two types of errors: false alarm and missed detection.

Equation (9)

Equation (10)

Equation (11)

We focus our theoretical performance analysis on PFA for two reasons. First, PDE is dominated by PFA, as demonstrated by

Equation (12)

where in general we have card(S X) >> card(X).
Second, the inequality (5) always holds if two users are within the threshold. Therefore, missed detection mainly results from users accidentally losing track of several satellites, which can happen indoors, in an urban canyon, or in other GPS-challenged environments.

We should note the existence of two types of false alarm:

• Nearby false alarm: A pair of users are incorrectly detected to be nearby; their actual distance is greater than t, but still close to t, and they may see the same set of GPS satellites.
• Distant false alarm: A pair of users are incorrectly detected to be nearby; their actual distance is much greater than t, and they may see totally different sets of GPS satellites.

In this article, nearby false alarm is not our major concern because “proximity” itself is a fuzzy concept in social networking. For example, if two users within a distance t are always deemed nearby, it is acceptable that two users within a larger distance (e.g., 1.5t) are detected to be nearby with a certain probability. The following analysis is about distant false alarm.

Probabilistic Model of Ranges.
Suppose we randomly choose a location on the Earth. At a random epoch the range to an arbitrary GPS satellite observed at this location is a random variable r. Let pdf_r(x) = ^_d/_dx Prob(r ≤ x) be its probability density function.

Here we use a uniform distribution to approximate the actual distribution of ranges, i.e, r ~ U(r_min, r_max). When the satellite elevation mask angle is set to 10 degrees, r_min ≈ 20,189 kilometers and r_max ≈ 24,619 kilometers. Let the spread of range measurements λ = r_max – r_min, and we have pdf_r(x) = 1/λ.

A fundamental assumption of this analysis is that ranges to different satellites are independent and identically distributed (i.i.d.). The validity and efficacy of this assumption has been demonstrated in our previous work (L. Heng et alia).

In this analysis, we ignore GPS range measurement errors because such errors are much less than the threshold distance.

Probability of False Alarm. Suppose user 1 reports an untagged range vector r₁ = [r₁⁽¹⁾, . . . , r₁^(K₁)]^T and user 2 reports an untagged range vector r₂ = [r₂⁽¹⁾, . . . , r₂^(K₂)]^T Both users are randomly chosen on the Earth so that with a very high probability they are far apart. Let c denote the number of matched range measurements. According to our discussion in the section on “Proximity Detection,” false alarm occurs when the match ratio m = c/ min{K₁,K₂} is grater than or equal to the threshold ζ.

Let us randomly shuffle r₂. Based on our i.i.d. assumption mentioned earlier, r₂^(k) ∼ U(r_min, r_max) for all k = 1, . . . , K₂. The probability of r₂⁽¹⁾ matching one of the elements of r₁ is given by

Equation (13)

If r₂⁽¹⁾ matches one of the elements of r₁, then the probability of r₂⁽²⁾ matching one of the remainder elements of r₁ has a similar upper bound 2t(K₁ – 1)/λ. Similarly, the ith match happens with a probability less than or equal to 2t(K₁ + 1 – i)/λ.

Let η = ζmin{K₁, K₂} be the required number of matched range measurements, where ⎡·⎤ is the ceiling function. Finally, the probability of at least η matched range measurements has the following upper bound:

Equation (14)

Since P_FA ≤ 1, we finally have the following upper bound:

Equation (15)

The equation shows that increasing λ (equivalent to using a lower mask angle) and/or decreasing threshold distance t will reduce the false alarm rate.

Experiment 1: Real Data from Global GPS Receiver Networks
We first validate our theory and algorithm using real GPS pseudorange measurements collected by the International GNSS Service (IGS) and the University NAVStar Consortium (UNAVCO). The two networks consist of more than 1,000 stations all over the world. Each station has one or multiple GPS receivers continuously generating GPS pseudorange measurement data. We obtained range measurements by removing receiver clock biases from pseudoranges. The experiment provides a more realistic assessment because the receivers occasionally lose GPS signals.

We treated the IGS and UNAVCO stations as nodes to test for proximity using the scheme outlined in the earlier section. These stations are usually very far (at least tens of kilometers) apart. With a proper distance threshold, the receivers at different stations can be seen as distant users, while the receivers at the same stations are nearby users.

We applied our algorithm to the IGS data recorded on January 10, 2014. The pseudorange measurements released by 1,171 stations around the world at the start of the day during one time epoch was used to aggregate the statistics for validation. Figure 2 shows the variation of probability of false alarm with the threshold distance.

In Figure 3, we see that the missed detection rate is below 0.05 for ζ ≤ 0.8. However, the false alarm rate is higher for lower values of ζ. This trade-off is succinctly depicted in Figure 4, which plots the probability of detection P_D = 1 – P_MD versus P_FA, also known as the receiver operating characteristic (ROC) curve. Thus, the empirical results clearly illustrate the viability of our scheme for efficient private proximity detection.

The results with real data demonstrate the robustness of our scheme. Occasional loss of satellites cause missed detection. With a proper choice of decision threshold ζ, we can still achieve satisfactory detection performance.

Experiment 2: Real Data from Android Phones
With the evaluation using the IGS tracking network, the “user” locations were fixed. Further, most pairs of stations were very far apart and the locations of the stations in the tracking network do not model the distribution of mobile phone users very well. Thus, we performed some local experiments to further validate the algorithm.

We developed an Android application to log GPS range data. Upon post-processing, we can evaluate the utility of our scheme. However, working with the Android API presents a fresh set of challenges. An additional evaluation using GPS receivers is presented to further strengthen the proof of concept.

In an ideal scenario, an implementation of a proximity-based service using an Android app would just have the Android app interacting with a friend-finder server as shown in Figure 5. However, several challenges arose while working with the Android location application programming interface (API), which is the only mode of accessing the underlying GPS engine. We provide a description of the challenges and the suitable modifications required below.

Pseudorange Measurements Not Available from the API. An app developer can only access geodetic coordinates (latitude, longitude, and height) information of an user. Thus, we had to set up an external server to continually download high rate ephemeris from nearby IGS stations. The high rate ephemeris within the last two hours from the stations NIST and GODS (shown in Figure 6) were downloaded and hosted on a server.

The Android app downloaded and updated the ephemeris from the server periodically. Further, Android provides a GPSSatellite class which reports the satellites currently in view along with the elevation and azimuth. Using the last known location, the satellites in view and the downloaded ephemeris, we find the range measurements and form the anonymous range vector for proximity detection.

Unknown Clock Bias and Time Sync Issues.
Another major issue is that the Android API does not report accurate GPS time. It only reports the coordinated universal time (UTC) time at fix and the clock error can be as large as a few minutes. In order to circumvent this problem, the app was forced to compute the satellite positions and range measurements at fixed time epochs.

Initially, we manually configured each of the devices used for testing to not have time lag more than 5–10 seconds. However, the range measurements vary rapidly as the satellites are continually moving. In order to ensure time sync between users, the app was forced to report range measurements at fixed time epochs, i.e, integer multiples of 20 seconds.

Fluctuating Satellite Set.
As stated earlier, the GPSSatellite class reports the satellites currently in view along with the elevation and azimuth. However, the satellites in view were very fluctuant and the 4-5 satellites reported changed very rapidly (on the order of seconds). Thus, we aggregated the reports over 10 seconds to find all the satellites in view.

Synchronous versus Asynchronous Implementation.
In an asychronous implementation, the user notifies the LBS server when he/she is looking for friends nearby. The LBS server can then notify the user’s friends and obtain their untagged range measurements to test for proximity. In a synchronous implementation, the users report their untagged range measurements periodically at fixed time epochs.

The former implementation is obviously better in terms of privacy because the users give away less information. Also, there is the extra communication overhead of continuously reporting data in the synchronous implementation. However, we resort to a synchronous implementation in this work for simplicity. Further, we do not implement a friend-finder server as we are simply recording data for research purposes. The Android app just logs all the pseudorange data in a text file. Figure 7 shows the final implementation used for this experiment.

We processed the files from all users after the recording to evaluate our algorithms. The app was distributed to six graduate students who lived and worked in Urbana-Champaign, Illinois. They primarily worked indoors and were close to each other for most part of the data collection. We present the results from this data collection in Figures 20, 21, and 22.

Figure 8 shows the probability of correct decisions, false alarms, and missed detections. Unlike the IGS stations, most users in this experiment were close to each other for the most period. Hence, we see a relatively higher ratio of false alarms as the denominator in equation (9) is considerable smaller. Figures 9 and 10 show an interesting comparison of the number of mismatches between the anonymous range vectors of two users using two different threshold distance parameters t. We see that the results are consistent with what we expect.

Experiment 3: Real Data from GPS Receivers
In the previous section, we elaborated on the challenges and our approach to Android data collection. We had to construct our own range measurements from the satellite ephemeris and the last known position.

We thus decided to perform another experiment with small L1 single-frequency GPS receivers known to output pseudorange measurements. Four graduate students from University of Illinois took part in this experiment. Two of them drove in opposite directions from the Urbana-Champaign campus for a few kilometers. Two others walked around on the campus. The paths of these users are presented in Figure 11.

Three accompanying figures present the variation of match ratio with distance for t = 750, 1,500, and 5,000 meters, respectively, for a pair of users. From Figure 12, we can see that there is a sharp decrease in match ratio when the distance between the users increases more than the threshold of t = 750 meters. This demonstrates the robustness of the algorithm. For the same pair of users, as can be seen in Figure 13, a higher threshold of t = 1500 meters gives appropriate results in terms of match ratio. In Figure 14, the threshold t = 5,000 meters is always above the distance between the two users. As a result, the match ratio is close to 1.0 most of the time.

Conclusion
This article proposed a novel private proximity detection method, which makes use of partial GPS measurement information. We developed an efficient algorithm for proximity detection. We theoretically analyzed proximity detection performance and derived an upper bound on probability of false alarm.

We further conducted experiments using globally and locally collected data. The empirical results demonstrated the efficacy and robustness of our scheme for performing private proximity detection.

Acknowledgments
This article is based on two of the authors’ papers: “Private Proximity Detection Using Partial GPS Information,” submitted to IEEE Transactions on Aerospace and Electronic Systems, and “GPS Privacy: Enabling Proximity-Based Services While Keeping GPS Locations Private,” in Proceedings
of the 27th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+ 2014), Tampa, Florida, USA.

Additional Resources
[1] Apostolico, A., and C. Guerra, “The Longest Common Subsequence Problem Revisited,” Algorithmica, vol. 2, no. 1-4, pp. 315–336, 1987
[2] Atallah, M. J., and W. Du, “Secure multi-party computational geometry,” in Proceedings of the 7th International Workshop on Algorithms and Data Structures, ser. WADS ’01, pp. 165–179, London, UK, Springer-Verlag, 2001
[3] Ghinita, G., “Privacy for Location-Based Services,” ser. Synthesis Lectures on Information Security, Privacy, and Trust, Morgan & Claypool Publishers, 2013
[4] Heng, L., and T. Walter, P. Enge, and G. X. Gao,
“Overcoming RFI with High Mask Angle Antennas and Multiple GNSS Constellations,” in Proceedings of the 26th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+ 2013), pp. 3433–3442, Nashville, Tennessee USA, September 2013
[5] Hoh, B., and M. Gruteser, “Protecting Location Privacy through Path Confusion,” in Security and Privacy for Emerging Areas in Communications Networks, 2005, SecureComm 2005, First International Conference on Security and Privacy in Communication Networks, 2005, pp. 194–205
[6] Krumm, J., “A survey of computational location privacy,” Personal Ubiquitous Computing, vol. 13, no. 6, pp. 391–399, August 2009
[7] Matsuo, Y., and N. Okazaki, K. Izumi, Y. Nakamura, T. Nishimura, and K. Hasida, “Inferring long-term user property based on
users location history,” in Proceedings of the 20th International Joint Conference on Artificial Intelligence (IJCAI ’07), 2007
[8] Misra, P., and P. Enge, Global Positioning System: Signals, Measurements, and Performance, 2nd ed., Ganga-Jamuna Press, Lincoln, Massachusetts USA, 2006
[9] Narayanan, A., and N. Thiagarajan, M. Lakhani, M. Hamburg, and D. Boneh, “Location privacy via private proximity testing.” in Proceedings of 18th Annual Network & Distributed System Security Symposium (NDSS 2011), 2011
[10] Qiu, D., and D. Boneh, S. Lo, and P. Enge, “Robust Location Tag Generation from Noisy Location Data for Security
Applications,” in Proceedings of the Institute of Navigation International Technical Meeting (ION ITM), September 2009
[11] Qiu, D., and S. Lo, and P. Enge, “Security for Insecure Times: Geoencryption with Loran,” GPS World, 2007
[12] Ravichandran, R., and M. Benisch, P. G. Kelley, and N. M. Sadeh, “Capturing Social Networking Privacy Preferences,” in Proceedings of the 9th International Symposium on Privacy Enhancing Technologies, ser. PETS ’09, pp. 1–18, Springer-Verlag, Berlin, Heidelberg, Germany, 2009
[13] Riley, P. F., “The tolls of privacy: An underestimated roadblock for electronic toll collection usage,” Computer Law & Security Review, vol. 24, no. 6, pp. 521–528, 2008
[14] Shin, K. G., and X. Ju, Z. Chen, and X. Hu, “Privacy Protection for Users of Location-Based Services,” IEEE Wireless Communications, vol. 19, no. 1, pp. 30–39, 2012
[15] Swets, J. A., Signal Detection Theory and ROC Analysis in Psychology and Diagnostics: Collected Papers, Lawrence Erlbaum Associates, Mahwah, New Jersey USA, 1996
[16] Zhong, G., and I. Goldberg, and U. Hengartner, “Louis, Lester and Pierre: Three Protocols for Location Privacy,” in Privacy Enhancing Technologies, ser. Lecture Notes in Computer Science, vol. 4776, pp. 62–76 N. Borisov and P. Golle, Eds. Springer Berlin Heidelberg, 2007

The post GPS Confidential appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

A GNSS receiver estimates 3D location and timing from pseudoranges from four or more satellites, assuming that these pseudoranges correspond to direct line-of-sight (LOS) paths from each satellite. In urban canyons, however, the signal from a satellite to the receiver suffers from multipath propagation and shadowing.

It is by now well known that GNSS-based localization in built-up urban environments can be extremely inaccurate. This is a fundamental problem that hardware enhancements cannot solve.

A GNSS receiver estimates 3D location and timing from pseudoranges from four or more satellites, assuming that these pseudoranges correspond to direct line-of-sight (LOS) paths from each satellite. In urban canyons, however, the signal from a satellite to the receiver suffers from multipath propagation and shadowing.

Even if the LOS path is available, the pseudorange estimate may be corrupted due to the presence of alternative paths. Furthermore, the LOS is frequently blocked except for satellites at the highest elevations; so, the satellite is either unavailable, or the strongest path seen by the receiver is a reflection off a building, leading to a pseudorange often significantly larger than that for the LOS path. These errors in pseudoranges lead to large errors in localization (e.g., up to 50 meters in high-rise environments such as New York City).

Attempts to enhance GNSS localization using signals from other sources, such as cellular and WiFi, have had limited success: it is difficult to use geometric techniques to infer location in a complex propagation environment, and received signal strength measurements are subject to significant fluctuations due to multipath fading and shadowing.

Shadow Matching
Over the past few years, several groups of researchers have realized that we can turn shadowing in urban environments from a bug into a feature, by using information already available in the GNSS receiver regarding the signal-to-noise ratio (SNR) that corresponds to each satellite it sees. If the LOS path from the receiver to a satellite is blocked (i.e., the receiver is in the shadow of a structure), then this SNR is likely to be small. Conversely, if the LOS path is available, the SNR is likely to be high.

If we interpret these SNR measurements with the aid of a 3D map of the environment, then we can significantly reduce localization uncertainty by performing shadow matching, as shown in Figure 1 (see inset photo, above right, for all figures). In its simplest form, the concept of shadow matching is summarized as follows:

Shadow Matching with 3D Maps
1. Compute shadows of nearby obstacles using satellite ephemeris data and 3D maps
2. Classify each satellite as visible (LOS) or blocked (NLOS) based on measured SNR
3. Match the location of the GNSS device to areas inside/outside of the various shadows

Any GNSS-capable Android smartphone or tablet can provide, via the location application programming interface (API), its estimated position with uncertainty, as well as the satellite coordinates and SNRs. Thus, shadow matching can be performed entirely in software without requiring any changes in GNSS receiver hardware or firmware.

The article by Wang et alia listed in the Additional Resources section near the end of this article provides an introduction to the basic idea of shadow matching. Wang and his coauthors reported promising results in mitigating GNSS cross-street errors using relatively straightforward techniques.

While shadow matching clearly provides information that is useful for urban localization, naive application of this idea using deterministic algorithms has significant limitations. First, the concepts of “high SNR when not blocked” and “low SNR when blocked” are inherently probabilistic, because SNRs exhibit large fluctuations due to multipath fading. Second, the range of SNRs seen by a device depends on the hardware implementation and physical realization of the GNSS receiver. Third, the potentially large errors in the raw location estimates output by the GNSS receiver are not captured by the typical computations of uncertainty regions based on dilution of precision (DOP), because these implicitly assume that paths to all satellites are LoS. Hence, it is unclear how large an area to perform shadow matching over and how to fuse shadow matching information with GNSS location estimates. Finally, shadow matching is based on having access to accurate 3D maps, which are not readily available in many settings.

In the work described here, we address the preceding challenges through an adaptive Bayesian framework for inference, localization, and tracking with two complementary ingredients:

• Real-time Localization: Where accurate 3D maps are available, we develop nonlinear filters for localization based on the following ingredients: (a) probabilistic, rather than deterministic, modeling of shadow matching; (b) on-the-fly adaptation to device characteristics; (c) probabilistic fusion of information from shadow matching and GNSS location estimates; (d) mobility modeling using state space techniques. Because of significant modeling uncertainties and computational constraints, our solution needs to go well beyond classical particle filters.
• 3D SLAM: When accurate 3D maps are not available, the same Bayesian framework, with additional computation, is used for generating and refining 3D maps using simultaneous localization and mapping (SLAM) based on crowdsourced GNSS data. Intuitively, we can do this by assigning likelihoods of blockage to many crisscrossing receiver-satellite rays based on measured SNRs and then stitching these rays together, accounting for the uncertainty in the receivers’ locations, into 3D maps. The framework can exploit “warm start” (e.g., based on public building data), if available.

Increased accuracy in urban localization clearly has many compelling applications, including car services, delivery services, navigation and guidance for vehicles (including for emerging automated and semi-automated operation), tourism, and hyperlocalized advertising.

We do not discuss any application in detail but instead focus on core technical aspects of our proposed solution in the remainder of this article. In particular, we describe the key concepts behind these algorithms in more detail, discuss cloud-based implementations with lightweight application layer modifications to mobile devices, and provide experimental results demonstrating the scalability and efficacy of these techniques.

Initial results based on these ideas, and the underlying technical details, were described in a series of publications last year. (See the articles by A. T. Irish et alia and J. T. Isaacs et alia in Additional Resources.) The experimental results reported here are based on significant enhancements to those techniques.

Real-Time Localization and Tracking
In Bayesian estimation, we compute the conditional distribution of a quantity of interest given a set of measurements. This is termed a posterior distribution, because we can only compute it after we see the measurements. In particular, we are interested in estimating the posterior distribution for the location of the GNSS receiver, based on measurements consisting of the GNSS position fixes and the satellite SNRs. In order to do this, we must model the conditional distribution of these measurements, conditioned on the actual location.

We consider a discrete time model, typically with samples spaced by one second. The true 3D location at time t is denoted by x_t, and the corresponding GNSS location fix is denoted by y_t. The satellite SNR measurements are denoted by z_t,n, where z_t,n is the SNR to satellite n at time t.

Modeling GNSS Location Fixes. The measurement model for the GNSS location fix is, in principle, straightforward:

y_t = x_t + e_t,

where the covariance of the error e_t can be estimated using standard DOP-style computations. However, these standard computations assume that the paths seen by the GNSS receiver are LOS, whereas many of the dominant paths in urban environments are actually strong reflections, with the LOS path blocked. Thus, in order to obtain satisfactory localization performance, we need to modify the model for {e_t}, resulting in a non-Gaussian statistical model.

Modeling Satellite SNR Measurements. We represent the 3D map m using an occupancy grid: space is divided into binary-valued “voxels” or “cells,” with m_i = 1 if the ith cell is occupied, and m_i = 0 if the ith cell is unoccupied. If the ray from a hypothesized location to a satellite crosses only unoccupied cells, it is classified as LOS. If a ray passes through at least one occupied cell), it is classified as NLOS (or shadowed).

Figure 2 illustrates this approach. On the left, we illustrate ray tracing: blue/red lines represent LOS/NLOS signal paths to satellites, light/dark grid cells approximate empty/occupied space. On the right, we show typical SNR distributions that have been found to yield good performance: Rician for LOS, and lognormal (with smaller mean and higher spread) for NLOS.

Note that we have made a drastic simplification in our model for NLOS rays, in that the distribution does not depend on the number of occupied voxels the ray crosses. Although it might be possible to improve performance using a more detailed model, our simplification leads to a significant reduction in computational complexity, while yielding excellent localization performance. Additional improvements can be obtained by introducing dependence of the LOS/NLOS SNR distributions on satellite elevation.

Tracking framework. We use a standard linear state space model for pedestrian motion, with the state consisting of position and velocity. However, because the measurement model is nonlinear and non-Gaussian, a standard Kalman filter cannot be used. Simple extensions such as the extended Kalman filter also do not work, as the posterior distribution of the state is often multimodal.

We therefore use a particle filter. Roughly speaking, it operates as follows: At each time t, the posterior distribution of the state is approximated by a discrete probability mass function putting weights at a set of hypothesized state values, or particles. These particles are propagated probabilistically to obtain a new set of particles and weights at time t + 1, based on the dynamics of the motion model, and the new set of measurements.

The particle filter has by now become a standard tool, but we have needed to make a number of modifications in order to account for modeling uncertainties.

Figure 3 shows example pedestrian results for Bush Street in downtown San Francisco. Figure 3(a) shows the mean trails and 68 percent confidence circles for the GPS reported fix (red) and the particle-filtered estimate (blue), with the ground truth path in yellow. Note that GPS makes cross-street errors which are corrected by our algorithm.

Figure 3(b) shows the SNR likelihood surface at the beginning of the trail: GPS starts out with a cross-street error, which our algorithm is able to correct because the SNR likelihood has a strong peak on the correct side of the street. The large greyish-green ellipse is the 3σ uncertainty estimated by GPS around its (incorrect) estimate, while the smaller black ellipse is the 3σ uncertainty estimated around the location estimate provided by the particle filter.

Finally, Figure 3(c) shows the composite SNR/GPS likelihood surface, which exhibits a peak at the correct location. We also show the satellite rays colored likely LOS (green) to likely NLOS (red) according to our SNR model; from the image, we can see that these do correspond to unblocked and blocked rays, respectively.

Vehicular applications require more sophisticated motion models, and knowledge of the road network can also be exploited. However, the essential ingredients, in terms of the measurement model and the particle filter, are the same. Figure 4 shows example vehicular results with the 68 percent probability tunnels for both the GPS reported estimate (red) and our filtered estimate (blue), along with a road centerline clamping of the filtered estimate (green).

SLAM for CrowdSourced 3D Maps
For 3D mapping, we are interested in using the noisy GNSS position measurements y and the SNR measurements z, to estimate the 3D map m. In particular, we wish to estimate the probability of each map cell being occupied or not, given all the measurements. That is, we wish to compute the marginal posterior distributions, p(m_i | y,z) for each cell i. However, because the GPS fixes are noisy, we do not know the paths x followed by the devices. Thus, in order to estimate the map, we must also estimate quantities of the form p(x_t^j | y,z), where x_t^j is the position of a particular device j at time t. In the robotics community, this is referred to as the SLAM problem.

Our approach for Bayesian estimation of the map is to use a factor graph representation of the measurements and the quantities to be inferred, which allows us to employ loopy belief propagation to approximately compute the map marginal posteriors. While a detailed exposition of factor graphs is beyond the scope of this article, we provide intuition into their applicability in our context.

Figure 5 illustrates the construction of the factor graph based on measurements corresponding to a single GNSS receiver over two consecutive time periods. The variables that we wish to perform inference on, denoted by circles, are the binary-valued map cell occupancies {m_i} (which is our primary interest in shadow-based SLAM) and the true locations {x_t} as a function of time t (which are, for the mapping task, nuisance variables to be averaged out).

The factors represent the statistical information about the variables provided by the measurements. In Figure 5, the factor g_t represents p(y_t | x_t), the likelihood function corresponding to the GPS fix at time t, and the factor f_t,n represents p(z_t,n | m,x_t), the likelihood function corresponding to the SNR to the nth satellite at time t.

Belief propagation can now be implemented by the standard sum-product algorithm described in the article by F. Kschischang et alia, passing messages back and forth between nodes. A message is simply a function whose domain is the set of values taken by the variable. Thus, for a binary-valued map cell m_i, a message is a binary vector.

For a position variable x_t, we simply quantize its possible set of values to a discrete set (e.g., with K possible values), and a message is then a K-dimensional vector corresponding to the values of the function evaluated at these K points. A message along an edge from a variable node to a factor node is simply a product of messages coming along all of the other edges.

A message from a factor node f to a variable node v is more complicated: it involves products of messages coming into that factor node, but also involves summing over (marginalizing out) all variables other than v. We refer the reader to the article by A. T. Irish et alia (2014a) for details, but it is worth mentioning that our simplified SNR model (in which the NLOS model for SNR is independent of the number of occupied map cells intersected by the ray) is key to tractable computation of messages from SNR factor nodes to position and map variable nodes.

The preceding procedure has been successfully used to construct 3D maps of the campus of the University of California, Santa Barbara (UCSB), as well as of downtown Santa Barbara. Here we will present sample results for the latter based on about 25 hours of input data from four Android devices. Figure 6(a) shows an aerial view, with GNSS traces in red and mapped region outlined in yellow. Figure 6(b) shows the cell occupancy estimates for 2D slices at multiple heights.

We employed a “warm start” with prior information based on 2D OpenStreetMap data employed to initialize the map for the first couple of slices. Note, however, that, given enough data, 3D maps can be generated without any prior information. (This was the case for the 3D maps generated for UCSB campus; see A. T. Irish et alia (2014a) for details).

Architecture and Implementation
Shadow-based localization and 3D SLAM can be easily deployed on existing mobile devices via a software development kit (SDK), which performs a very simple function: send GNSS data to “the cloud” at regular intervals and receive improved position information (along with an estimated uncertainty) back. The GNSS data includes the estimated latitude and longitude coordinates, along with the azimuth, elevation, and SNR of each satellite in view. As shown in Figure 7, all of the heavy lifting for both real-time localization and 3D mapping is done by means of Internet cloud computing.

The preceding architecture naturally scales to include additional sources of information when available — for example, inertial navigation, WiFi, and cellular. Such information can be sent from mobile device to cloud, and can be incorporated into the Bayesian computations done in the cloud. Such extensions are also under development.

Conclusions
The accurate urban localization provided by enhancing GPS with shadow matching has the potential to create significant commercial benefits, including for car/taxi services, delivery services, location-based advertising, e911, tourism, and automated and semi-automated navigation.

Several academic research groups worldwide have demonstrated the promise of shadow matching over the past few years. However, a key innovation in the approach described here is that, in contrast to the ad hoc algorithms used previously, it is based on a Bayesian inference framework that systematically accounts for significant modeling uncertainties and nonlinearities, and seamlessly extends to incorporate additional sources of information as they become available.

The approach proposed here has been demonstrated to provide significant improvements in localization over raw GNSS estimates in challenging high-rise environments such as downtown San Francisco, with latencies of the order of 100 milliseconds. With current (and ever decreasing) cloud computing costs and the inherent parallelizability of the core computation bottlenecks (such as ray tracing and particle sampling techniques), providing such a service is estimated to cost less than one U.S. penny per hour per device.

On the mapping side, shadow-based SLAM provides a method for creating 3D maps from scratch and an effective means of refining them from a warm start. The former is probably most important for military applications.

For most urban environments worldwide, 2D maps together with public municipal data on building heights can be used to build a coarse 3D map, which can subsequently be refined by shadowbased SLAM. Even in areas where accurate 3D building models are available, shadow-based SLAM provides the ability to continuously monitor and capture changes in the environment (e.g., new construction, demolition). It also captures features such as trees and lampposts. As vehicles migrate towards automated operation, a fine-grained mapping of features in the environment become increasingly important for timely decision making in navigation and path planning.

While the cloud-centric architecture described previously in the “Architecture and Implementation” section enables immediate deployment of shadow-matching technology on mobile devices via a lightweight SDK, it is also of great interest to explore means of efficiently migrating some of the algorithmic functionalities to a mobile device. Accomplishing such a migration would reduce reliance on network connectivity by exploiting the computational resources available on today’s smart phones. Self-contained operation with limited or no network connectivity is also an important feature for incorporating these advances in urban localization into vehicular navigation systems.

Additional Resources
[1] Ben-Moshe, B., and E. Elkin, H. Levi, and A. Weissman, “Improving Accuracy of GNSS Devices in Urban Canyons,” in Proceedings of the 23rd Canadian Conference on Computational Geometry, 2011
[2] Groves, P., “Shadow Matching: A New GNSS Positioning Technique for Urban Canyons,” Journal of Navigation, 64(3):417–430, 2011
[3] Irish, A. T. (2014a), J. T. Isaacs, F. Quitin, J. P. Hespanha, and U. Madhow, “Belief Propagation Based Localization and Mapping Using Sparsely Sampled GNSS SNR Measurementsl”, in Proceedings of IEEE International Conference on Robotics and Automation, 2014
[4] Irish, A. T. (2014b), J. T. Isaacs, F. Quitin, J. P. Hespanha, and U. Madhow, “Probabilistic 3D mapping based on GNSS SNR measurements,” in Proceedings of IEEE International Conference on Acoustics and Signal Processing, 2014
[5] Isaacs, J. T., and A. T. Irish, F. Quitin, U. Madhow, and J. P. Hespanha, “Bayesian Localization and Mapping Using GNSS SNR Measurements,” in Proceedings of IEEE/ION Position Location and Navigation Symposium, 2014
[6] Kim, K., and J. Summet, T. Starner, D. Ashbrook, M. Kapade, and I. Essa, “Localization and 3D Reconstruction of Urban Scenes using GPS,” in Proceedings of IEEE International Symposium on Wearable Computers, pp. 11–14, 2008
[7] Kschischang, F., and B. Frey, and H.-A. Loeliger, “Factor Graphs and the Sum-Product Algorithm,” IEEE Transactions on Information Theory, 47(2):498–519, 2001
[8] Park, M., and H. Kim, S. Lee, and K. Bae, “Performance Evaluation of Android Location Service at the Urban Canyon,” in Proceedings of IEEE International Conference on Advanced Communication Technology, pp. 662–665, 2014
[9] Thrun, S., “Simultaneous Localization and Mapping,” in Robotics and Cognitive Approaches to Spatial Mapping, pp. 13–41, Springer, 2008
[10] Wang, L., and P. D. Groves, and M. K. Ziebart, “Urban Positioning on a Smartphone: Real-Time Shadow Matching Using GNSS and 3D City Models, In Proceedings of ION GNSS+, 2013

The post Urban Localization and 3D Mapping Using GNSS Shadows appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.

The current stage of GLONASS evolution is aimed at meeting future user requirements of which the most important is the improved accuracy of positioning.

During the implementation of the GLONASS Space Segment Modernization Program (2012–2015), the GLONASS team is facing the situation in which it is not feasible to launch new navigation satellites because the existing constellation is comprised of GLONASS-M satellites operating beyond their guaranteed design lifetime. Nine more GLONASS-M satellites are in ground storage.

The current stage of GLONASS evolution is aimed at meeting future user requirements of which the most important is the improved accuracy of positioning.

During the implementation of the GLONASS Space Segment Modernization Program (2012–2015), the GLONASS team is facing the situation in which it is not feasible to launch new navigation satellites because the existing constellation is comprised of GLONASS-M satellites operating beyond their guaranteed design lifetime. Nine more GLONASS-M satellites are in ground storage.

In these circumstances the decision was taken to install L3 code division multiple access (CDMA) navigation payloads onboard the GLONASS-M satellites # 755–761 — only the first of which has been launched. In the meantime, the GLONASS developers’ team conducted analysis and experimental testing of the options to improve performance of the GLONASS services based on the current GLONASS-M satellites transmitting the traditional frequency division multiple access (FDMA) signals.

Analysis of GLONASS Evolution Prospects
As a result of the GLONASS Federal Program in 2002–2011 resulted in the creation of a considerable scientific and technological backlog. Technological solutions for augmenting GLONASS were implemented primarily based on modernized software to enable the improvement of system performance. The most significant of these augmentations is the ground network of reference stations for the System of Differential Correction and Monitoring (SDCM).

A trade-off analysis for GLONASS services development identified the need to create a global ground network to improve the accuracy of orbit and clock prediction when implementing techniques for estimating the combined ephemeris and time parameters using one-way measurements by the ground segment.

Associated efforts initiated in 2007 resulted in deployment of a ground network with 19 reference stations in the Russian Federation and 4 stations abroad for the SCDM, which enable observation of the GLONASS satellites in both hemispheres. The main disadvantage of the network is the lack of overlap for satellite visibility zones from the stations in the Russian territory and the stations abroad. This results in the need to attract additional international stations on the territory of the Indian Ocean to provide for orbit and clock data estimates along the whole satellite track.

Despite these limitations, the prerequisites for improving the current GLONASS performance still exist. That is why the task of estimating the potentially achievable GLONASS performance in terms of accuracy becomes utterly vital. The main issue is identifying the minimum number of reference stations needed and determining the dependence of orbit and clock data errors on the periodicity of their update on board a satellite.

GLONASS Performance Improvement Options
We conducted an experimental study to assess options for improving GLONASS performance. To do this we used the adaptable high-accuracy hardware and software of the Information and Analysis Center (IAC) for Positioning, Navigation and Timing residing in the Central Research Institute of Machine Building of the Russian Federal Space Agency.

Based on International GNSS Service (IGS) weekly estimates, the IAC determined the orbital error of GLONASS satellites to be about 0.02–0.025 meter (RMS). In the technological cycle employed in IAC we also estimate the systematic errors of the reference stations’ receivers. The delay of the IAC final data is seven days.

We selected a 45-day period starting February 2, 2015 for the estimation of of potential GLONASS performance. Figure 1 (see inset photo, above right, for all figures and tables) shows the network of selected SDCM reference stations expanded with several IGS stations. These stations provide a data stream at one-hertz frequency.

We conducted the orbit and clock data estimation using two-day interval measurements with six-hour periodicity. The estimated parameters included the GLONASS satellites ephemerides, the stochastic corrections to satellite and reference station timescale parameters, the troposphere model parameters, the coordinates of the stations, and the Earth rotation parameters.

At the end of each solution the predicted orbit and clock data for a 24-hour interval were generated. It is worth mentioning that the clock corrections were generated using an adaptive model without high-accuracy referencing to the GLONASS time scale.

The equivalent user range error (UERE, disregarding propagation errors and user receiver biases) was estimated against the postprocessed precise orbit and clock data generated by IAC at 15-minute intervals. We estimated UEREs for all satellites in the constellation without limitations.

Figure 2 presents the estimates of Allan deviation for the clocks on all operational satellites in the GLONASS constellation.

The following five optional modes of operation for onboard and ground capabilities were studied when estimating UERE (Table 1):

• estimating, predicting, and uploading orbit and clock data every 6 hours with 6-hour delay from the latest measurement (orbit and clock data prediction at the 6–12 hour interval
• estimating, predicting, and uploading orbit and clock data every 6 hours with an hour delay from the latest measurement (orbit and clock data prediction at 1–6 hour interval instead of the 1–7 hour interval)
• estimating, predicting, and uploading orbit and clock data every hour with 2-hour delay from the latest measurement (orbit and clock data prediction at the 2–3 hour interval)
• estimating, predicting, and uploading orbit and clock data every 2 hours with an hour delay from the latest measurement (orbit and clock data prediction at 1–3 hour interval)
• estimating, predicting and uploading orbit and clock data every 15 minutes with an hour delay from the latest measurement (orbit and clock data prediction at 0.25–1.25 hour interval).

The obtained results demonstrate that the accuracy of user position solutions can be improved by employing the orbit and clock data determination technique based on the measurements from both the limited ground network and the FDMA navigation signals. The improved orbit determination and clock synchronization accuracy is achieved through the faster calculation of the orbit and clock data and its delivery to users.

GLONASS Use for Precise Orbit Determination of LEO Satellites
For decades before the global navigation satellite systems came into being, the position of any satellite had been determined through a dynamic method incorporating a priori specified models (geopotential, tidal deformations, atmosphere, and so on). Such models were validated along with the satellites orbits, while solution accuracy was estimated based on measurement residuals or laser ranging data. With the advent of GNSS, onboard carrier phase measurements and precise point positioning (PPP) technology in particular have enabled a change in the approach to the solution of the aforementioned tasks.

Such solutions are based on high-accuracy GNSS satellite orbit and clock data, resulting from the sophisticated processing of one-way navigation measurements obtained from global networks of commercial geodetic-grade receivers.

In the 1990s such solutions were carried out only for geodynamic research in analysis centers (such as those located in Pasadena, Ottawa, Bern, Potsdam, and Darmstadt) united by the International GNSS Service (IGS) with its Central Bureau located at the NASA Jet Propulsion Laboratory (JPL). During that period methods and models of processing phase measurements were being improved using a posteriori approaches.

By that time it had already become clear that GNSS phase measurements could potentially enable the determination of user position in real time with significantly better accuracy. High-accuracy GNSS satellite orbit and clock data is the only thing needed to accomplish this. Still, one substantial restriction on carrier phase positioning is the need for a continuity of phase measurements at a relatively high rate providing cycle-slip detection.

Therefore, it seems reasonable that the evolution of GNSS measurement techniques could support orbit determination of low Earth orbiting (LEO) satellites. The position dilution of precision (PDOP) at a LEO altitude of 500 kilometers is almost the same as that at the ground level, while atmospheric influences, the major error source for a ground-level user, are substantially lower.

In order to estimate the accuracy of a LEO satellite position determination based on GLONASS measurements, we chose the 24-hour interval (January 7, 2015) of measurements by a navigation receiver on board the Resurs-P satellite. This Earth remote sensing satellite is the first in a series of three spacecraft designated for updating maps, supporting economic activity of some user groups, as well as collecting data to support environmental monitoring and protection.

The Resurs-P satellite carries four GNSS navigation receiver sets on board, providing a unique opportunity to estimate both the spacecraft’s flight path and its orientation pattern based on navigation measurements. Figure 3 shows the exterior of the Resurs-P satellite.

Estimation of the Resurs-P flight path error may only be evaluated indirectly using high-accuracy ephemeris and time information based on phase measurements of GLONASS satellites’ signals. The availability of several navigation receiver sets onboard the spacecraft enables us to compare flight paths obtained and processed independently and thus evaluate the accuracy of orbits. Figure 4 shows the difference of flight paths from separate GNSS receivers. Figure 5 provides a typical example of phase measurement residuals.

In order to improve orbit quality all four receivers were processed in one filter together with strong constraint on the distance between each pair of receivers and other parameters. Taking into account that the noise error of residuals for the ionosphere-free combination phase measurements are at the level of 0.01 meter (RMS) and that the flight paths from the various receiver sets, obtained through independent processing, align at 0.05 meter (RMS), we can estimate the 3D final flight path accuracy for combined processing at the level of 0.10 meter (RMS).

The four receiver sets also allow estimation of the satellite’s attitude based on GNSS measurements only. Figure 6 shows the values of the Resurs-P roll axis angular displacement off the Earth’s center direction. Thus, the noise component of the satellite attitude determination is at the level of 10 angular minutes (RMS).

Conclusion
Allowing for the current onboard clock performance of GLONASS-M satellites and augmentations are capable of driving the orbit and clock error down to 0.3 meter in real time. This allows use of GLONASS with existing FDMA signals for high accuracy navigation.

Additional Resources
1. Lyskov, D., Deputy Head of the Russian Space Agency, Roscosmos, “Directions 2014: New Horizons of GLONASS,” GPS World, December 1, 2013
2. The Central Research Institute of Machine Building Informational-Analytical Center, Service for the comprehensive analysis of GNSS performance [Electronic resource]. (date of access: April 10, 2015)

The post GLONASS for Precise Navigation in Space appeared first on Inside GNSS - Global Navigation Satellite Systems Engineering, Policy, and Design.