Rx For Satnav Obscuration

Apr 2012 | No Comment
It is increasingly clear that our industry has an opportunity to pool our information from multiple sources and mitigate the current high level of risk. Let’s just do it.

James L Farrell

Vigil, Inc.,
Severna Park, Maryland, USA

Various publications have lately drawn increased attention to the industry’s need for more timely adaptation to change. While recent commentary has largely focused, understandably, on spectral issues, it is widely understood that the overall need is multifaceted – i.e., in addition to the highly important spectrum-related investigations already widely circulated, there are further steps that could reduce vulnerability to coverage limitations. To the existing tracts on the adaptivity theme I wish to offer my own observations. As musings prompted by one individual’s limited experience, these reflections cannot by any means represent a complete prescription answering this global question. Nevertheless, significant value can be gained from a response offering meaningful solutions wherever needed. In fact, solutions offered herein are asserted to give performance improvements that are urgently required.

Need for broader receptiveness

Past and current happenings provide instructive insights. PayPal founder Peter Thiel noted that “If Einstein sent a letter to the White House, it would get lost in the mail room and be treated as a joke.” I’m no Einstein – far from it – but even from a much lower level, a need for change is easily sensed. For relevance of Mr. Thiel’s statement, consider: Einstein wasn’t mainstream. To connect that fact with the present, statements below exemplify utterings from three of the best-known people in today’s navigation community:
• “… can’t see the forest for the trees” (regarding a popular methodology)
• ” … it’s amazing how slowly good ideas get adopted!”
• ” … it’s like pulling teeth to get a good idea accepted today”

Some resistance to innovation is attributed to an installed base, but upgrading from now through the future does not have to imply a demand for retrofitting. Inertia is another major impediment to acceptance of innovation – people are creatures of habit. A past experience described below illustrates that.

One Saturday night in early January 1981 I was nearly killed. Next to the window at the “Top of the World” restaurant atop the World Trade Center, even the edge of the brick outside was obscured by fog. A foreign airliner, essentially blinded by it, very nearly collided with the building (that time it would have been accidental). Still, without action by an alert air traffic controller, I would have been gone. It occurred to me then: data bases with buildings and all other potential obstructions could easily be compared vs projected fiight paths to avert danger. Controlled Flight Into Terrain (CFIT) was already a known concept; why wasn’t prevention a routine procedure? After years of talking about this to as many people as possible, I finally stopped. Then in 1996, U.S. Secretary of Commerce Ron Brown was killed as a plane collided with a mountain. CFIT increased in importance and finally became widely recognized. Moral of the story: timely acceptance of existing available solutions could prevent, or at least reduce, tragic happenings.

Inertial augmentation: Perceptions

Wider recognition and increased importance are now being attached to protection against GPS/GNSS vulnerability. Again, much important spectrum work – already documented – needs no repetition here. Instead, attention is now drawn to methods solving other critical problems. One area is GPS/Inertial integration for which a recent list of urgent needs [1] included (among other issues) the following perceived shortcomings:
• defining models for the inertial sensor errors

• cost

• attitude accuracy

• complexity of use

• availability (or unavailability) of other sensors/systems.

Responding to those perceptions, with sufficient depth despite the necessity for brevity, can be done only by reference to other material. Citing excerpts from my own writings by no means implies that one individual has all the answers (in any case, not all the solutions originated with me), but it facilitates addressing the issues succinctly and from a consistent perspective.

Gyro and accelerometer error models The first item enumerated above often arises from characterization of IMU errors as sets of randomized inputs to various transfer function models (e.g., with positive and negative slopes of different order on plots vs frequency). The concepts are valid but the effects they cover are often overshadowed by a host of motion-sensitive degradations. A “tip-of- the-iceberg” is discussed in [2] while Chapter 4 of [3] plus Addendum 4.B of [4] provide extensive coverage. I’m pursuing it further in an ongoing funded project.

Cost Item #2 is addressed here for systems not needing extended coast capability (which covers a substantial percentage of applications). For that case it is fairly widely accepted that a low-cost system can provide useful performance. Taking full advantage of that, however, requires exploiting methods not at all in the mainstream today. Further elaboration of that fits naturally with the next points.

Attitude accuracy Item #3 is one key reason to go far beyond methods that are almost universal today. In practically all published flight or ground test results (not simulations) with GPS/INS, leveling accuracies are on the order of a few tenths of a degree (and accompanying velocity accuracies do not at all reflect inherent satnav capability). In marked contrast to that, the fight-validated methods documented (with no proprietary strings) in [4] show state-of-the-art RMS leveling accuracies (a few tenths mrad), with cm/ sec RMS velocity errors, from roughly an hour of flight (and also from earlier van tests) under severe vibration with low-cost IMUs. A brief description appears in [5].

Complexity of use Item #4 in many cases arises from attempts to “graduate” from loosely coupled to tight or ultratight integration. The latter, and often also the former, involve carrier phase – which many perceive as simple in concept but difficult in practice. Two main reasons for that perception are (1)incomplete understanding of a specific receiver’s implementation and (2)intrinsic risk, even with flawless design, of incorrect ambiguity resolution causing catastrophic error. For the first of these, choose only receivers that meticulously form integrated doppler and not approximate deltaranges formed by sampling partial (but not complete) phase history. For the second, usage of 1-sec sequential changes – rather than carrier phases themselves – provides the precise dynamic accuracies just described [5] while eliminating a host of problems (interoperability, acceptance of high intermittency, no masking nor ambiguity). An additional reason for perceived (not intrinsic) difficulty in usage involves the Kalman filter design itself, specifically the settings for process noise. As with other facets of this topic, there is room for only a short reference here: Section 4.5 of [4]. Also worth noting here – the no-strings documented formulations in [4] contain solutions to “problems” cited elsewhere (including one publication cautioning its readers with over ten “caveats” for carrier phase usage – all of which had successful solutions already fl ight-verified).

Availability / unavailability of other sensors/systems The last of the enumerated items is a basic integration issue concerned with widening the scope. Notice the reference to “sensors/systems” instead of just sensors. Systems (actually subsystems) give solutions based on subsets of information available to each individual (sub)system. Sensors provide observables that can be combined centrally to produce solutions based on all available information. Ever since day one of Kalman filtering, the best way to extract full information from whatever is available has called for using raw data for all observables. A commonly employed alternative is to combine outputs of “systems” – again, actually subsystems – instantly identifiable with loose coupling approaches. In many cases it works adequately but, as widely acknowledged, it is unprotected from scarcity of data – e.g., incomplete or marginal satellite coverage becomes “loss of satellite navigation.” Another major drawback is inflexibility; a turnkey system is convenient until a change is desired (e.g., to accommodate new conditions or measurements to be added). Modification costs then rival the price of the original procurement. Solution: insist on availability of raw measurements and use public domain algorithms to manage them.

Promises, promises – and deliverance

All methods prescribed in discussions that follow are fully consistent with the guideline just stated. Downloadable examples capitalizing on that for satellite navigation with marginal coverage include recent discussions about maintaining operation under adverse conditions, [6] and [7].

For those interested I’ll point the way to further information. Depending on how you count, I’ve attacked vulnerability in multiple areas for some years [8]; a decade [9]; two decades [10,11]; over three decades [12]; – and I’m still at it, e.g., now for collision avoidance [13- 16] (with a YouTube link on [13]) – not too soon with unmanned aircraft about to enter the airspace. Much of the vulnerability we face today stems from old habits. Conventional operation hinges on dependable full fixes with RAIM/ FDE-supporting geometry. Inescapably that reliance precludes robustness, just as loose coupling precludes both robustness and highly accurate dynamics (and in fact, tight coupling is only the beginning in the list of features needed). Understated in so much present and future system planning is intermittency due to the severe full range of reasons (jammers, masking, attenuation, IONO/tropo/multipath, vulnerability from weak signals and any other missing link (e.g., communication, etc.) – plus whatever unforeseeable threats emerge in the future.

Receptiveness – then vs now

Earlier advances (e.g., Differential GPS and conventional usage of carrier phase) were readily and widely accepted but, as the inventory of existing systems has grown, constraining habits have inhibited acceptance of later advances, even as they addressed requisite adaptation to changing conditions such as SA removal, processing advances, new adaptations; e.g.,

• exploitation of sequential changes in ambiguous carrier phase for Dead Reckoning [5]
• extension of DGPS to Relnav with no stationary participant required [9]
• means of accounting for measurement error correlations due to differencing [4]
• operations and interoperability facilitated by singlemeasurement RAIM [17]
• progress in coordinating surveillance with nav – DME, AGPS, • • •

• innovations via technology advances – three benefits of FFT processing [18].

Among myriad examples that could illustrate advantages of practices advocated I’ll cite one here for SV’s either about to vanish or just emerging over the horizon: Recall the insensitivity of 1-sec sequential carrier phase changes to Iono/Tropo effects. No masking is needed; only those observations affected by multipath need be extracted by the single-measurement RAIM just mentioned. The geometry benefit is self-evident. These and many additional issues were discussed at ION-GNSS-11 panel [19] with my presentation documented in ION-GNSS 2011 Proceedings as “Processing of Measurements for Robust Operation under Adverse Conditions.” For that presentation I was allowed to borrow from Ohio University designers a short video showing a road test they had made earlier with satellites flickering in and out due to presence of trees. Comparison was made vs a conventional receiver that used correlators and track loops, not FFTs. No contest. The same 1-sec sequential changes in carrier phase that provided 1-cm/sec RMS velocity accuracy in flight with a low cost IMU (and 1-decimeter/sec RMS without it) again were crucial. The brief glimpses that weren’t enough to maintain lock in the conventional receiver were fully adequate with the advanced approach using FFTs and 1-sec phase changes.

Numerous other publications could be cited (including a 90-minute tutorial at, free to ION members, plus columns I wrote for InsideGNSS and GPSWorld). Again, those represent one man’s experience; other individuals can recall other experiences and cite additional examples illustrating a similar theme: Much of what is “mainstream” is duplication; the industry needs to open up its acceptance of readily available means that are in urgent demand. As the industry moves further toward alternatives to GNSS – using DME, Wide Area Multilateration (WAM), and pseudolites [20] – some of these concepts (e.g., measurements-not-coordinates) offer substantial performance benefits in those approaches as well.

Another kind of availability

As a kid I laughed hard at movie scenes starting with Bob Hope trudging through a desert, desperately uttering “water, water” – and finding himself moments later waist deep in a stream, mumbling “mirage, mirage” – with his desperation undiminished as ever. Parallels between that comic sequence and not-so-funny real life occur repeatedly in multiple areas, including our own. The good news is: solutions are available. Will we accept them? It’s in that spirit that I wish to respond positively to current needs.

Originally that last paragraph was intended to close this writing – and then the March 2012 issue of Coordinates arrived with its excellent cover story. Advocating needed steps (e.g., robustness enhancement, test standardization, • • • ), [21] offers
• questions such as “Do we really need to wait for a catastrophe before taking action against GNSS vulnerabilities ?”

• statements such as “It’s all solvable. Major portions are already solved.” Amen to all that. It is increasingly clear that our industry has an opportunity to pool our information from multiple sources and mitigate the current high level of risk. Let’s just do it.


[1] InsideGNSS webinar Feb 29 2012

[2] “Gyro Mounting Misalignment”

[3] Farrrell, J.L., Integrated Aircraft Navigation, Academic Press, 1976 (now in paperback) –

[4] Farrrell, J.L., GNSS Aided Navigation and Tracking: Inertially Augmented or Autonomous, American Literary Press, 2007 –

[5] “Dead Reckoning by GPS Carrier Phase”

[6] “Robust Design for GNSS Integration,” ION-GNSS-08

[7] “Aging SV’s – We Have Solutions,” ION-GNSS-09 http://

[8] “ADSB (2nd-) Best Foot Forward?” Air Traffic Control Journal v50 n3 Summer 2008, pp 17-18

[9] “Send Measurements not Coordinates,” ION Journal Autumn 1999 Publ. #66 of

[10] “That All-Important Interface” (coauthor with Prof. vanGraas, ION-GPS-90)

[11] “System Integration: Performance Doesn’t Measure Up” (NAECON 1993 and IEEE-AES Systems Journal) Publ. #55 of

[12] “Keeping Pace with Avionics Innovations” (NAECON 1977)

[13] Collision Avoidance By Deceleration

[14] Runway Incursions

[15] “Simulation Model to Evaluate Collision Avoidance Methods using Raw Measurements in the Automatic Dependent Surveillance – Broadcast,” (coauthor; presented by Maarten UijtdeHaag), ION-GNSS-11

[16] “Unfinished Business: Glaring Absences from the Achievement List,” IEEE PLANS-04

[17] Single-measurement RAIM

[18] GPS Receiver Mechanization Pioneered at Ohio University



[21] Dixon, C.S., Hill, C.J., Dumville, M., and Lowe, D., “GNSS Vulnerabilities: Testing the Truth,” Coordinates VIII n3, March 2012, pp. 13-20.

My Coordinates
Starting another decade…
No Coordinates
Drawing ‘the difference’
Mark your calendar
April 2012 TO November 2012

1 Star2 Stars3 Stars4 Stars5 Stars (4 votes, average: 1.25 out of 5)

Leave your response!

You must be logged in to post a comment.