Author Topic: Are accurate clocks really the limiting factor in cheap Inertial Navigation?  (Read 8591 times)

0 Members and 1 Guest are viewing this topic.

Offline rs20Topic starter

  • Super Contributor
  • ***
  • Posts: 2322
  • Country: au
Hi,

In this article, in which scientists have apparently developed some sort of portable atomic clock, they claim that with this clock, the need for GPS satellites is rendered obsolete:

Quote
Dr. Alessia Pasquazi from the EPic Lab in the School of Mathematical and Physical Sciences at the University of Sussex explains the breakthrough: "With a portable atomic clock, an ambulance, for example, will be able to still access their mapping whilst in a tunnel, and a commuter will be able to plan their route whilst on the underground or without mobile phone signal in the countryside. Portable atomic clocks would work on an extremely accurate form of geo-mapping, enabling access to your location and planned route without the need for satellite signal.

Now of course you need more than just the time to figure out one's location; on the surface these claims make no sense. I sent an email to the author for clarification, and got this response:

Quote
Long story short, precision clocks are needed for what is called 'inertial navigation', where you map your position with time using accelerometers.  ships have used for a long time, and a  version of this navigation is already on our phones. it is how they get the direction when you start the navigation.
because you calculate the path by measuring time and acceleration/angular moment, if you do not have a good clock you make a lot of errors, and this is why portable atomic clocks are needed. exciting  research by our colleagues on quantum  gyroscopes should make this thing even better, but the first bottleneck at this point is to make a portable high accurate clock.

So in short, the claim is that we could have dead reckoning, non-satellite backed INS systems in our cars (maybe even mobile phones), except for one hurdle: modern (quartz?) clocks being too inaccurate. My intuition/understand was the MEMS 6-axis sensors, while amazing feats of engineering, aren't all that accurate and certainly aren't hampered by the inaccuracies presented by a quartz clock.

Is my understanding out of whack? Or are the claims in the link completely unreasonable?
 

Offline PA0PBZ

  • Super Contributor
  • ***
  • Posts: 5259
  • Country: nl
Even when the time is absolute correct then still the smallest error from the accelerometer will keep multiplying, so I don't see how this is going to work. If you use it in a car and keep correcting based on where the roads are.. .maybe.
Keyboard error: Press F1 to continue.
 
The following users thanked this post: rs20

Offline Berni

  • Super Contributor
  • ***
  • Posts: 5050
  • Country: si
Does not make sense to me too.

Sure time is a factor, but we can already make very precise TCXOs that are easily orders of magnitude more accurate than needed. But accelerometers and gyroscopes are way too noisy and drifty to make this work. Acceleration integrated gives you velocity and velocity integrated gives you position. Any slight offset error in the accelerometer causes the velocity integral to drift away in time, and since this feeds into another integrator it means it will drift faster and faster until it just flies off towards infinity. Its an incredibly unstable system.

We do make use of intertial navigation tho. Intercontinental missiles for one have a dead reckoning system that allow the missile to keep on course even if all outside navigation signals are jammed. But they use very fancy accurate accelerometers and special light interferomentry gyroscopes to get it accurate enough to not drift off in the few minutes it takes the missile to hit the target.
 
The following users thanked this post: rs20

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 5569
  • Country: us
In every navigation application I have encountered clock errors are several orders of magnitude below the other error sources.  By my rough estimate even the incredibly exotic inertial navigation systems used in submarines have at least an order of magnitude between the clocks and other error sources, though I don't really have any definitive information on the components of those systems.

While he didn't admit it to you, the professor may have one real application up his sleeve which he is hiding for proprietary reasons.  A really good clock can be used to estimate altitude.  Which combined with a known starting position, good geodata and a crude inertial system could give you excellent results.  Unclear how much better they would be than a similar system with just the known start, good geodata and a crude inertial system.  My GPS units do a pretty good job of estimating position in tunnels based on these types of inference. 

But my own prejudice, based on many encounters with research professors, is that they are often incredibly naive outside their specific expertise and that he had never actually dealt with real world errors in navigation systems.  My favorite example of this naivety came in a grad class in the early 1980's.  The prof was actually a noted expert in his field and as well as I could judge the reputation was well earned.  But my respect for his overall knowledge came when he announced in class that there was a new invention out, something called a microprocessor that might someday allow data operations in small fielded units.  Many of us in the class were already employed in industry and had fielded such devices years ago.  The prof was at least three product cycles behind the curve.
 
The following users thanked this post: rs20, Frank

Offline StillTrying

  • Super Contributor
  • ***
  • Posts: 2850
  • Country: se
  • Country: Broken Britain
But accelerometers and gyroscopes are way too noisy and drifty to make this work.
...
very fancy accurate accelerometers and special light interferomentry gyroscopes

I'd agree that some other very accurate other stuff is needed as well as the accurate time. Perhaps they haven't figured them out yet so they thought it best not to mention them.

"which range from partnerships with the UK aerospace industry, which could come to fruition within five years,"

Sounds like they're after another 5 years of funding!

"We will be starting work on these projects on 1st December 2014 (i.e. next Monday) and the initial funding is for five years."
https://telescoper.wordpress.com/2014/11/26/quantum-technologies-at-sussex
« Last Edit: March 25, 2019, 03:35:55 pm by StillTrying »
.  That took much longer than I thought it would.
 
The following users thanked this post: rs20

Offline IDEngineer

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Accurate, long term dead reckoning is a staggering technical challenge. Even the most accurate (read: and most expensive) inertial navigation systems have some sort of recalibration technique back to a reliable reference. Spacecraft and military aircraft, for example, have (had) optical systems that reset their INS's based on the positions of reference stars. I'm at least as impressed by the design of those optical miracles as I am by the INS's they are (re)calibrating. If INS's with those kinds of budgets still weren't accurate enough over time to not require external references, I'm confident no MEMS sensor is either. At least today.
 
The following users thanked this post: rs20

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10031
  • Country: gb
You can't render GPS satellites obsolete until your atomic clock, and all the kit needed to go with it, can match the tiny power consumption of a current GPS receiver. People have come to expect a really high level of portability.
 
The following users thanked this post: rs20

Offline Berni

  • Super Contributor
  • ***
  • Posts: 5050
  • Country: si
Tho it does make me wonder how hard is it to make a optical fiber interferometry gyroscope. Perhaps something along the lines of shining the laser head of a CD drive into a coil of fiberoptic cable since CD drive laser assemblies are pretty much cheap tiny interferometers.
 
The following users thanked this post: rs20

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10031
  • Country: gb
Tho it does make me wonder how hard is it to make a optical fiber interferometry gyroscope. Perhaps something along the lines of shining the laser head of a CD drive into a coil of fiberoptic cable since CD drive laser assemblies are pretty much cheap tiny interferometers.
Most optical gyros don't actually use fibre, because they can achieve greater stability by reflecting light off mirrors. The base plate and mirrors form quite a heavy assembly, to achieve the rigidness needed for accurate navigation.
 
The following users thanked this post: rs20

Offline IDEngineer

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Ring gyros are a fun topic. Their challenge is loss of registration due to extremely slow rotation rates... there is always some rate of rotation small enough that the pattern shift is smaller than the resolution of your sensor.
 
The following users thanked this post: rs20

Offline IDEngineer

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Most optical gyros don't actually use fibre, because they can achieve greater stability by reflecting light off mirrors. The base plate and mirrors form quite a heavy assembly, to achieve the rigidness needed for accurate navigation.
There are versions that use a single piece of Pyrex, with the mirrors silvered directly onto polished faces in the proper alignment. Solves quite a few challenges with respect to thermal stability, etc.

However, my earlier comments regarding the sensor still apply. That's where the real challenge lies, in having sufficiently high enough sensor resolution to not lose registration due to very low rotation rates.
 
The following users thanked this post: rs20

Offline Doctorandus_P

  • Super Contributor
  • ***
  • Posts: 4004
  • Country: nl
Inertial navigation was accurate enough to hit London from the west coast of the Netherlands in WW2.
The amount of drift you can expect of "modern" sensors is related to price you pay for them.

One way you could copensate for drift in road traffic is to map the accumulated data to the road map, ans assume you can only travel over the roads.

If you accumulate enough left and right turns you could even get the location from that even without prior knoledge about the starting position.

No atomic clocks needed.
 
The following users thanked this post: rs20

Offline jc101

  • Frequent Contributor
  • **
  • Posts: 729
  • Country: gb
There is a reason why part of the checklist in aviation is to verify the position of an aircraft before departure, all gates have their specific Lat Long available as cross check.  This is to verify the aircraft INS (Internal Navigation System) which does dead reckoning has a known staring point.  It will drift during the flight.  The INS is one navigation system, and it combined with radio Nav beacons and GPS.

 
The following users thanked this post: rs20

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10031
  • Country: gb
Inertial navigation and time consuming journeys don't go together very well. An ICBM can reach its target accurately by inertial navigation. A bomber needs corrections, manually applied, along the way. A cruise missile uses inertial navigation to move between way points, where it uses a ground mapping radar to realign itself against a stored map.
 
The following users thanked this post: rs20

Offline ejeffrey

  • Super Contributor
  • ***
  • Posts: 4032
  • Country: us
Ring gyros are a fun topic. Their challenge is loss of registration due to extremely slow rotation rates... there is always some rate of rotation small enough that the pattern shift is smaller than the resolution of your sensor.

That isn't quite the problem.  The detectors have plenty of resolution to detect a small fraction of a fringe.  The problem is that at DC, the two modes (clockwise and anticlockwise) are close enough in frequency to cause injection locking of the laser from scattered light or other means of crosstalk, This leads to zero signal.  The usual (partial) solution is to mechanically dither the gyro. This isn't perfect and also introduces a signal you have to filter out.
 
The following users thanked this post: rs20

Offline RoGeorge

  • Super Contributor
  • ***
  • Posts: 7011
  • Country: ro
If you have a precise clock and are allowed to look at the stars to recalibrate, than it's possible to find out the precise location.  It was a huge prize for a precise enough clock (or other method to navigate in the open sea, without following the shore line).  Many methods were presented.  One of them was looking at the star and precisely knowing the time, except there were no accurate enough clocks back then.

Very long but interesting documentary/dramatization:



In theory, an extremely precise clock can be used for inertial navigation (without reference stars).  In practice, it won't be easy to beat the GPS.
« Last Edit: March 25, 2019, 06:37:01 pm by RoGeorge »
 
The following users thanked this post: rs20

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 5569
  • Country: us
Clocks were a huge problem a couple of hundred years ago.  Not so much today.  The magnitude of the time error in celestial observation is roughly the rotation rate of the earth surface in the celestial reference times the time error.  More or less 500 meters per second of error.  For a garden variety quartz wristwatches that means about 250 meters of position error per day since last setting of the clock.  For high precision wristwatches that becomes something more on the order of 500 meters per month.  Go to a commercial Rubidium standard and you cut those errors by another couple orders of magnitude.

For an accelerometer to contribute the same 250 meters per day as the garden variety wristwatch it has to have offsets and gain errors on the order of 3E-08 g.  Such accelerometers are not garden variety.  Similar story for rate sensors.

For those who like laser ring gyros look up Sagnac effect. 
 
The following users thanked this post: rs20

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
Quote
exciting  research by our colleagues on quantum  gyroscopes should make this thing even better, but the first bottleneck at this point is to make a portable high accurate clock.
I read it: there might be some super duper technology in the future which certainly wouldn't be possible without our invention and although it may still not happen with our invention if it does happen it will totally be some awesome stuff. And we gotta invent some impressive sounding justification for our research to make beancounters happy.
 
The following users thanked this post: rs20, Siwastaja

Offline Mechatrommer

  • Super Contributor
  • ***
  • Posts: 11714
  • Country: my
  • reassessing directives...
Inertial navigation was accurate enough to hit London from the west coast of the Netherlands in WW2.
The amount of drift you can expect of "modern" sensors is related to price you pay for them.
true only if path is near straight line or nice parabolic as errors dont accumulate that much. retry with zig zag motion, the rocket may end up in seabed or top mountain desert, if not on civilians occupied area. as mentioned, even the most expensive spaceship will need correction based on landmark or beacon/gps. if this idea is correct, people should have already proved the concept with special vehicle setup with Cesium clock on board. the portable atomic clock maybe just a dream or another 21st century buzz. but then an "atomic gyroscope" maybe something else.

One way you could copensate for drift in road traffic is to map the accumulated data to the road map, ans assume you can only travel over the roads.
If you accumulate enough left and right turns you could even get the location from that even without prior knoledge about the starting position.
it may work as this is not purely dead reckoning. a simple AI, or path alignment can be embedded as driver can still see the road. pure dead reckoning may end up vehicle turning into grass field, not with a human driver.
Nature: Evolution and the Illusion of Randomness (Stephen L. Talbott): Its now indisputable that... organisms “expertise” contextualizes its genome, and its nonsense to say that these powers are under the control of the genome being contextualized - Barbara McClintock
 
The following users thanked this post: rs20

Offline StillTrying

  • Super Contributor
  • ***
  • Posts: 2850
  • Country: se
  • Country: Broken Britain
It seems strange that a University of Sussex dept. the Prof. and phys.org don't realize that a lot more than a miniature atomic clock is needed for the none-GPS ambulance or mobile phone navigation, but we all do!

Debunked. :)
« Last Edit: March 25, 2019, 08:41:39 pm by StillTrying »
.  That took much longer than I thought it would.
 

Offline barry14

  • Regular Contributor
  • *
  • Posts: 102
  • Country: us
A quick Google search turned up a portable cesium clock introduced by Symmetricon in 2011 that is the size of a matchbox, weights 35 grams and uses 115 mw of power. It's based on work done by NIST.
 
The following users thanked this post: rs20

Offline ejeffrey

  • Super Contributor
  • ***
  • Posts: 4032
  • Country: us
That said, one application I can see for cheap/portable atomic clocks is not really dead reckoning, but navigation based on local beacons where a GPS clock isn't available even to the beacons. So for indoor underground positioning you could deploy fixed local beacons and allow mobile receivers to triangulate relative to those beacons.  Or you could do the reverse for wireless network monitoring.  In a mesh wifi network, each client can be seen by a handful of APs.  They could in principle do triangulation to allow you to locate unidentified devices or rogue APs.  Like GPS you would still have to periodically update the clocks due to their non-zero drift, but a cesium clock would have less drift than an OCXO.

I am still skeptical.  Inexpensive portable rubidium oscillator already exist, and as I recall the stability crossover point between a Rb oscillator and a GPS signal is several hours which is plenty of leeway to keep your mesh network synchronized.
 
The following users thanked this post: rs20

Offline hamster_nz

  • Super Contributor
  • ***
  • Posts: 2812
  • Country: nz
The only use I can see is to augment GPS signals - if you know the time accurately then you have one less unknown in your solution.  Kalman filters and all that stuff. But going from a few ppm time error to 0 isn't going to change much.

When I was working on a system for trucks that used counting hubometers as an absolute distance reference, tire pressure and even tread wear were visible in the data.  The firmware had to calibrate the tire size when travelling in a constant direction for a long enough interval and long enough distance, while GPS fix was maintained, and could then use that value when GPS was lost.
Gaze not into the abyss, lest you become recognized as an abyss domain expert, and they expect you keep gazing into the damn thing.
 
The following users thanked this post: rs20

Offline tomato

  • Regular Contributor
  • *
  • Posts: 206
  • Country: us
It seems strange that a University of Sussex dept. the Prof. and phys.org don't realize that a lot more than a miniature atomic clock is needed for the none-GPS ambulance or mobile phone navigation, but we all do!

Maybe you and all the others that "know better" should read what the author wrote:

Quote
Long story short, precision clocks are needed for what is called 'inertial navigation', where you map your position with time using accelerometers.  ships have used for a long time, and a  version of this navigation is already on our phones. it is how they get the direction when you start the navigation.
because you calculate the path by measuring time and acceleration/angular moment, if you do not have a good clock you make a lot of errors, and this is why portable atomic clocks are needed. exciting  research by our colleagues on quantum  gyroscopes should make this thing even better, but the first bottleneck at this point is to make a portable high accurate clock.

 
The following users thanked this post: oPossum

Offline rs20Topic starter

  • Super Contributor
  • ***
  • Posts: 2322
  • Country: au
Maybe you and all the others that "know better" should read what the author wrote:

Quote
Long story short, precision clocks are needed for what is called 'inertial navigation', where you map your position with time using accelerometers.  ships have used for a long time, and a  version of this navigation is already on our phones. it is how they get the direction when you start the navigation.
because you calculate the path by measuring time and acceleration/angular moment, if you do not have a good clock you make a lot of errors, and this is why portable atomic clocks are needed. exciting  research by our colleagues on quantum  gyroscopes should make this thing even better, but the first bottleneck at this point is to make a portable high accurate clock.

The problem is, read what you bolded. The first bottleneck. That implies that this innovation is immediately applicable and useful on its own.

Regardless, even if you interpret that sentence differently, the entire original article gives absolutely no hint that there's another entirely new innovation required to provide the claimed benefits. It's essentially one group claiming the credit for hypothetical future work done by someone else.

Keep in mind that if you had a choice between getting the atomic clock in your phone OR this hypothetical quantum gyro made by someone else; based on what I'm seeing in this thread, you'd absolutely want the quantum gyro. The MEMS sensors are the first bottleneck, not the clock. Hence your bolded sentence is factually wrong.
 
The following users thanked this post: Siwastaja


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf