new gps integrated Inertial Navigation Unit

[cobaltdan]

http://www.atairaerospace.com/inu/

gps + triaxial magnetometer + triaxial accelerometer + triaxial gyro + barometric + microprocessor + extended kalman filter :)

all in a package slightly bigger than a match box :eek:

[Unterhausen]

That package is very similar to the Xbow micro-nav.

[sesat]

How much? Unter, how much is the Crossbow unit? I have written to AT Air, just hoping someone else has done so already.


Ram.

[Unterhausen]

Ram,
the crossbow is $1500. I've been looking at Sparkfun wondering if I want to buy their inertial nav unit for personal use. Still $500 once you add in gps
Just noticed that you have to add in a magnetometer with sparkfun too. And there are no pressure sensors.

http://www.sparkfun.com/shop/index.php?shop=1&cart=410062&cat=71&

[cobaltdan]

the atair inu is about $2000

it has superior performance to the crossbow unit and is significantly smaller and lighter, outperforming units on the market in the $8k + range


the xbow is 3"*3.75"*3" and weighs 1.3 pounds compared to the atair unit which is 1.50” x 2.00” x 0.75” with a weight: 1.6 oz

additionally the atair unit incorporates a barometric altimeter and serial bus for fusion of additional sensors, e.g. thermal leveling, doppler, etc...

access to source code is also granted with development agreement.

[Unterhausen]

actually, the crossbow micro-nav weighs about the same. Crossbow makes a lot of INUs, this isn't one of the ones in the heavy gold boxes.

[sesat]

Dan, ATAir has responded to me that their INU costs $4k. What would someone need to do to get it for $2k?

The xbow unav code is open to users too.


Ram.

[cobaltdan]

A decision was recently made to offer evaluation units at a higher volume price point to facilitate new customer and product development. The exact price of the Atair INU is $2375 each. Rick Zaccarri is the sales contact for this product at Atair Aerospace and can be reached at 718-923-1709 or rzaccari@atairaerospace.com

[OregonGPSINS]

I have reviewed both products and find the Xbow to be far, far superior.

For background, I have worked extensively with the Xbow micro-nav in many UAV applications and there isn't a better product on the market. Very robust, small and inexpensive. Support is very good, if they call you back.

[danstrider]

Very nice looking unit cobaltdan! Thanks for sharing.

Dan

[Unterhausen]

For background, I have worked extensively with the Xbow micro-nav in many UAV applications and there isn't a better product on the market. Very robust, small and inexpensive. Support is very good, if they call you back.
Are you using this with a stargate? I have found very little information about using the uNav, so I figured that I'd basically have to do everything myself. One of the problems is that googling for Stargate and UAV is a non-starter, since they use UAV's on the tv show :eek:

[air]

What would the user need along with the Xbow unit to get to a working UAV platform?
How would it compare with the U-Nav 3300 for example?

[Unterhausen]

What would the user need along with the Xbow unit to get to a working UAV platform?
I don't think the ATMega128 on the xbow uNav has enough horsepower to run a flight control, although I may be wrong. If I'm right, you need a processor with 2 serial ports that can run your autopilot. It's built to hang off of a Stargate, which is an Xscale processor. The minimum cost of a Stargate is over $600.


How would it compare with the U-Nav 3300 for example?

I don't know, you have a link?

[air]

Ok, the 3300 is a fully featured ready to go autopilot:
http://www.u-nav.com/3300.html

[Bg~]

I don't think the ATMega128 on the xbow uNav has enough horsepower to run a flight control, although I may be wrong

The Paparazzi project uses the ATMega128 to run nearly all of their flight control stuff (minus PPM decoding and other servo stuff).

[danstrider]

The Xbow and the Atair products are both inertial measurement units. That is, they both give you the attitude of the sensor (presumably bolted to your craft) based on reading the rotations and accelerations of the unit. The Atair product goes a couple steps further and uses GPS for actual position and magnetometers for actual heading combined with the accelerometers and gyros. The Kalman filter fuses all these pieces of data together and the end result is a single location and aircraft state.

Inertial navigation alone builds up inaccuracies because the unit simply integrates the roll/yaw/pitch and accelerations to find your current aircraft state. When integrating, recall from your calculus days that a constant is tacked onto the end and it is this constant that builds up over time.

A submarine would use an INU sensor to determine its position after being submerged for a long time without access to GPS updates or stars (call back to the old days :-) For aircraft, an INU is possible, but inertial navigation can be supplemented with even a periodic "actual position" update from GPS. Aircraft such as the C-130 used inertial nav to fly w/o GPS in bad weather and carried such sensors as the Xbow.

Why not simply use only GPS you ask? GPS only updates at a finite rate, typically 1Hz (once per second), though smaller and better GPS units are updating at 10Hz advertised. During this one second delay, your aircraft could turn, flip upside down, and reverse course and your GPS unit would still suggest your aircraft was travelling in a straight line at an instant before the next update. Also, GPS heading is calculated as a straight line between the current and previous update; this method is not correct in a steady turn. So GPS has limits of a low update rate.

Kalman filtering is worthy of its own graduate level class. This filtering method is used to combine sensor data and help reduce the inaccuracies of using a single sensor. There is some open-source class documentation online if you're interested in reading more about this filtering method.

In summary, the GPS & magnetometer corrected INU gives you the exact location, heading, and attitude of your aircraft at any time, corrected for integration inaccuracies and provides high-quality estimates between GPS updates.


So to answer the real question, this box gives position and attitude data over a serial line. You would still need a processor to crunch these numbers and use them in a control algorithm for the aircraft. At this point, you're building your own autopilot :-)


If you're still interested and reading, but lack the budget to play with such expensive toys, I suggest hooking up with a controls company or a university and find someone else who shares your passion!




Someone with more knowledge double check me; nobody is perfect, nor would I remotely claim to be a controls expert ... a trained amateur perhaps?

Dan

[Bg~]

When integrating, recall from your calculus days that a constant is tacked onto the end and it is this constant that builds up over time.

I think this is not the case. I think as long as the INU is started from a known reference condition, this integration constant is not important. The cause of drift in INU measurements is due to gyro bias change (gryos are accurate in the short term, their measurements drift over the long term, and can change with temperature), noise in the measurements, and possibly sampling effects in the analog-to-digital conversion (I'm just guessing here for this one).

If you had perfect measurements, integration of angular rates would be perfectly accurate.

Also, there has been some university projects (Georgia Tech I think) that used a GPS only autopilot. The requirement is that your airplane must be very stable, since this type of autopilot can't provide any real stability, only navigation. I think there also has been some experiments on determining aircraft attitude based on two separated GPS antennas (say on each wingtip) by comparing the phase difference of the GPS carrier signal (way over my head).

[Unterhausen]

The Atair product goes a couple steps further and uses GPS for actual position and magnetometers for actual heading combined with the accelerometers and gyros. The Kalman filter fuses all these pieces of data together and the end result is a single location and aircraft state.
Dan

Actually, the xbow has magnetometers and a gps, the Kalman filter was intended to be implemented off-board. As Bg notes, it's not the constant of integration that kills you, it's the bias in the reading which integrated over time gives you an infinite error. You can estimate this error and reduce its effect considerably, but of course the best way is to correct it from the gps. Crossbow provides a test application that combines the measurements somehow (probably not a kalman filter) and the main noise source on ours is one of the magnetometers.

I know I'm sounding like an advocate of the xbow, other than we have one at work and I'd rather not go it alone getting it to fly. It seems to me that people should disclose any commercial interests when they post here.

Bg, that's a good point about the Papparazi project using an ATMega, but they have essentially closed the loop around some thermopiles with a pid controller, so the processing is considerably less.

[Bg~]

they have essentially closed the loop around some thermopiles with a pid controller, so the processing is considerably less.

Ahhh, I didn't even think about that. :) I did see a paper at the recent AIAA Infotech conference for a 8bit processor controling some type of quadrotor equipped with an IMU. The trick was not to use a Kalman filter, but just PID loops to get kinda the same effect at much less processing reuquirement. It seems to work well. (the details are fuzzy)

Re: xbow... I'd love to get one of these to work with, mainly since its open source.

[OregonGPSINS]

My associate at Southwest Research Lab (SWRI) notified me that ATAIRs guided parachute systems, which are driven by this very same INS, failed miserably at a big demonstration last week in Arizona. He said that over 80% of the parachute systems went haywire and flew off into never never land. Thats a lot of money to be wasting. Looks like I'll be avoiding these units like the plague.

[Unterhausen]

...ATAIRs guided parachute systems, which are driven by this very same INS...
That's an interesting data point, but it doesn't really mean that they got lost, their control strategy may not be too great. I can see all sorts of reasons why a guided parachute might fail besides the inu. Makes you wonder how much they tested before their demo.

This has been a very weird thread.

[cobaltdan]

....

[cobaltdan]

My associate at Southwest Research Lab (SWRI) notified me that ATAIRs guided parachute systems, which are driven by this very same INS, failed miserably at a big demonstration last week in Arizona. He said that over 80% of the parachute systems went haywire and flew off into never never land. Thats a lot of money to be wasting. Looks like I'll be avoiding these units like the plague.


yes we had 4 of 8 systems not perform well at a recent demonstration. this was quite new to us as we have not missed a target in over several hundred drops. the preliminary analysis indicates that a new mounting component acted as shock amplifier and overranged our accelerometers on parachute opening form the c130. this was not noticed earlier as we test at 130 knotts but at this demonstration we were droped at 150-170 kias.

the problem has now been fixed.

sincerely,

daniel preston
atairaerospace.com