Precision Landing

[reedchristiansen]

Hello,

I thought I would show you guys some of recent work. We stuck a laser range finder in a 48 inch flying wing. The goal is to hit a lat/lon/alt point on the ground with some precision. Of course for peacfull purposes (Delivering food to refugees really acuratly... :) )

Here is the resulting video:

www.procerusuav.com/video/PrecisionLanding.wmv

The landing point is specified in lat/lon/alt. The standard deviation for touch down is under 5 meters. You can see quit a bit of aircraft motion in the video, this is due to turning up the gains to get tight control (maybe too tight).

Reed

[kd7ost]

COOOOOL,

Dan

[Tom Harper]

Impressive - how was the range finder connected to your control system.

Tom

[reedchristiansen]

Tom,

The range finder was pointed mostly straight down. It provided relativly long term(3 second time constant) corrections to the barometric altimeter.

Reed


Impressive - how was the range finder connected to your control system.

Tom

[icebear]

Really impressive!

Any more detalis (please!)

/Bjorn

[Tom Harper]

Thanks....I'd like more details also. Did you tap into a commercial range finder for the signal?

[reedchristiansen]

Here are some more details:

We used an oem range finder. It is big and expensive ($700, 5-6 ounces), but it has a serial output. It gives data at 10 hz and is accurate to about .5 meters.

The control is pitch for altitude and roll for heading. Heading is computed using the course to landing point. Xtrack error provides a heading bias.
throttle controls airspeed, but the aircraft is in an overspeed situation during most of the glideslope (airbrakes would be nice to steepen things up a bit).

The altitude is computed from the glideslope. The glide slope terminates at the touch down point. The motor is shut off when the aircraft is withen 10 meters of touch down point. Typically the motor is off anyways because the aircraft is above desired airspeed due to the "sled ride" down the glide slope.

Shallow glide slopes work best (this is counter intuitive). Due to the laser range finder, most of the error is lateral. Before we had the laser range finder, most of the error was longitudinal (due to baro drift), so steeper glideslopes worked better.

Reed



Thanks....I'd like more details also. Did you tap into a commercial range finder for the signal?

[treehog]

Also interested in that kind of problem so this link supplied

http://www.rcgroups.com/forums/showthread.php?t=589533

might help out as extra back up if there is lazer beam splatter issues from fog or smoke

also marine boats sonar sender equipment I found can detect fairly accuratly the distance to ground from about two feet high but you would need to miniturize it

keep up the good work

how much do you recon the whole lazer rig out costs

the AP alone is my annual modelling budget eeeekkkkk


Ralf

[Tuner]

Yah Laser range finders can have a wide range of Problems.
I started to design my own and learned the hardway just how difficult they are to get working properly.

One easy solution is to use a NON-Eye Safe system and that is impratical.

Another is to pay good money for a systems that is designed to work on many surface textures and colors. Black is usually the one surface color that kills these systems. Also Daylight can cause some issues.

Good Luck! 5-6 ouces is not that bad but I was woundering if an Ultrasonice Range finder would have better performance and lower weight??? Ive never used one?

Scott

[reedchristiansen]

It would be great if ultrasound would work. It is a much prefered solution because of size, weight, and cost.

After much fideling, we gave up. It was unreliable over different surfaces and gave bogus data at times.

I have tried allot of stuff - here is my list of best to worse for height above ground sensing:

1. Laser - This is WAY better then the rest.
2. Optic flow (mouse sensors re-tooled with wider aperature lenses)
3. Ultrasound from mr robot (panasonic unit).
4. Ultrasonwith with seperate send and recieve.
4. homemade proximity sensor using laser pointer and ccd array (similar to sharp proximity senors).

A tough problem. Someone needs to convince DARPA to dump a couple $mil of our taxpayer money into making a micro laser range finder.

Reed


Yah Laser range finders can have a wide range of Problems.
I started to design my own and learned the hardway just how difficult they are to get working properly.

One easy solution is to use a NON-Eye Safe system and that is impratical.

Another is to pay good money for a systems that is designed to work on many surface textures and colors. Black is usually the one surface color that kills these systems. Also Daylight can cause some issues.

Good Luck! 5-6 ouces is not that bad but I was woundering if an Ultrasonice Range finder would have better performance and lower weight??? Ive never used one?

Scott

[Unterhausen]

Reed, that movie looked really good. But everyone wants to know if you are the manly-man on the left that has confidence in his work, or the wimp on the right that hits the deck on every landing?

[Tuner]

Your item 2 Optical Flow.
I have been studying this for 3 years and the best I can tell Optical flow is by fare the best methode for this however it is probably the most complicated to get tuned and working.
I would recomend using the Ace 16k Camera Sensor. This chip was designed specifically for this type of application. Analogic-Computers and Anafocus are the two companies to involve in this.

The only reason I am so enthusiastic about optical flow is it can help with the AHRS aka autopilot or autonomous flight as you can use the optical flow to increase the accuracy and resolution of the AHRS.

My 3 cents.

[reedchristiansen]

Reed, that movie looked really good. But everyone wants to know if you are the manly-man on the left that has confidence in his work, or the wimp on the right that hits the deck on every landing?


I would be the manly man on the left. My partner Josh bails about every time.

Reed

[reedchristiansen]

Optic flow is great, but it requires lots of knob tuning like you mentioned.

we are using specialy designed optic flow chips (the ones from optic mice). It is all self contained in the airplane, no ground station involvment. The optic flow update rate is variable depending on height above ground - anwhere from 2 hz to 20 hz.

The problem with optic flow that we have experienced is:

1. too close to the ground and the registers fill up - you cannot read the sensor fast enough.

2. it does not work well on smooth surfaces, and many surface look smooth above 10 meters (asphalt, for example is virtually featureless above 5 or 6 meters even with a 2 degree lens.

3. it does not work at night.

Laser does not suffer from any of these problems. Of course it is about 10 times heavier, and $800 instead of $8

A few more cents.

Do the chips you mentioned have built in processing or do they require additional cpu?

I would like to hear more about the work you have done.

for small mavs, I think optic flow is the only solution.


Reed




Your item 2 Optical Flow.
I have been studying this for 3 years and the best I can tell Optical flow is by fare the best methode for this however it is probably the most complicated to get tuned and working.
I would recomend using the Ace 16k Camera Sensor. This chip was designed specifically for this type of application. Analogic-Computers and Anafocus are the two companies to involve in this.

The only reason I am so enthusiastic about optical flow is it can help with the AHRS aka autopilot or autonomous flight as you can use the optical flow to increase the accuracy and resolution of the AHRS.

My 3 cents.

[brianhomer]

woowee..
I like the last landing. left hand side (I don't know who's that man). Seems very confidence about the precision landing.

Is good and hope to share more!

[Birddog]

Are you flying at Spanish Fork Airport?

[Vindication]

Reed-

What about just open loop landing using altitude information from map files along with the barometric altimeter?

-Vind

[reedchristiansen]

Are you flying at Spanish Fork Airport?


We fly by the Kawasaki dealership south of Spanish Fork Airport. We are close enough that we have to be real carefull not to encroach on the 2 mile boundery. we are also hemmed in on the east by the freeway. But lots of swampland out west and north to crash into.

Reed

[reedchristiansen]

Reed-

What about just open loop landing using altitude information from map files along with the barometric altimeter?

-Vind


Vind,

That is our standard landing. However, because of baro drift, the altitude could easily be off more then a few meters. Because of the shallow glide slope, this causes errors of 30 meters or more.

For really tight landings, it is nice to have something pinging the ground.

Reed

[Unterhausen]

Too bad the Hokuyo laser scanner weighs so much and has a distance limit of 5.6m. The previous 4m limit has been increased. They also say it is only for indoor use. We have run ours outdoors and it seems to do fine. If I remember right, it does ok with black, but shiny objects drive it crazy. I would like to put one on a plane though.

[treehog]

i know a radar gun is realitivly heavy but if you carried that on a UAV that would probaly be the best awnswer

Also might be able to miniturize it

Those type of solutions I presume are to cater for a location where you dont have or dont wish to use a ground based helping solution

I hink if one checked into it it should be possible to make a solution that helps for spot landing if the presision landing equipment was on the ground like some form of ground based AGS assisted glide slope

It might be possible to have a lazer beam or micro wave beam on the gound guiding the AUV glide slope better

Also the GPS system in USA can be enhanced a lot with a extra input from a geostationary satilite that improves accuracy to some few metres for the GPS ( service not available outside north america) and that might be eneogh without to spend $800 on a onboard LAZER


I am looking into robotics and I found a link which suggests sterio vision capabilties are possible with ordinary PC (how good I dont know )

http://www.leafproject.org

subset of forum descussion robotics
http://www.alsrobotics.co.uk/ and http://www.robotcafe.com/ where thre might exist other similar links

subset from ezone thread
robotics forum http://www.rcgroups.com/forums/showthread.php?t=577677

and similar theme simulator forum http://www.rcgroups.com/forums/showthread.php?t=458466

so miniturising this down could be the way forward in the long run mayby with web laptop wifi uplinks to do the realy number crunching work initialy then later onboard baby boards (saw a 300mghz board size of pack of cigerates so ultra small m?B exist that run ordinary MS or linux )


Anyway I shall continue to pursue slightly cheaper solutions like pop a drag chute if I risk overshooting as my budget for UAV are lot less than $1000 as a handy add on feature


Ralf

[Vindication]

Vind,

That is our standard landing. However, because of baro drift, the altitude could easily be off more then a few meters. Because of the shallow glide slope, this causes errors of 30 meters or more.

For really tight landings, it is nice to have something pinging the ground.

Reed

Reed-

What about using a sensor like: http://www.sparkfun.com/commerce/product_info.php?products_id=8128 which claims to have 17 bits of resolution (9 cm of accuracy). Filtering between the altimeter and the GPS shouldn't allow the drift to be so bad during the landing-- should it?

[Vindication]

Another option would be to put an altimeter in your groundstation- you may be able to compare the altitude between the plane and the groundstation.

[reedchristiansen]

Another option would be to put an altimeter in your groundstation- you may be able to compare the altitude between the plane and the groundstation.


Not a bad idea. I am not sure where most of our drift is coming from. Since we charicterize each sensor in a temperature chamber, I assumed that most of the drift is due to the actual atmosphere changing. A similar baro sensor close to the landing point would help null that out.


Definatly worth looking into.

Thanks,
Reed

[senosy]

It is cool and you can go further with this.

[treehog]

http://www.maplin.co.uk/Search.aspx?criteria=park&menu=1&WorldSearch=Y&doy=11m1&ShowPics=true&Stock=28&MinPrice=0&MaxPrice=200&SD=true


these devices go on car rear bumper and beep faster as you get closer to objects

this could be changed to feed back into plane and cause flare out from steeper dive

Ok might not beep so good over tall grass or wet ground
but figure over close cut grass or tarmac conctert road should be suffiennt

A lot cheaper than Lazer :D


but if I had the money yeah i would get the LAZER as well :D

Ralf

[wsmalley]

The idea of auto-landing has fascinated me for a long time, but I've never seen much interest from the electonic gurus. I bought one of the cheap garage ultrasonic gizmos which has red, yellow, green lights as you approach. Thought about how one might hack into that. I found a site once-never could find it again-where the Germans had done auto landing in 1939 with 'state of the art' radio gear. Anyone know how, or what type of gear, the military UAV's are using for their stuff? Enjoyed the video, glad someone has finally saddled that beastie! It can only improve from there, I think.

[_helitron_]

Hi guys,

what's about this new device from Bosch:

http://www.bosch-pt.com/productspecials/blue/dle50/de/de/function/index.htm

(unfortunately in german but I think there is an english site also anywhere)
this is an incredible tiny laser distance measuring device for only € 177,- with max. distance of 50 m and an accuracy of 2 mm. Weight is approx. 170 g with case. I don't know if there is an interface to outside world on the tiny one but there are a bit bigger devices with interfaces.

Maybe it helps :) ...

Erwin

[Tom Harper]

Malley,

You might start with something like the old AN range system. You could use a pair of the small xmtr/rcvr pairs that SparkFun sells for about $12. They are supposed to get 300' with clear line of sight and proper antenna orientation. You would set one on either side of the field. They then transmit synchronously on the same frequency. The transmitter on the left transmits an 'A' = dot dash. The transmitter on the right transmits an 'N' = dash dot. So if you are to the left of the glide path you hear an A, to the right you hear an N. On the glide path you get a solid tone. I suspect the Germans used something similar. It was hot stuff in the forties.

In this digital age we might improve on the technology without adding cost. At 300MHz a couple of dipoles shouldn't be too difficult. The signals would be easy to sort out with a cheap microcontroller.

For ranging you could use a ground based unit, in the center of the runway, that simultaneously transmits an RF and ultrasonic pulse. The rcvd time difference between the two will give you distance to touchdown. The range should be 100 ft plus since you are line of sight and the sonic pulse does not have to bounce off of anything. GPS altitude would be sufficient to calculate initial glide angle.

With position and altitude from GPS you should be able to align with the runway, drop the throttle and settle onto the ground. Err...that's the idea anyway.

The military has infinite dollars and many bands in the spectrum. They have some kind of ILS that gives them glide angle. They live in a different world.