paparazzi-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Paparazzi-devel] Re: Paparazzi-devel Digest, Vol 68, Issue 30


From: Bart Remes
Subject: [Paparazzi-devel] Re: Paparazzi-devel Digest, Vol 68, Issue 30
Date: Sat, 14 Nov 2009 13:11:41 +0100

Paper and presentation availeble on

http://www.emav2009.org/

OptiPilot: control of take-off and landing using optic flow

Antoine Beyeler, Jean-Christophe Zufferey, and Dario Floreano


Bart Remes

Message: 3
Date: Sat, 14 Nov 2009 09:24:00 +0000
From: Gareth Roberts <address@hidden>
Subject: RE: [Paparazzi-devel] distance measurement for landing.
To: address@hidden
Message-ID: <address@hidden>
Content-Type: text/plain; charset="UTF-8"

I believe the paper I saw presented built on the work done by
Christophe.  The technique was radically different to the usual methods
(ccd cameras and fpgas/sbcs).
The sensors were from the bottom of optical mouses, cost nothing, were
incredibly lightweight and the control processor consisted on a
pic/atmega with about 20 lines of code, as the sensors simply give an
analogue output in line with detected motion (imagine how simple optical
mouses are).  He used 7 (or maybe 9) mounted across the bottom of the
aircraft.  I found the paper I was talking about
http://infoscience.epfl.ch/record/140575/files/emav09_optipilot_web.pdf
Although the presentation he gave also showed how the aircraft pitch and
roll were controlled, as a side effect.
I really think this is a great candidate for small/light/simple landing
systems, when combined with the known speed.
--G

On Fri, 2009-11-13 at 07:45 +0200, gisela.noci wrote:
> The problem with this method is that to an optical/visual sensor ( CCD
> Camera or the like) with the field of view,autofocus and dynamic range of
> the eyes would be quite costly and the optics might weigh a bit. Not really
> feasible for low budget home grown UAV,s. Slow flyers and rotary wing are a
> lot easier.
>
> The same problem with the moiré pattern system.
>
>  On one of our large UAV's we use a laser ranger with associated optical
> guidance, located at the landing point. We fly @ 140kmh, and 'land' by
> flying into a vertical net, no runway required. The A/C is 6meter wingspan,
> 145kg, and the net is 14meters wide by 6meters high. We hit the center of
> the net within a 1.5meter circle, every time...
>
> The A/C has a reflector on the nose, and the ground laser ranges the
> approaching A/C at 3km away. The laser system also has a camera which then
> 'acquires' the A/C and locks onto it, and expects the A/C to fly in on a
> specific trajectory, and relays flight control command up to the A/C via the
> datalink. Works very well, and is used where runways are impossible. This
> systems costs...70K  USD.
>
> Joe.
>
> -----Original Message-----
> From: paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> [mailto:paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> ] On Behalf Of antoine drouin
> Sent: Thursday, November 12, 2009 11:02 PM
> To: address@hidden
> Subject: Re: [Paparazzi-devel] distance measurement for landing.
>
> I believe this also exists (existed?)  for landing on ground, known as
> PAPI and VASI
>
> http://www.abbysenior.com/aviation/vasi&.htm
>
> also I had heard about this
>
> http://dspace.mit.edu/bitstream/handle/1721.1/37951/144589001.pdf?sequence=1
>
> they estimate not only distance and position to a target, but also the
> attitude of the vehicle using moiré patterns
>
> the problem with vision based system is that you're very dependant on
> environmental conditions, especialy lighting
>
> Regards
>
> Poine
>
>
> On Thu, Nov 12, 2009 at 9:27 PM, Elden Crom <address@hidden>
> wrote:
> >
> > Joe, your bumming me out on the altimeter route to precision
> landing....... Painfully you make a convencing argument....
> >
> > Perhaps another route. No cammera, only light intensities needed.
> >
> > Do you know how a/c land on an aircraft carrier (old style anyway)?
> > One of the systems that they use is to have a set of light and Fresnel
> lens on the deck of the carrier.
> > The pilot looks at the lights (the ball), if they are on the correct glide
> slope it appears green, too low or too high, left or right, it appears to
> them as a diffeent color. (not exactly right but for my purposes....)
> > (For a better explaination
> http://everything2.com/title/How+to+land+a+jet+plane+on+an+aircraft+carrier
> >  , http://answers.yahoo.com/question/index?qid=20071105145150AAYcwq8)
> >
> > Anyway,
> > Perhaps a simmilar system with a box of lights on the ground and 6 light
> sensors on the A/C.  Three colors for up and down. Threee colors for left
> and right. (when I say 'colors of lights' I proabably mean different
> frequencies of infrared.)
> >
> > Therory of Operation (only describing up and down for now,left and right
> will be similar):
> > The box of lights shines 3 beams of light. The beams diverge at ~3degrees.
> > The too high light is red.
> > The too low light is blue.
> > The just right light is green.
> > X RRRRRRRR
> > X      RRRRRRRR
> > X AC       RRRRRRRR
> > X GGGGGG       RRRRRRRR
> > X       GGGGGG     RRRRRRRR
> > X             GGGGGG   RRRRRRRR
> > X BBBBBBBBB         GGGGGG RRRRRRRR
> > X          BBBBBBBBB     GGGGGG RRRR[----Box---]
> > X                  BBBBBBBBB  GGGGGG[----of----]
> > X                          BBBBBBBBB[--Lights--]
> > The A/C approachs the landing site via GPS (+/-20meters=GPS + flight
> error?) at ~1 kilometer out the A/C sees the green only.
> > The navigation system swithes to moth mode (ie follow the lights).  If it
> sees red, more elevator; if it sees blue, less elevator; if it sees green,
> do nothing.  As the thing gets closer to the box, it will see pairs and
> maybe all three colors because of bleed. So basically just fly the plane
> such that Red and Blue have the same intensity.
> >
> > Come to think about it, the desire for the Green is more a human
> thing...perhaps just Red and Blue and fly the path that keeps them the same
> intensity.  (This will be all sorts of fun getting the PID loop constants
> adjusted!)
> >
> > Back into 3D space,
> > So that would just require, 4 light sensors with different color filters
> on them to be on the AC. Less than 20 grams? They would all be aligned the
> same (downward at about 3 degrees) so they could be flat on one PCB.  The
> box of lights could just be 4 flash lights with lens....maybe even 1 flash
> lights with a four color lens.
> >
> > Or even better, one ground light with a lens that is in thirds and only 3
> light sensors on the AC.
> >
> > It might even (maybe, perhaps, questionably?) be possible with just one
> color, but now the AC would be constantly drifting left/right and up/down
>  trying to cross the highest intensity portion of the beam, kinda like WiMax
> (WiFi's bigger bother) does in the time domain......but the SW algoithm gets
> a bit compilcated!
> >
> > Enough babling for now.....
> >
> > Elden
> >
> >
> >
> >
> >
> >
> > -----Original Message-----
> > From: gisela.noci [mailto:address@hidden]
> > Sent: Wednesday, November 11, 2009 12:58 PM
> > To: address@hidden
> > Subject: RE: [Paparazzi-devel] distance measurement for landing.
> >
> > Hi Eldon.
> >
> > Ok, had a look on the web site. Pity they do not show the other side of
> the PCB....Anyway, as I intimated, the only way they can achieve 140mm
> resolution is by amplifying the sensor, and limiting the range( as they say,
> from 50 to approx 2000meters)  to squash more bits per meter in. They
> certainly cannot have a 16 to 18bit A/D on that little PCB! The lower 4 to
> 5bits would be in the noise... And, most important, they talk of resolution,
> which is fine for maintaining an 'altitude' whatever that might be , but
> they do not mention the units accuracy and that is what is important in
> trying to use it for AGL determination for autoland. There is simply no unit
> accurate enough, and repeatable enough, to be able to use for autoland. In
> addition, the atmospheric model used for pressure- altitude calculation
> varies all over the world...
> >
> > That is why Pilots are told by control tower to set their altimeter to a
> specific 'altitude' in that towers controlled airspace. It might not be an
> accurate altitude, but at least all other aircraft in the same airspace are
> referencing the same 'inaccurate' altitude. Resolution is important, to
> insure you do not drift up or down without knowing it, but accuracy is so
> poor, that it is not used in absolute terms.
> >
> > It seems that sensor is 'read' by observing the led flash count, in which
> case the output rate is going to be VERY slow - seconds - rendering it
> useless even for altitude control.
> >
> > Our autopilot flies to pressure altitude, which is augmented in a kalman
> filter against GPS Alt to get some semblance of reasonable altitude. The
> sensor sample rate is at 60Hz, while the GPS is at 4Hz. The pressure Alt is
> used to determine a climb/descent rate, to feed the pitch/altitude loops.
> > So, sampling at a rapid rate is essential for good altitude control
> response.
> >
> > However, as to what sample rate is best for landing by means of pressure
> alt sensor, all I can say is - Don't.... The only way to land with a
> pressure sensor as your only reference is to sample  at least at a 50Hz
> rate, and to use that to derive a descent rate and then to approach from a
> long way, descend to  a 'safe' altitude, say 30 or 40 meters AGL at least,
> and then just keep wings level, speed constant at 3 or 4meters/second above
> stall, and fly in at a controlled descent rate of around 1 meter/second, and
> wait till you reach terra-firma. It works, I have done this a number of
> times, but you need space, lots of it. For example, at 16meter/second
> approach, with a 1m/sec descent rate you will cover 40X16 meters = 640
> meters. If the accuracy of the 'safe' altitude was such that is was actually
> 50 meters AGL, then you will cover 800meters......No magic bullet here, I
> fear. A 1/4 mile is to short for the above scenario; no room for error. If
> your approach speed is 5meters/sec then maybe, but at that speed, just catch
> it in your hand as it passes by......
> >
> > I am not sure what you mean by not wanting to use 'radar' because of
> sloping terrain; that is in fact the best reason to use sonar/radar, and
> then to fly at a fixed height above ground,ie, terrain following flight, and
> then descend when at the correct point, reducing height AGL at
> 1meter/second, flying an actual trajectory to the touchdown point. That is
> what we do, and we touch down at 20m/s ( a HEAVY plane!!!), within
> plus/minus 15meters each time of the designated touchdown point. That
> requires better than 1 second accuracy in descent rate, speed and pitch
> control.
> >
> > We do the same on a much bigger plane, 6meter wingspan and 145kg all up
> weight..
> >
> > And if you want to use radar to detect the height AGL of the flat patch,
> just use it to land with......
> >
> > I have watched a $300K UAV with a infrared camera in the nose try to land
> 3meters below ground level, at 4meters/sec descent rate - no more landing
> gear and droopy wing resulted.......
> >
> > But, don't give up. If you really want to try to do landing with a
> pressure sensor reference, I would suggest that you use a sensor such as the
> Honeywell ASDX015, a 4volt output for 100 to 1013millibar sensor that is
> temp compensated pretty well, amplify and offset the output to give you good
> resolution for your desired altitude range, make 2 identical units, keep one
> on ground and use any change on that one to correct the airborne one. This
> should give you a short term accuracy in the 1 to 3 meters range, and then
> land by approaching to 5 or so meters AGL, and then wings level, kill
> throttle and glide in. Works (almost) every time ..
> >
> > Happy landings
> >
> >  Joe
> >
> > -----Original Message-----
> > From: paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> >
> [mailto:paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> > ] On Behalf Of Elden Crom
> > Sent: Wednesday, November 11, 2009 7:44 PM
> > To: address@hidden
> > Subject: RE: [Paparazzi-devel] distance measurement for landing.
> >
> >
> >
> > Thanks Joe,
> >
> > The link must have clipped, the long way is http://www.hobbycity.com On
> the left side choose DataRecording Then ShadowWing Then How High
> >
> >
> http://www.hobbycity.com/hobbycity/store/uh_viewItem.asp?idProduct=6776&Prod
> > uct_Name=WS_HowHigh_Altimeter_Feet_&_Meter_(New_Version)
> >
> > I just go it in the mail, but I haven't had time to play with it much yet.
> > The output is already digital and they taught me something.
> > (off topic) LEDs work in reverse! (same band gaps as a solar cell, I
> should have already know this) The output led also a light sensor.  Try it
> yourself, just hook up a digital meter (ie HiZ) and shine a light at a
> random LED in your junk drawer, I got about 0.5 volt with a red LED and a
> flash light.
> >
> > Anyway, they also have a 'reader', so I'm hoping to have TWOG read the
> already digitized data (from what I'm hearing, it will need to be as farway
> from the antennas as possible!)
> >
> > Another thing I don't know is how slow the reading will be. (I'm guessing
> that after internal filtering it will be slowish, like 1 sec or worse, I've
> got my fingers crossed for 100milli-sec)
> >
> > Any idea what sampling rate I must have to be able to use it as landing
> data?  If I have a long approach (1/4 mile) does this sound plausable?  My
> issue for not using a radar/sonar range finder is very slopped terrain with
> a smallish flat spot to land on.
> >
> >
> > My worst case may be a radar/sonar first pass over flat spot to adjust the
> altemeter and a second pass to actually land.
> >
> >
> > -----Original Message-----
> > From: gisela.noci [mailto:address@hidden]
> > Sent: Tuesday, November 10, 2009 12:27 AM
> > To: address@hidden
> > Subject: RE: [Paparazzi-devel] distance measurement for landing.
> >
> > What 'Baro' sensor?
> >
> > The maths does not lie.
> >
> > We cannot generalize when categorizing sensors in this class of
> performance.
> >
> >  The picture I painted below is typical of a fairly good quality MEMS type
> airdata sensor. For example, the Honeywell ASDX015 is a 0 to 103.4kPa
> sensor, fully temp compensated from 0 to 85degC with 4V output span, and
> 2%/V accuracy. This means that 4V output covers the full 10,000m altitude
> pressure range range and that means 4/10000 = 400uV per meter. To achieve
> 14cm, you need to be able to sample around 50uV. If you have ever tried this
> it is not easy, esp in the noisy environment of RF modems and video
> transmitters, etc
> >
> > In addition, the inherent sensor accuracy is typically in the order of 1
> to 2% per volt, ie, at sea level, lets say at a sensor o/p of 1volt  the
> error
> > could be 10mv, or 25 meters!!
> >
> > One can amplify the sensors output, and offset the output, to generate a
> large sensor voltage change for a smaller altitude change, thereby reducing
> the number of sampling bits of the A/D, althought then the sensor range is
> limited. This is often done, since most of our applications do not go to
> 10,000meters,(I suspect this is what they have done in the sensor in the
> > quadrotor) .....But, there is no free lunch here - you have to use
> amplifiers that have very low input and output noise, very low DC offset,
> and good temp stability. If you have access to a high accuracy pressure
> calibration system, you could put the sensor pack in the environmental
> chamber, and cycle the temp from 0 to whatever, and vary the pressure to the
> sensor at each temp, and generate a calibration curve to eliminate all the
> errors,ie, temp, offset, drift, etc but the cost.....(takes many hours to
>
> > do)
> >
> > In essence, obtaining repeatability, high accuracy, and temp stability, to
> achieve centimetric pressure alt accuracy, is what $10k sensors are all
> about.
> >
> > We do this type of work at our company ( we are a UAV manufacturer) and
> all this come from painfull experience..
> >
> > Have fun all
> > Joe
> > PS - I could not access the sensor referenced by Eldon - seems the only
> one I find is a 1meter 'accuracy', and 400m or so range??
> >
> >
> > --Original Message-----
> > From: paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> >
> [mailto:paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> > ] On Behalf Of antoine drouin
> > Sent: Monday, November 09, 2009 10:48 PM
> > To: address@hidden
> > Subject: Re: [Paparazzi-devel] distance measurement for landing.
> >
> > this is done with a baro and a 10 bits ADC
> >
> > http://poinix.org/video/booz2_vert_ctl_v0.2_perturb.mp4
> >
> > On Mon, Nov 9, 2009 at 9:31 PM, gisela.noci
> <address@hidden> wrote:
> >> Baro_altitude sensors can certainly measure to that kind of accuracy.
> >> The difficulty arises in digitizing the output ( via an A/D convertor)
> >> and ensuring that the noise on the signal is very low. Some typical
> >> sensors
> > give
> >> a 0-4v output for a pressure variation from 1013millibar (close to sea
> >> level) to around 200millibar ( a difference of 813mb, 81kPa, about
> >> 8000meters alt change). This means to achieve 14cm resolution, you
> >> need 8000m/0.014m = 57000 bits, or increments. This implies at least a
> >> 16bit
> > A/D,
> >> and then remember that each bit is 4v/57000 = 70uV !!!   Couple this
> >> with the sensor output variation over temp ( not insignificant!) and
> >> the noise
> > on
> >> the signal, you will be VERY lucky to get 1meter accuracy, and that at
> >> a specific temp.
> >>
> >> Ask any real pilot - land with your eyes, not your altimeter!!!
> >>
> >> Ground ranging (Radio-Altimeter, radar, ultrasound, etc is the only
> >> way to go.
> >>
> >> Joe
> >>
> >> -----Original Message-----
> >> From:
> >> paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> >>
> >
> [mailto:paparazzi-devel-bounces+gisela.noci=ate-international.com@nongnu.org
> >> ] On Behalf Of Elden Crom
> >> Sent: Monday, November 09, 2009 7:57 PM
> >> To: address@hidden
> >> Subject: RE: [Paparazzi-devel] distance measurement for landing.
> >>
> >>
> >> I was considering the pressure/altitude route
> >>
> >>
> >
> http://www.hobbycity.com/hobbycity/store/uh_viewItem.asp?idProduct=6776&Prod
> >> uct_Name=WS_HowHigh_Altimeter_Feet_&_Meter_(New_Version)
> >>
> >> But I think this may require one on the ground station as well to deal
> > with
> >> a cold front coming in and changing the barometric pressure.
> >>
> >> Supposedly, this thing can measure altitude accurately to 14cm.
> >> Does anybody believe that? (I have on one order to see if it's
> >> actually
> > that
> >> accurate)
> >> I wonder how to mount it such that it is not affected by wind speed.
> >>
> >>
> >>
> >>
> >> -----Original Message-----
> >> From: Dirk-Willem van Gulik [mailto:address@hidden]
> >> Sent: Fri 11/6/2009 1:04 PM
> >> To: address@hidden
> >> Subject: Re: [Paparazzi-devel] distance measurement for landing.
> >>
> >> Roman Krashanitsa wrote:
> >>
> >>> The http://www.maxbotix.com/ founder did some comparison for
> >>> performance of his sensors with Sharp sensors you are linking to. As
> >>> far as I remember, there are some plots in the FAQ section and in
> >>> "Preformance Data" section that might be useful for you.
> >>
> >> I found that even very simple DIY shop ultrasonic meters; like below:
> >>
> >>
> >
> http://www.sella.co.nz/general/building-renovation/tools/other-tools/503tx7/
> >>
> >> which can be had for 5 to 10 euro's at the local DIY market - are
> >> easily hacked and not that unreliable.
> >>
> >>
> >> Thanks,
> >>
> >> Dw
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> _______________________________________________
> >> Paparazzi-devel mailing list
> >> address@hidden
> >> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
> >>
> >
> >
> > _______________________________________________
> > Paparazzi-devel mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > Paparazzi-devel mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > Paparazzi-devel mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
> >
>
>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>
>
>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel






------------------------------

_______________________________________________
Paparazzi-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/paparazzi-devel


End of Paparazzi-devel Digest, Vol 68, Issue 30
***********************************************


reply via email to

[Prev in Thread] Current Thread [Next in Thread]