Multiple Ultrasonic Transmitters + Interference

[dknguyen]

To give the heli and absolute ground reference during landing, indoors, and when it's too low that the horizon sensors don't work, I was going to have 3 or 4 (maybe 5?) ultrasonic transceivers pointing outwards and downards from the base in a cone shape to detect the ground.

I was going to have them all fire transmit simultaneously and then listen for an echo (I figure I can assume they are all at a single point and I want maximum range on this thing so I'd prefer to fire them all at once rather than alternating), but I was wondering if this would cause desctructive interference? Constructive interference is what I want, and if I just fire them using a single control pulse, shouldn't that do it?

I'm more worried about the presence of desctructive interference rather than the lack of constructive interference- except the wavelength is significantly smaller than the distance between the transducers so I suppose synched transmitters it could still cause destructive interference due to sensor positioning

 

[jpanhalt]

When I have used an ultrasonic yardstick, if it is aimed at the wall at other than a perpendicular angle, I get odd readings. Won't your side angled transmitters give the same type problem? That is, reflections won't be directly back to the receiver and the point from which the reflection comes will be measured on the slant. Moreover, you have no way to know which echo is from which transmitter. John

 

[dknguyen]

Yeah, if the beam's divergence is too low, then the reflected signal isn't wide enough to be picked up by the transmitter on a slant. The narrower the beam, the closer it has to be to perpindicular to be picked up by the transmitter. I ran into this problem with Devantech's SR235. I thought it was broken for the longest time (I had shocked it and superheated the transducer with the soldering iron by accident), but it turns out that my angle was just not perpindicular enough to give a reading.

I'm not sure if I Need to know which echo is being received by each transmitter. THey are going to be fairly close together, enough that I think I can assume they are all the same transmitter and I can probably echoes ater the first since they are probably from the other transmitters whose return times would be longer since they are pointed in the other direction. The goal is to get readings in a cone below the chopper to measure it's orientation with respect to the ground.

Ultrasonics do have a pretty wide cone...so it might be pointless to have multiple ones (or too many). But I do need more than at least 3 readings from separate points on the ground to calculate the attitude.

I might have to trial error to check how far off horizontal it can be before the ultrasonics stop receiving their echoes.

 

[kchriste]

If you reduce the number of transducers to 3, you could take your readings in sequence. Tx #1, wait for your echo or max echo return time (whichever comes first), Tx #2 wait for echo/max time, Tx #3 wait for echo/max time, repeat.
"max echo return time" = max distance you want to measure.
I don't think you'll have any luck Txing from more than 1 transducer simultaneously. What would work would be a very wide beam Tx transducer in the middle surrounded by several narrow beam Rx transducers aimed in different directions.

 

[dknguyen]

Yeah, my original plan was two either have 3 outward receivers and one centered transmitter facing down, but then I started thinking, why not make them all transmit? (then maybe I could rid of the center receiver) Coherency problems with destructive interference and stray echoes that might happen when using simultaneous transmitters might be a nasty, unforseen, and very bad thing if it happened since the chopper is probably very close to the ground when it happens.

THe big problem I just thought of with a single transmitter surrounded by receivers is like what jpanhalt's case. If the chopper was at an angle, the receivers wouldn't detect the ground at all. The heli is tilted way too much, and it'se sensor setup can't work! And since I don't have super wide transmitters or narrow receivers, I guess that setup is crossed off for good.

So yeah, I guess I'll just use the 3 sequencing transmit/receives then. Maybe I should still have one pointing straight down? Although a center transceiver would has more potential to interfere with the first 3 "coned" transceivers and slows down my sampling rate. Maybe just 3 coned ones then (or 4), but no center.

THe problem is that at max range of 10m (20m for round travel time), my maximum sampling frequency is only 17Hz. If I start alternating transmitters, then that leaves only 5Hz per transmitter. That is definately too slow. However, it is fairly far from the ground at that point at the IMU can "fill-in-the holes". An absolute reference every 200ms is good enough to prevent the IMU drifting.

I could set it up though to transmit as soon as an echo is received so that as it gets closer to the ground the sample rate increases so about 50Hz per transceiver at 1m from the ground since that is when it's most critical.

Even if I had a dead-time between transmissions equal to the sensors to let all the stray echoes die out, that's still 25Hz per transceiver at 1m from the ground, and it's not too important at 10m since the IMU is there. PLus wind at 10m probably makes it useless anyways.

Now.. should it be two pointing fron-down-outwards and one pointing straight down-back? Or vice versa? Or just 4 one on each corner?

 

[kchriste]

Now.. should it be two pointing fron-down-outwards and one pointing straight down-back? Or vice versa? Or just 4 one on each corner?

That depends. If you are planning on flying backwards a lot, then 4 may be necessary. You might also need to add a little extra tilt to the forward transducers to scan ahead and to compensate for the heli's forward tilt when it is flying forward.

 

[dknguyen]

Well I guess bad control would make it travel backwards at least a little bit on landing. Might as well make it 4 then- easier translation of the X and Y axis orientations that way.

I'm not entirely sure about forward sensing yet- I've dropped it completely for now since I don't think sound is long enough range for the possible speeds this thing can travel at to be of much use. Maybe someday...obstacle avoidance cameras or optic flow or something. But not now.

 

[ericgibbs]

hi dkn,

A point to consider with ultrasonic sound beams, they can be 'diverted' by a side draught of air.. perhaps the downdraught from the heli blades could be a problem.

What frequency range are you looking at.?

 

[dknguyen]

hi dkn,

A point to consider with ultrasonic sound beams, they can be 'diverted' by a side draught of air.. perhaps the downdraught from the heli blades could be a problem.

What frequency range are you looking at.?

Yeah, I had thought about the blades interferring with the ultrasonics, but I have no idea if it will or not. THe frequency was 50kHz because that's what the transducers work best at (and give the most range). If I went infrared (as in the Sharp IR parallax rangefinders), it would be unaffected by the wind and rotor thrust, but it would be affected by sun, have a larger minimum range, and lower maximum range. I just have no idea right now how much of a problem it would cause.

 

[ericgibbs]

hi,
Not knowing your heli and landing gear configuration, I'll suggest the following.

What if, centrally located, an pulsed active downward pointing ultrabeam, with passive detectors located on the outer edge of the landing gear.

So in effect you have a downward sound cone, with two offset detectors.
It should be possible to derive some indication of ground distance under the heli.

 

[dknguyen]

I'm starting to think that without accurate-laser rangefinders, I might not be able to accurately sense the orientation of the heli with only rangefinders alone. Too many unknowns.

I'm unsure if 4 downard, outward facing transceivers would be better (in theory I suppose they would be) or 4 corner mounted receivers with a single center-mounted transmitter (due to the possible interference). But they are very similar as far as hardware is concerned, so I probably should just build the setup and run a bunch of tests.

The wide beamwidth might render the 4- downward vertical receivers and 1 downard vertical center transmitter ineffective since the beamwidth is so wide- but not wide enough so that I can spread out the receivers far enough so there are large altitude changes between receivers for a given angular change in orientation. That's one of the things I like about the 4 outward-down pointing transicevers, each one is it's own transmtter and has it's own dedicated scanning area indepenendent of the other transceivers so that maybe, just maybe the wide beamwidths would be able to "map" 4 separate elevation points to determine the ground plane. They could be more sensitive to the rotor downdraft than the single center mounted transicever though...

It would seem that I would need to keep the chopper high enough so that the horizon sensors are effective and can prevent IMU drift, and only go below that minimum altitude during landing and travel fast enough through that "dead altitude" where the horizon sensors are not effective quickly enough so the IMU can still be depended upon, using only the ground-rangefinder data + IMU orientation data to figure out where the ground is, rather than only relying on IMU data to figure out orientation and altitude near the ground. If that's the case only one centered mounted downard rangefinder would be needed.

I guess...it's trial and error now!

(Maybe I should just make BIG wide landing gear with a shroud underneath which the ultrasonics are mounted to help block rotor downdraft. Less manueverability...but who gives a crap about that anyways?)

 

[ericgibbs]

(Maybe I should just make BIG wide landing gear with a shroud underneath which the ultrasonics are mounted to help block rotor downdraft. Less manueverability...but who gives a crap about that anyways?)

If you block the downdraught with a heli mounted panel, won't that spoil the lift.?


You could fire the transponders sequentially, as you know velocity of sound in air about 330m/sec.
Should be enough time to scan and process 4 transponders OK.

Check that the heliblades are not pushing out any ultrasound at your frequency.

 

[dknguyen]

If you block the downdraught with a heli mounted panel, won't that spoil the lift.?


You could fire the transponders sequentially, as you know velocity of sound in air about 330m/sec.
Should be enough time to scan and process 4 transponders OK.

Check that the heliblades are not pushing out any ultrasound at your frequency.

Yeah, that had occured to me too. Only one way to find out., although the I don't think the lift at the center of the heli is very much because of the fuselage and the slower speed of the inner sections of the rotor blades.

Due to compactness, I have to deviate from my previous design (for ground vehicles) which used high voltage boost converters to using small step-up transformers. Their duty cycle only allows for 5Hz transmission rate though so the speed of sound is not the limiting factor. I do not know if it is sufficient. I suppose I could just rely on the IMU and use the ultrasound to stop drift, as I do with the horizon sensors and IMU when it is in the air.

All that aside, can anyone think of a way to detect translation movement in a helicopter (relative to the ground)? Accelerometer integration won't work obviously, and landing is seriously botched if you are level, but still moving relative to the ground. The only method I can think of is a very computationally intensive optic flow method (or other imaging methods).

 

[killivolt]

Is it possible to deploy a known device that has had complex circuitry. A simple camera with an auto focus ? Of course there would have to be room to mount it between the two rails ?

If I were to use the camera's focus ring and put a gear on the outside that could inter-phase another to operate a variable capacitor or dash pot ? It would automatically observe the positioning relevant to aperture once it is close enough to ground maybe you would want to eliminate it and deploy a more sensitive device ?

 

[dknguyen]

Is it possible to deploy a known device that has had complex circuitry. A simple camera with an auto focus ? Of course there would have to be room to mount it between the two rails ?

If I were to use the camera's focus ring and put a gear on the outside that could inter-phase another to operate a variable capacitor or dash pot ? It would automatically observe the positioning relevant to aperture once it is close enough to ground maybe you would want to eliminate it and deploy a more sensitive device ?

Huh? I don't understand what you just said. But the problem with most cameras is that htey have too many pixels and I can't afford the tools to work with processors powerful enough to brute-force analyze the image. But how does the auto-focus on a camera help you keep track of absolute position? Doesn't it just help you determine distance? (Or are you talking about the ultrasonics and not how to keep the helicopter stationary over the same patch of ground during a landing?)

 

[Krumlink]

Why not use Infrared? Wind cannot blow it off course, and it is perhaps better in this situation as a distance measurer, as ultrasonic cannot be very effect as proven in FRC, we used IR in a type of control circuitry, and we NEVER had a problem with it. Ultrasonic in other robots (team 148) have had problems. IR is better for distances, and Ultrasonic is better for object sensing.

 

[dknguyen]

Ir might be affected by sunlight, which may or may not be worse than the wind. My theory is that sunlight gives you a false reading, while ultrasonics would give you no reading (hopefully). I'm not sure what is worse:

-responding incorrectly to a false IR reading
-or not reacting at all to a noised-out ultrasonic echo because you it doesnt think the ground is there at all

My theory is that the ultrasonic echo would be able to be picked up once it got closer to the ground, when it matters most in bad conditions but still let it see the ground farther away in good conditions.

But the main reason was because of the lower minimum and higher maximum range.

BUt I don't know how each would fare right now. I admit that I am also more biased to ultrasonics as I trust TOF more than parallax measurements and like bats. (THey seem to use ultrasonics for UAV automated landings. I'm fairly certain it's because you have a range envelope for a single sensor.)

I do have some Sharp IR sensors lying around though, and it's far easier to imeplement them than ultrasonics because everything is already done for you.

 

[Krumlink]

why not test each method?

Test in:
Sunlight
Wind
Noise
Other things?

 

[dknguyen]

Your insistence on IR has given me the idea to use the largest IR rangefinders I can find:
https://www.acroname.com/robotics/parts/R302-GP2Y0A700K0F.html

To use them as side-sensors rather than ground sensors. THe ground sensors need a low minimum range so it doesn't lose sight of the ground just as it's about to land, and long range so it has a good lead-time to know that the ground is coming.

But for side sensors, so it doesn't side drift and smack into me or anything else while it's taking off or landing, the minimum range of 1m of the IR sensors is okay (since the rotor blades are 1m in diameter anyways, and if anything gets withing 0.5m of the rotating blades...it's all over anyways. I think outward facing sensors would be more affected by the rotor downwash than downard facing sensors, so IR is probably better than ultrasonic for that case.

THey need to be calibrated though which is a bit of pain. Another reason I like TOF measurements.

Still not sure how to actually make sure it's stationary over the same patch of ground as it's landing though. That side thrust from the tail rotor is a pain.

 

[killivolt]

My thing of imagery was not to take a picture. It simply has to focus on an object. If you had a distance at 20' to an object as you approach it the focus ring turns. If you had a 0-20' measurement you could attach a gear on the outer edge and then add output gear attached to a shaft of electrical adjust resistance or something measure it giving it a distance ? or magnetic showing how many clicks equal to X distance ?

Orientation I'm not sure. Tail yaw I think of Magnetic north as 0 and 360 degree orientation on the outer edge.