360° Object Detection with Diodes

[DigiTan]

I have a few questions about a method of object detection I want to try, but can't find much documentation on. Basically, I want to design a robot sensor that can detect objects across a 360° range using an LED emiter and receiver pair.

For the initial design, I want to try a basic IR proximity detector --much like **broken link removed**--and have it rotate in a 360° sweep using a stepper motor.

For the later design, I want to replace the IR diode with an IR laser to increase my resolution.

My first question is, where is the best place to get a very small IR laser, and generally, what is their price in U.S.C.? If all IR lasers are expensive, is there a practical way to narrow the beam of a diode and still keep the brightness. (I'm guessing a long tube with a reflective lining would work?)

Second, I was curious about rangefinding. Since light travels 1 meter in 3.33ns, I was hoping to either use a fast μprocessor to measure the delay time, or simply purchase a small pre-built rangefinder. If I decided to purchase one, where is the best place to look, and how expensive are they? This is a small robot that would weigh 10-15 pounds, so anything small in size would be preferred.

 

[Roboticinfo]

reply

Why don't you go with ultasonic instead, all you need is 4 to make 360.

 

[MrMikey83]

Well, a simple search on Digikey.com tells me that they have a 5.6mm IR laser for as low as $28. An IR sensor is less than $10. Your biggest problem is getting power to and reading the signals from the assembly while its spinning on the servo.
Good luck man!
~Mike

 

[Oznog]

Second, I was curious about rangefinding. Since light travels 1 meter in 3.33ns, I was hoping to either use a fast μprocessor to measure the delay time, or simply purchase a small pre-built rangefinder. If I decided to purchase one, where is the best place to look, and how expensive are they? This is a small robot that would weigh 10-15 pounds, so anything small in size would be preferred.

What kind of resolution can you live with? A 40MHz PIC could in theory read it with 100nS resolution, giving you a resolution of 30 ft, since that measures light travelling there-and-back that would mean a distance of 15 ft, but I'd be surprised if this actually worked as such. The sensor would have to be really, really fast.

 

[Nigel Goodwin]

Second, I was curious about rangefinding. Since light travels 1 meter in 3.33ns, I was hoping to either use a fast μprocessor to measure the delay time, or simply purchase a small pre-built rangefinder. If I decided to purchase one, where is the best place to look, and how expensive are they? This is a small robot that would weigh 10-15 pounds, so anything small in size would be preferred.

Generally you don't measure the time for the beam to be reflected, the speed of light is too fast - the short range Sharp IR sensors measure the angle of the reflected beam in order to measure the distance.

This is why ultrasonics are usually used for distance measurements.

You may be interested in the current issue of EPE, it has an article for a 360 degree distance sensor using ultrasonics and a stepper motor.

 

[DigiTan]

That's a good point. I did some thinking about it and I may avoid the rangefinding portion until the other issues are settled. Most likely, if I try the timing method I might use some kind of PLD as a "coprocessor" since I don't have any high-end processors lying around.

This project is just in the brainstorming stage, so I'm not certain about resolution. Most likely, available parts will let me work with step angles as low 3.6°.

MrMikey83 also brought up another question that I forgot to ask. Generally how do you power something that's constantly spinning? I have a vague idea of what I need to do, but I was hoping someone might a link or a diagram for these gizmos. If it complicates the project too much, I might just use a left-right sweep instead of this 360° search pattern.

 

[Nigel Goodwin]

MrMikey83 also brought up another question that I forgot to ask. Generally how do you power something that's constantly spinning? I have a vague idea of what I need to do, but I was hoping someone might a link or a diagram for these gizmos. If it complicates the project too much, I might just use a left-right sweep instead of this 360° search pattern.

You usually use slip rings, or a rotary transformer, or probably a mixture of both?.

It's a LOT easier to use a left/right sweep instead!, and actually looks better - more menacing!.

 

[MrMikey83]

Rotary transformers seem to be easy enough to construct. You just need to figure out how to mount one side close enough to the other side to really transfer the power you'll need.
Then, once you have power, you'll need to transfer the info from the range sensor circuit. You would have to either use IR led's to transmit data to a receiver but this would have to be in the very center. You could also use an RF signal to transmit a short distance to the main board.

Complicated enough yet?
I'd say you best bet would be to use a sweep. You could do a 360 degree sweep anyway, just start somewhere, rotate 360, reverse polarity and go back the way you came from. Just make sure you have a flexible enough wire harness for your power and signal cables.
~Mike

 

[pittuck]

perhaps you could try a laser + simple CCD???

I know that some old digicams need external IR filters, removing the filter + lens u should / might be able to get a point of reflection, with some trigonometry you can then work out the distance...

 

[DigiTan]

Is this rotatary transformer also called a resolver in some situations? I did a little googling on them and it seems like a good possibility. Most likely I would use an enclosed IR channel to transfer data from my lidar or sonar. Otherwise, I would use a traditional servo to create a left-right sweep pattern.

Also, I was curious about CCD and CMOS cameras, and recalled a past robot I has seen that used this for precision object detection. My main problem so far is that I'm unfamiliar with video standards. Does anyone have any tutorials to recommend for getting started with this kind of equipment?

 

[MrMikey83]

I'm not sure how the video would work to find distance unless there were two cameras.
BUT, you you can do this with one camera, you wouldn't even need to spin it at all. Get a cone shaped mirror and center the point of the cone dirrectly above the camera lense pointed straight up. This way, the camera can see in all dirrections at the same time. You would just need to program the controller to read the data correctly.

Some security cameras currently use this mirror cone to view 360 degrees without having to motorize the camera, so the software does exhist. The advantage other than not needing a motor is that you are able to 'move' the camera view even if you are watching a recording.

If you could get your software to view all 360 degrees at the same time, it would be the same as using the rotating laser rangefinder and it will be constant resolution and not limited to the speed of the motor.

http://www.versacorp.com/vlink/product/axialref.htm

~Mike

 

[pittuck]

ok lets start off with a laser (with a truly fine beam ) and its on a flat plane with the CCD.

Now lets call the distance between the laser spot on the CCD and laser diode centre x meters and the angle of the laser towards the CCD Y degrees and the distance to the object z meters. Also the CCD is a VGA 640x480 pixel one and arranged so that the 640 side is parralel to the plane. With most camera's u can select the region to view, so u can choose 640x10 pixels to cut down processing.

Anyhow, some simple trigonometry gives: tanY = z / 0.5x => 0.5x.tanY = z

And volla, only difficult things are getting the distance to CCD, but it would be much easier to just use some experimental data to create a lookup table for a desired value of Y.

 

[DigiTan]

Those are really good strategies. 8) I was thinking of maybe using a flat mirror if a cone is too hard to locate. Plus, I'd imagine I'd save a ton of energy rotating that tiny mirror around instead of the whole sensor assembly. I've always been interested in stereo vision. For now I want to restrict the project to methods I already understand so I can keep the costs down. Anyway, I'm looking into all possibilities.

So if I got my hands on a CCD or CMOS camera, what type of output format should I expect from it? I was hoping to get my hands a cheap Gameboy Camara, and use its memory-mapped output to track the beam. Do most other digital camara's output this format, or should I expect something a bit more like a television format?

 

[panic mode]

there could be some problems...

 

[pittuck]

yup, but the same will happen with ultra sonnics unless u use a very wide 'beam' which would lose much of the resultion needed to create a sensor.

 

[bloody-orc]

well if he beam should "fly away" like that then why doesen't the beab fly away in sharp Ir sensors? they get accurate readings even if the object is angeled 100degrees and tilted upwards... it uses IR beams also, doesen't it?

 

[panic mode]

well, i wouldn't call it accurate. in fact it's accuracy (Sharp sensor) is quite poor (i do have few of them).
you just have to evaluate from case to case if the sensor is satisfactory for the application. decent sensors are often very expencive but the output is not affected by temeprature, color of the target etc and they will have linear output and visible red beam for easy alignment. however the price tag of $1000-1300 per sensor will make you think twice before placing purchase order