IR follower

[james1]

I was thinking of making a picaxe car that would move by itself and avoid obstacles using IR sensors. I also wanted it to follow a transmitter of some sort. I thought using a IR led for a transmitter (TV remote) and four IR receivers on the car (front, back, left, right). Depending on which receiver has the strongest signal, the car will go in that direction.

For the IR sensors I would use one led to shine and another to pick up the light reflected off a object. When the light reaches a certain level the car will reverse and turn then carry on.

Has anyone tried this or anything similar?

(I remember someone said they made one that followed light)

Edited by - darthtader on 18/03/2007 07:25:16

 

[jplinteau]

You may be interested by this
<A href='http://www.hvwtech.com/products_view.asp?CatID=114&amp;SubCatID=148&amp;SubSubCatID=0&amp;ProductID=88' Target=_Blank>External Web Link</a>

J-P

 

[BeanieBots]

The Sharp GP2D12 range finder described in the reply above is an ideal choice for object avoidance but can't be used for IR detection.
If this is a first project, I would aim a little lower and go for something easier such as the light follower you described.
Start off by making a simple robotic platform that can get the PICAXE to move in a straight line, left and right. Maybe just a set timed sequence to start with. Until you have that perfected, there is no point trying to introduce any sensors and/or intelligence.
Then add object avoidance.
This could be as simple as a few microswitches, the sensors described above or maybe even ultrasonic. (start simple).
Finally, think about adding sensors to detect what you want to follow.
Initially, I would suggest just using two sensors and I would also suggest using visible light to start with. It is possible to &quot;see&quot; IR by using a CMOS CCD camera but it is far easier to test when you can see directly yourself what your robot is 'looking' for.
Two LDRs mounted at the front of the robot facing out at about 45 degrees apart is a good starting point.
Then it is a simple question of turn towrds the sensor that gives maximum signal.
This principle can be expanded on from a simple &quot;if left sensor greater than right then turn left&quot; up to a more complex but better and smoother, &quot;if left sensor is X% greater than right then turn a radius of Y&quot;.

Above all, get a good stable platform working first. Then worry about sensors. The worlds best sensors read by the worlds most powerfull computer with the best software won't be able to control a robot that can't be controlled.

 

[Dippy]

Darthtater, you want two things here as I gather.

1. A sensor 'array' which can detect the strongest signal from a handheld thingy and follow it (or turn towards it).
2. Range finding for obstacle avoidance.

For the first bit, I tend to agree with BB about trying light sensing to experiment with. Why? I assume that as you plan to use a TV R/C you will be using one those IR sensing modules on the robot? Well the trouble is that they have a huge dynamic range and an on/off logic output. So you could easily get 2 or more sensors detecting your TV R/C simultaneously. (You know how you can wave your R/C round and the TV still changes channel - that's huge dynamic range.)
You would need some clever electronics (prob using op/amps to detect 'level' and unless you have experience/knowledge your amplification &amp; filtering would be very tricky).
A modulated I/R would be far better but blooming tricky to get right - espcially if you haven't experience of op-amps.

On the second point, I would first point out a basic error in JPL's linked website description. It implies the Sharp sensor uses light time-of-flight to calculate distance. That is WRONG. It uses a triangulation method with beam reflected back to a p/d linear array type thing. And laser rangefinders will cost hundreds.
Having said all that it is a popular device.

PS. A question: If you use IR for your direction control, could that interefere with the Sharp rangefinder? I don't know, but its something to think about.

Edited by - Dippy on 18/03/2007 12:40:09

 

[james1]

I think it would be cheaper to make my own range finder than buy one.
The direction control shouldn't interfere with the range finder because the range finder needs a higher level of IR light. It would not be able to pick the light from the controller at a diatance.

I was going to use a RC car body. Then I could put one IR reciever on each side of the car. That way only one should have the strongeat signal.

I might try using LRDs first and see how that works.

 

[slurp]

are you thinking of a straight un-modulated signal?

This is usually ok for line sensors mounted close to the surface but if you're looking at object detection then the IR signal will be swamped in daylight and with some artifical light sources.

The modulation allow better discrimination between background levels and the signal (to a limit!)

There are a variety of sensors &quot;off the shelf&quot; which range some simple &quot;bump&quot; detection to ranging sensors that give a signal proportional to distance. The former are easier to build than the latter but not always cheaper. Sharps range sensors are relatively consistent regardless of the reflecting surface (again to a limit).

TV remote components can be useful but as they designed for a different purpose it can be interesting getting the signal right and understanding just what's going on.

regards,
colin

--
www.minisumo.org.uk

 

[Dippy]

Darthtader, how do you plan to make your own rangefinder?
An interesting project, let us know how you get on.

 

[BeanieBots]

My thoughts too. There's a lot more to it than just measuring the strength of a reflected signal. I'd be impressed with a design that was cheaper than the GP2D12.

 

[Rickharris]

<A href='http://www.societyofrobots.com/sensors_sharpirrange.shtml' Target=_Blank>External Web Link</a> A good description of how the sharps IR sensor works - Without the ability to mould the lenses I think you may as well cough up the cash. the system seems very sophisticated.

A simple reflective IR system is quite short range in my experience - I tend to use ultra sonic. Touch works well on the bench top.

 

[rolsen]

Does anyone know who would stock the GP2D12 Sharp rangefinder in Australia.

 

[inglewoodpete]

Heaps of places in Oz, including RS-Components on line. About A$25 to A$35. (Cover your eyes Dippy): Google it.

 

[moxhamj]

Those links are very good, and the prices mean it isn't worth reinventing the wheel. I designed a reflective IR system a few years back that took a sample of the light level just before sending out a pulse, then sent out a pulse and measured the difference. It certainly extended the range and made the system much more immune to interference from flouro tubes. All things being equal, I'd buy one &quot;off the shelf&quot;.

 

[Lliam20789]

Hay everyone,

I'm still working on my mapping robot.

Just wondering if anyone knows of a simple circuit I could make with basic parts and IR LED's and revivers that would provide about 15-cm of range.

The plan was to use it as a backup to the sonar rang finder, SRF05, which will be mounted on a sweeping servo. The IR would simply tell the servo to &quot;look&quot; in the direction of an obstacle if it wasn't already...

I do not have a very big budget so if anyone has a simple circuit than that would be great!

Thanks,
Lliam.

 

 

[james1]

Here is a robot that uses IR sensors
<A href='http://www.iirobotics.com/catalog/product_info.php?products_id=312' Target=_Blank>External Web Link</a>

 

[Dippy]

Is that the basis of your design? Or was it a couple of hours Googling? Anyway regardless, it looks very nice.

I note: &quot;2 direction-discriminating light sensors.&quot; Light not IR.
I thought you wanted IR 'following' sensors? I suppose you could double/up and use them to receive IR commands too. Though with standard IR receiver (demodulator) devices I still think your device will pick up from more than one direction, unless you use it outside.
I be interested to see how those IR rangers (eg from Sharp) behave outdoors. Home-made IR sensors that work outdoors and indoors I would have thought would be pretty tricky.

Looks a very interesting project, go for it.

 

[jodicalhon]

As Dippy mentioned earlier, the IR receiver modules have a large dynamic range and cannot discriminate between an emitter that is 1 metre away and one that is two metres away. Or five metres away.

&quot;The Robot Builder's Bonanza&quot;, by Gordon McComb, talks about making a (not sure how accurate) range sensor by varying the emitters modulation frequency about the centre frequency. Like 38kHz plus or minus a few kHz. I can't remember exactly. I don't have the book so I can't check - the library wanted it back.

Apparently, the receiver's sensitivity depends on the emitter's frequency and its distance from the receiver. Meaning that, even if the emitter frequency is off the centre frequency, the receiver will still trigger if the emitter is close enough.

So, a receiver that is closer will trigger sooner and reset later as the emitter frequency swings about the centre frequency.

(This is all from memory, so sorry if it's a bit skewiff.)

I don't think this is accurate enough for a good range sensor, but could maybe be used for differential sensing with two sensors.

wilf_nv has also posted, elsewhere, an amplitude modulating scheme that uses a 100Hz sawtooth to modulate the 38kHz carrier. The receiver that is closest triggers earlier.

Just some thoughts.

 

[Dippy]

I tried a method which modulated O/P power some time ago for people-detecting and it worked to a reasonable degree. I ramped up the LED-voltage using pwm-rc and had a low sided FET to modulate.
(Actually it was a laser diode, so for the required oomph for an LED you will require buffering).

Modulating the CF may work too I hadn't thought of that. Its worth someone playing with that.

I just guess it depends on the reliabilty you require in different environments. To get it good you will need to put in some work.

Edited by - Dippy on 23/03/2007 10:24:46

 

[BeanieBots]

To &quot;get it good&quot; will require unimaginable hard work. The &quot;big boys&quot; have already invested hundreds of man years and so far the best available on hobby budget is the likes of the GP2D12.
If absolute distance is not required, then the method described by Jo C is very effective. Namley, detecting if the left or right side of the robot is closest to the object.

The question about the GP2D12 working outside.
Direct sunlight blinds it completely.
This can be solved with a simple visor.
In a very bright (sun lit) room with white walls, it perceives objects to be up to 30% closer than they really are.
This reduces to about 5% when the wall is not in direct sunlight but is still very bright.

Another interesting effect is that if the PICAXE is run from batteries, as they run down, the robot sees objects ealier. This is because the PICAXE ADC becomes more sensitive at reduced voltage but the GP2D12 gives a reasonably constant voltage for the same distance as its power rail drops.

Another advantage of the GP2D12 is its non-linear response. OK, the reverse characteristic at very close range is an absolute pain and makes a close robot steer hard into the object, but the response in the 'normal' phase is just what's need. Increased sensitivity as you get closer. This saves having to calculate exponents and/or squares within the PICAXE if you want to do proportional object avoidance rather than just making sudden 90 degree turns when your predefined limit is exceeded.

No, I don't have shares in Sharp. I just think it is ideally suited to PICAXE robots.