Let's Make Robots!

Make your own IR obstacle detection sensor

Detects objects at close range. Can be used for object tracking.
AttachmentSize
OBJECT_TRACKER_.BAS1.24 KB
OBJECT_TRACKER_V1.7.BAS1.89 KB
MPSA13.pdf154.35 KB
MPSA14.pdf43.7 KB

Now with video of the sensor being used as a Mintvelt inspired object tracker! This sensor is a short range obstacle detector with no dead zone. It has a reasonably narrow detection area which can be increased using the dual version. Range can also be increased by increasing the power to the IR LEDs or adding more IR LEDs

The photo below shows my test setup with some IR LED's (dark blue) as a light source and two phototransistors in parallel for the reciever. You could use one of each but I wanted to spread them out to cover a wider area. This setup works like a FritsLDR but with IR. It has a range of about 10-15cm (4-6 inches) with my hand as the object being detected.

I'm only running my LEDs about 20mA. My LEDs are capable of 50mA continuous and some LEDs are capable of 100mA (see "Getting the most from a LED").

I'm using this setup on Junior as a general purpose object advoidance sensor to prevent him backing into anything. I'm getting a good response with less than a volt when my hand is up close and reflecting the IR and over 4.5V with no IR.

IR_amp_closeup.jpg

To get this to work well with an A/D input it needs to have a much lower impedance (needs to let more current through). You can do this with an op-amp but most op-amps like more than 5V and are usually more expensive than my one transistor and three resistors. This is a simple one transistor amplifier that gives my ADC good resolution. Click on the schematic for a larger picture.

IR_sensor_Schematic__small_.jpg

Starting from the left you can see my two IR LEDs with a resistor and transistor in series. The transistor allows the processor to turn the LEDs on or off. This is necessary to tell the difference between the ambiant IR from daylight and indor lighting and the reflected light from the LEDs that indicates the presence of an object.

Next are my two phototransistors in parallel with a 1M resistor in series. You could use only one but I wanted to cover a wider area so my transistors will point in slightly different directions. If either one detects IR it will allow more current to flow. Since volts=current x resistance, even a small increase in current will create a reasonable increase in voltage across the 1M resistor. Unfortunately the low input impedance of many AD converters will act like a small resistor in parallel with the 1M resistor and dramatically reduce the output to the processor. This is where our BC549 transistor comes in to save the day. In conjunction with the 1K and 10K resistors it amplifies the signal so that the analog input on your processor gets a nice strong signal.  The BC549 is not too critical, just about any general purpose signal transistor should do. My transistor had a hfe of 490 when measured with a multimeter. You should probably have a hfe of at least 200-300.

IR_sensor__small_.jpg

As you can see my sensor is made from liberal amounts of hotglue. Click image for a bigger picture. This has the advantage that you can flex the leds and transistors outward to cover a larger area. This is Juniors reversing sensor to prevent him reversing into anything and as such will cover a wide area. I will make single Led/Phototransistor sensors for front left and front right. This will allow him to avoid crashing into obstacles when his rangefinder/object tracker is looking elsewhere.

Note that the phototransistors are slightly forward of the blue LEDs. This helps stop stray light from the LEDs being detected.

 

 

 

 

 

 

Below is the sensor hooked up to Juniors mainboard which has three of my amplifiers built in.

IR_Sensor___Mainboard.jpg

 Using a simple test program that turns on the IR LEDs, stores the value of the ADC, turns off the LEDs, reads the ADC again and then subtracts the stored value from the recent value I was getting readings from 6 to 940. This was with the curtains closed and the lights off. When the reading was 6, my hand was about 300mm (1ft) away. With the lights on the values ranged from about 60 to 940 with a value of 60 being with my hand only about 150mm (6inches) away. Considering the max possible resolution with a 10bit ADC is 0 to 1023, I thought 60-960 with the lights on was a very good result.

After a comment about using sleeves I repeated these test with heatshrink sleeves on the LEDs and phototransistors. The sleeves actually had a negative effect and reduced the range. After I removed the sleeves I did not get the same reduction in range with the lights on. I don't know if it is because during the first test it was daylight outside and the curtains didn't block it all or if it was the way I held the sensor but the second set of test gave an almost identical range of approximately 300mm (12 inches) reguardless of the lights being on or off. I'll have to try again tomorrow when it is daylight again. It seems my initial test was at fault, maybe the way I held the sensor?

IR_Sensor_Single__small_.jpgThis is the single version of the sensor and will cost about half. In the photo you can see the current limiting resistor for the LED. Ignore the value as I had different requirements for Junior. Use the values shown in the schematic.

I've joined the positives together so there is only three wires going back to the mainboard.

Note that the phototransistor is slightly in front of the LED to prevent stray light from the LED being detected.

 

IR_Sensor_Single_2__small_.jpgOnce again I've used hotglue and heatshrink to make it solid and well insulated.

 

 

 

 

 

 

 

 

 

 

 

This is the schematic for the single version. Click on it and the photos for larger images.

IR_single_sensor_Schematic__small_.jpg

Because this sensor only has a single phototransistor it isn't quite as sensitive. To compensate I've increased the current to the LED to almost 50mA which is the maximum continuous current allowed. Because the LED is pulsed on and off this is quite safe and could have been increased to 100mA. The problem with pushing a LED to its limits when controlled by a proccesor is that if a fault occurs in the software then the LED could be destroyed.

When tested, The readings from the ADC of the picaxe ranged from about 100 - 910 reguardless of background lighting. Despite the slightly reduced resolution due to a single phototransistor the range was about 400mm (16inches). This increased range was due to the increased power to the LED.

Make certain your LED and phototransister are parallel to each other for good range.

It was asked how wide is the detection area. Using my hand as the object at a distance of aproximately 300mm (12 inches) from the single sensor the detection area was about 150mm (6 inches) wide. The double sensor can detect a wider area if the phototransistors are spread out at different angles.

Using my hand sideon to the single sensor the detection area was only about 60-70mm (2-3 inches). This is reasonably narrow due to the lenses in the LEDs and the phototransistors.

It should be noted that this is not a linear sensor because the intensity of light from the LEDs is 1 divided by distance squared. In other words, when the object is twice the distance away, the IR from the LEDs is 1/4. As a result, the closer the object, the better the resolution.

This would be a useful sensor to fill in for the dead zone of other IR sensors such as the SHARP GP2D12. To prevent interferance, one should be disabled when using the other.

 


 

As mentioned at the start, I've also experimented with using two of these sensors for a simple object tracker inspired by Mintvelt's "four eyes". This version can't tell the size or distance of an object but can track an object well enough for a robot to recognise a moving object and give chase. Wish I still had a cat, imagine a robot with a waterpistol chasing a cat around the house :

I've attached the code used in the video as well as an improved version (V1.7) that eliminated the servo jitter.

 Object_tracker2__small_.jpg

 

 Good luck and enjoy :)

 


Sunday 4-1-2009

 

This is the latest version of my object tracker as used in SplatBot. I've used 20 IR leds to increase the range. They are limited to 50mA at the moment so that they can't be damaged by faulty code. If I was to push them to their limit then the range could be increased further but they could then be damaged by something like an interupt routine occuring when the LEDs are on.

IR_sensor_array__small_.jpg

This is the schematic.

IR_sensor_array_Schematic__small_.jpg

Click on it for a larger picture. I found with all The LEDs on that the sensors were swamped by reflected IR from my hand even at a distance of about 400mm. The circuit works fine and I definitely get a lot more range but I'm going to have to remove the sensors from the board and mount them seperately so that I can adjust their distance relative to each other to optimise tracking and so I can better shield them from ambiant IR.

This is a work in progress.

 


Updated: 19-1-2009

I've experimented with improving and simplifying the detection circuit. This will give you better range.

IR_obstacle_detector_3_Schematic.jpg

The MPSA13 is a high gain darlington transistor with a hfe of over 5000. If you get the MPSA14 it has about twice the gain. By adjusting the 500 ohm trimpot you should get much better range than the old circuit.

 

 

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

The diagram is exagerated, the angle is narrow because the sensors have lenses moulded in. The downside is that the LED and the phototransistor need to be aligned optically. In my code for the IR object tracking I had to calibrate the sensors. I found at least part of the reason for this was that with one of the sensors, the hot glue set with the LED/phototransistor noticably out of alignment. Correcting this drastically reduced the problem of getting both sensors to respond equally. Other factors such as different amounts of transistor gain mean that some callibration in the software will always be nescessary.

The real question is what do you want your robot to do with this ability? Is long range tracking usefull? The greater the range, the more objects that will be detected and confuse the readings. You need to either do high resolution scans (which require more proccesing power than a picaxe) with a precision scanner (which is what I was trying to do with the laser range finder) and determine whats moving and what isn't or stick to short range and simple sensors that detect fewer objects and simplify programming.

So far from what i've seen of the more complex robots that use cameras and high end processors, they don't do any more than our very simple object tracking systems except determine the colour of an object. Three LDRs with different coloured cellophane filters (Red, Green, Blue) and a white LED would work well enough to tell if an object is orange or pink or purple etc.

Oh crap! I'm rambling again. Sorry :|

True it isn't the same as your tracker but I felt you deserved credit for the inspiration. It can't determine the size or distance of an object but set it up with a waterpistol and the cat's in for a nasty suprise :D

I've edited the intoduction text accordingly.

This kind of writing is what we need  on LMR! Forget about that pulsy thingy. We need high quality walkthroughs!

This would be one of them. Thanks Oddbot. Your schematics are starting to become your trademark. In a good sense.

One request though: could you name your sensor after its goal, rather than after its means? Sure it uses IR, but in the end we all want to build a distance sensor, or an obstacle detector.I guess it's all about that purpose thing again....

8ik

Wikipedia Affiliate Button

Sorry Rik, I did do the tag thingy, now I'll work on the name thingy :D

I'm about to update with the single version of the sensor so I'll do a rename while I'm at it.

now it's just plain perfect

This configuration would probably have a nice wide beam. Dit you test the width of the detection area? 

I've read that the sharp IR rangefinder uses some sort if modulation to make sure it gives good readings in differtent lighting conditions. I asume the frequency of the light is modulated, but how do you do that with an LED?

My sensor, like the FritzLDR and I presume the sharp sensor from what I've seen of its block diagram all work by pulsing the IR LED on and off, This is necessary to eliminate false readings due to background IR from indoor lighting and daylight.  It is not the frequency of the light itself that is modulated.

I've now updated the tip/walkthrough to include the width of the detection area but it is only aproximates as it could vary slightly with different components used.

You could use pwmout  to modulate the led or a 555 timer chip, but you'd then need a modulated receiver as well(though those are sold in a nice small package like a tsop4840).
If you put sleeves on the IR LED's so it won't shine directly into the recievers it may improve performance :)

I tried using small black sleeves of heatshrink that fitted snugly over then LEDs and the phototransistors. This actually reduced the response. I then tried very short sleeves on the phototransistors only, with just the body (not the rounded front) covered. This did reduce the fluctuations caused by the fluroesent lights but also caused the value to occasional go negative (jumped to 65535).

The fluctuations are because the fluro lights pulsate rapidly with the AC from the wall. In Australia it is 50Hz. Because the picaxe is sampling at a different rate the readings fluctuate.

The sleeves are not  necessary because of the lenses moulded into the LEDs and the phototransistors combined with the fact that the phototransistors are slightly forward.