IR & Sonar sensors precision

You can homebrew your own SONAR ( Polaroid ) using parts from an old Polaroid, I got one for $15 on ebay. Here is a really cool link with application code to OOPic:
have fun...

Basically the more you spend the better it gets.
Sonar generally has better range over IR, but sonar is sensitive to air movement (such as wind). Currently the small Devantech sonar range finders work out to about 6 feet or 12 feet depending on the model. But the sonar signal is cone shaped, so it gets larger the farther away it gets, thus more than one object would reflect the sound wave back to the receiver. Usually you cannot get under 1 foot or so with sonar as with a single transducer (acts as both speaker and microphone), you have pulse the output sound wave out, wait a bit for the transducer to settle down, and then listen for the return pulse. If you use two transducers, one to send one to receive, you can get a little closer down to maybe 4 inches or so, but you still have to blank the receiver to avoid noise from the transmitter.
IR on the other hand can work down to less than a inch away from something. IR can usually get out to around 72 inches max range. But this depends on the environment. IR interference from florescent lamps, sunlight, etc can reduce or negate or jam out the IR sensors completely. Other IR sensors can also interfere with one another was well. Now some guys have used IR sensitive photdiodes with amplifiers and such and can get the range out to 50 feet or so, but this gets pretty advanced. IR sensors also have the same problem with a IR signa; going out as a cone shape signal, thus more than one object can be detected at longer ranges.
Both sonar and IR can have problems with some things like curtains or drapes dampening the signal and not reflecting it back. Some objects like chair or table legs can reflect little or no waves back causing the object to not be detected until it's too late in some cases.
IR range sensors are more inexpensive as compared to ultrasound range sensors, which are a lot cheaper versus laser rangefinders. Another interesting sensor is the object imaging system, which does work pretty neat too, very sensitive to odd things like chair and table legs and stuff.. Freescale semiconductor also has a MC33794 E-field imaging device, but it is much less sensitive as it is intended for car seat passenger detection.

Thanks, Dale. Interesting how many sonars usually needed to cover 360 degree view.

Hi
I don't know, as I am a newbie to it all...but if you look for the Polaroid SONAR datasheet it will give the scan range ( Polaroid used to make just the sensor ) that should help. Maybe you could mount it on a servo and oscillate it?
Cheers
Dale

You best place I've found for old Polaroid cameras is in thrift stores. The prices are typically in the $5-10 range (don't pay more than $10 - they're not worth it) and you don't have to pay shipping charges.
~WEC

I should note that with the Devantech units, the minimum range is actually down to around 2 cm or so. The wide cone and minimum range of these units makes them excellent for obstacle detection. Other units, such as those based on ther polaroid rangers have narrower cones and a much larger minimum range (whichg can be lowered some as you mention). The narrower cone of the latter makes them a bit more suitable for mapping. I use the devantechs quite a bit and they work pretty well, especially for obstacle detection and location.
Probably the best performing units for the money are the Sharp GPXXX modules. They DO have a pretty narrow cone of detection which makes them less than ideal for obstacle sensing (although I have used them as "invisible whiskers" with some success).
Note also that sonar units (on multipel robots) can also interfere with one another, but the effects at least for a couple of robots can generally be mitigated somewhat in software.
The OP may consider using both in concert. A single Sharp unit combined with a devantech sonar can do a pretty good job of both mapping AND obstacle detection.
Has anyone actually used the thereminvision sensor on a working robot yet? I have the unit, but haven't had a chance to play with it as of yet.
Cheers -- tAfkaks

Geez -- I really need to remember to spell check.
I forgot to mention that while the Sharp sensors perform reasonably well, you do need to take care powering them if you intend to use more than a couple, as they can induce a substantial amount of noise on the power bus. While the current draw of these units is typically listed at 30ma, they actually flash the IR LED source at 1khz, with a 10% duty cycle. When the LED is on, the unit draws around 300ma -- thus the 30ma figure is actually an average.

Thanks all for reply. Good tips, pros and cons.
looks very attractive with use of single numbers of IR and sonars. I am going to open another thread and ask if anybody was successful in using thereminvision sonsors.
I am trying to assemble a robot similar to thing scares me that it uses 12 sonars that cost $360. Interesting if Devantech sells them cheaper when purchased in bulk.
On the other side all sonars (in above project) are connected to I2C and do not consume valuable analog inputs that is great solution. Is there a project with schematic that connects IRs without consuming several I/O lines?
Ek.

You might try the site out for buying sensors. They were pretty mnuch the first and have good doc's too for many items. Another site is the site for sensors as well. these are sort of the Portland Robotics people, and getting a nifty MarkIII robot is a good deal too. It is a great way to get your feet wet.
One way to use a bunch or IR sensors is to use another MCU to handle the IR sensors and query it using I2C or TTL serial, such as those found at the PAKIII and PAKIV I/O processors. these work well and are easy to hook up and use.

The Sharp GP2xx units are nice, but they do have one drawback in that they don't seem to be super reliable on a moving bot, or when scanned on a servo turret/etc. This is related to the fact that they use triangulation on the reflected beam, and average several results over 40 msec - IIRC, 20 samples or so. If the reflection changes significantly over this period, then the readings can be fluke-y.
Some of this is discussed in the following appnote [I believe this is the right one] ...
Note: Distance Measuring Sensors
I'd be interested in hearing whether anyone else feels there is a reliability problem when using these sensors on a moving bot. - dan michaels =====================

Thanks again all!
What is resonable set of sensors for robot with 12"x12" size?
What would be your selection? Please be concrete if possible.

Thanks again all!
What is resonable set of sensors for robot with 12"x12" size?
What would be your selection? Please be concrete if possible.

You didn't mention the height, which is generally significant as well. You also want to consider exactly waht you want the robot to do and where you intend to operate it. But offhand I'd consider:
Motor stall sensing for when all else fails (though wheels will often skid in the presence of obstacles rendering this useless)
A mechanical bumper. The more the area covered, the better. You do want to be careful to design it so that it doesn't end up getting caught on stuff and torn off of your robot. A single wrap-around piece is ideal.
I like the devantech sonar units for onstacle detection and limited mapping. They have a relatively wide cone. At least two.
For longer range sonar mapping, a single polaroid sonar mounted on a server works well.
There are lots of other possibilities, depending on what you want to do. You can get a small wireless camera, temperature sensors, etc. Passive IR sensors are easy to find (and cheap) to detect the presence of (moving) humans or animals.

Thanks again all!
What is resonable set of sensors for robot with 12"x12" size?
What would be your selection? Please be concrete if possible.

"The Artist Formerly Known as Kap'n Salty" wrote I think to minimize the height as much as possible down to 12inch. Video camera (firewire) will be mounted on top or on regular stand attached to the top (up to 20"). I have in mind a few applications, but first want to make a robot that maps surrounding (inhouse) area and avoids obstacles, and manipulated through web interface via wireless connection. It will have a charging station and will carry tools (attached manually) one at a time for different purposes (vacuum cleaner, electrical ant killer (do not laugh), etc.)
Is there any good link on mechanical part of the bumper? I want to start making it immediately :-)
You know the height of the robot now, would you mind to give me an example where you would locate sensors and their types based on the data available from this thread? I would like to have practical examples to start from. Every post/email is very valuable for me; I am collecting opinions to make a decision within short time.

What about compass, wheel encoders? Are they a must?

What about compass, wheel encoders? Are they a must?