Ultra-sonic vs Light sensors

[duvivier]

Hi folks, I'm want to develop an automated car that always moves forward, and turns when it reaches obstacles.
I have doubts about what kind of sensor of proximity should I use. I don't need to know the distance from the obstacles, just to turn the sensors ON/OFF while they get close to the walls.
I've heard that the IR sensors changes their range of activation accordingly to the color of the surfaces of reflection, and that both ultra-sonic and light sensors doesn't activates if the surface of reflection has a big angle with the sensors (this is my most important doubt).
Which of these do you recommend since the car should work outdoor??

 

[Roboticinfo]

reply

Have you heard of the competition called the DARPA challenge?

Well what you are doing sounds like it won't work due to the complex
Programing required.
Since you are asking such a simple question for a complex project I have my disbelieves.

People spend 30,000 to 200,000 dollars on this sort of project and they end up not even working.

I recommend you due a little project like a small rc car or a robot.

 

[zachtheterrible]

Complex programming!?!? Just how hard can something like this be??

I've never built a robot and I could probably make this entire thing from the ground up with discrete components! In fact there are cheap kits that you can buy that do just this.

duvivier, I guess that it kind of depends on what "obstacles" you plan to be dodging. If they are flat, then ultrasonic would be a sinch, but I would imagine the sound might be absorbed by a blanket or something soft. And yes, if the angle is large enough, the sound waves would probably reflect away.

I would personally use IR (but thats just me :lol: ). You could have a 38khz modulated beam and an IR sensor out of a TV that will only pick up that 38khz IR and no other light, to a reasonable extent (don't put it in direct sunlight without some kind of shade). Depending on how far away you want the robot to stop and turn from the object in front of it, you could adjust the intensity of the beam. That is true what you said about it reflecting off different colors, it would probably stop and turn closer to a black surface than a white surface because the sensor would have to be closer to the black to see the IR reflection.

 

[Roboticinfo]

reply

Zach this is a car not a small toy, I can assure you there is no kits out there for an automated car.

 

[zachtheterrible]

Car? as in one that you drive in?

I thought he was talking about a small robot car.

 

[Roboticinfo]

reply

No he is talking about a life size car, that is why I said make something small.

 

[DigiTan]

Well, I don't have many details for you, but this is what I gather so far...

Laser and IR will give you the smallest beam width and highest pulse rate, which normally translates into better resolution if you're using a sweep pattern. The receiver would have to be specially designed for outdoor uses or else there's a great chance the signal will be masked by sunlight, even if you modulate it. A nicely-focused laser like the type used in police lidar should scatter well enough return from some targets. Without a rangefinding function, the sensor will be misled by target color.

Right now I'm experimenting with a sonar device from EDPCompany.com called "SonaSwitch Mini A/S." It's an embedded sonar and it can reportedly detect objects as far as 40 ft, once configured for that. The "ping" rate however, is much slower that what you could get from a laser, so that is already extending my sweep time. The echos do not scatter as well as light, so some targets will remain hidden.

Well, I guess this isn't too imaginative, but my advice is to use both in a redundant fashion. There are a lot of UGV competitions on the web, and generally the ones that perform the best do so only because they are able to combine sensing methods.

 

[Roboticinfo]

reply

I recommend that you do an easier project
because that is a project that is harder then you think.

 

[DigiTan]

Eh...that's a little patronizing don't you think? I always say it's only as difficult as people make it.

But this is a difficult project, especially since you'll have a lot of non-ideal conditions to work with. Maybe you can tell us what type of objects will be on the test course. 'Might narrow some things down.