Need to make a roof/sky sensor, and What's the Blue Enhanced Photodiode good fo

Dennis Ferron

For a robot contest, I have to make a robot circumnavigate a building through its parking lot. It occurred to me that if the robot could reliably "see" the dividing line between the roof of the building and the sky behind it, it could use the angle to triangulate its distance from the base of the building, and thus maintain a safe distance all the way around as it circles the building. (I'm building a pan/tilt mount, and the robot will scan around with a single sensor until the sensor reads the high or low value that indicates the transition from roof to sky.)

I bought the 10 degree FOV noncontact temperature sensor, and I'm hoping it will see a different temperature for the roof than the sky, but I'd like to look at other methods as well in case that doesn't pan out. I've been looking at this: http://www.parallax.com/Store/Sensors/ColorLight/tabid/175/CategoryID/50/List/0/Level/a/ProductID/176/Default.aspx?SortField=ProductName,ProductName

It's called a "blue enhanced photodiode" but it has no part number or datasheet associated with it (hey, Parallax, where's the documentation on this item?). What's this meant for? Is it supposed to be part of a triplet of photodiodes for color sensing? Will it respond to blue sky over grey building?

I'm also planning on making a CdS array sun-sensor, based on this:
http://www.seattlerobotics.org/encoder/jun99/dusan.html

That might not help me locate the building roof, but might help me with dead reckoning by finding the sun, and could tell me when I'm in the shadow of the building.

Ugha

How tall is the building? A long range IR or sonic rangefinder might work if the building's not too tall...

Your best bet though is most likely just a compass module and a preprogrammed course.

PS: I'm curious as heck about the blue enhanced CDS cell as well.

Tracy Allen

The photodiode is (or very similar to) the Advanced Photonix PDV-113, data sheet attached.

Blue enhanced has nothing to do with color discrimination per se. It simply means that its response in the blue is slightly improved over that of a standard silicon photodiode. The peak response is still in the infrared, around 950 nm. If you want to use it to detect blue, you'll have to add filters. It is an interesting problem. What if it is cloudy? Bees detect the polarization of blue sky and use that for navigation.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Tracy Allen
www.emesystems.com

SRLM

I wouldn't rely on weather at all (blue sky, sun sensor, etc.) since that can change. I assume you're working towards the Sparkfun Autonomous Robot challenge? You know the parking lot, why not just make a corridor based on GPS that you can travel along? Sparkfun even sells a 5Hz GPS that would be perfect for this. Anyway, back to the weather: they said that they'll run the contest rain, snow or shine, so you probably want to plan for 100% of all possible conditions, not just 33%.

BTW, if it is the Sparkfun contest, good luck!

Dennis Ferron

Haha, small world, yes SRLM it is the Sparkfun contest! And thanks. Rather than repeat myself, I'll invite you to read my blog post about my plans to go to the contest:

dennisferron.blogspot.com/

One thing I have going for me is my robot is so big, nothing other than actually running into a parked car is going to stop it. It drives right over curbs, brush, and saplings. Only problem is dead reckoning navigation based on the wheel encoders isn't going to be very accurate because the wheels always spin and slip a lot when it's getting going; maybe an accelerometer could detect that and correct for it.

Thanks for pointing out that sky-based navigation could be foiled by inclement weather. Although, perhaps cold clouds could actually make for better temperature based differentiation. Someone suggested bouncing a laser to detect the roof line. This could work but I'm worried about alignment; it seems like it would be really hard to get it so the sensor lined up with the reflected light from the beam, and then too much jostling about could knock it out of alignment again.

I do have a compass module, so maybe the sun sensor isn't all that useful, unless I want to correct for magnetic anomalies in the compass. Lots of people have suggested GPS but I'm worried the accuracy of the GPS will be too sloppy to get an accurate fix, and I don't know if I have time to program for all the sensors I've already got. I figure the Parallax GPS sensor will probably be the easiest to interface, but might not be the most accurate compared to newer ones Sparkfun has. Basically I haven't wanted to spend the money on the GPS, although I've spent just as much on other things on the robot...

Thanks for explaining the blue enhanced photodiode Tracy. Kind of what I thought it was, but I wasn't sure. It would work for visible light stuff but I think for something like the sun sensor I'd still want a CdS photocell because they have a much more linear response. It looks like with the photodiode it would go from nonconducting to conducting so sharply that it would be difficult to get a reading from other than yes/no. Still, with its faster response, it could be useful for motion detecting in some applications.

SRLM

Well, the GPS is a winning idea. Carnegie Mellon used it for the first two DARPA Grand Challenges: they got the waypoints of the route and from that created a much more detailed route, with a more coordinates, corridor widths, and top speed constants. Although they didn't win (due to other malfunctions), the scheme worked. Just a thought. I don't think dead reckoning will be accurate enough for what you need (unless you have a really nice accelerometer/gyroscope unit...). I haven't looked at the rules recently, but I assume you can't place beacons on the course?

Nick McClick

I remember reading, a long time ago, about a robot challenge where robots had to identify poles and boxes. Most teams went with visual recognition, inertial navigation, etc. The guy who scored 2nd place spent $100 on his robot;

Imagine a pole, about 6 feet high with a pizza pan attached horizontally to the top. The pole is attached vertically to an RC car. When the pizza pan makes contact with something, it triggers a switch and the RC car changes direction.

I thought it was a pretty cool idea, especially since it beat teams that spent hundreds of hours on their design and spent a lot of money, too. I don't remember the rules of the sparkfun contest, but it's just a thought.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Gadget Gangster - Share your Electronic Projects - Sign up as a Designer and get a free 4-pack of Project Boards!

Tracy Allen

Hi Dennis,

I agree that the blue sky detector isn't viable for the contest But on topic, the photodiode is actually exceptionally linear, over 6 or 7 orders of magnitude, in its response of current vs illumination. Far more so than a CdS photocell. That is why photodiodes are used for quantitative light measurement. But linear is not what you need. The levels of illumination experienced by a robot might change over several orders of magnitude. A log-linear response would be better, where the sensor output makes a linear increase with each doubling or tenfold increase of light level. The open circuit voltage of a photodiode, or the CdS cell, are more like that.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Tracy Allen
www.emesystems.com

John Abshier

Carnegie Mellon used a differential GPS that provided better accuracy than the Parallax GPS or the ones Sparkfun sells. The differential GPSs require a paid subscription for the correction data. Stanford's Urban Challenge vehicle used an Applanix GPS set with 35 cm positional accuracy.

John Abshier

W9GFO

Dennis Ferron said...
Although, perhaps cold clouds could actually make for better temperature based differentiation.

I think you'll find that clouds are a lot warmer than clear sky. It is 32 deg here in Everett now, completely overcast and snowing. The temp of the sky is 20 deg. A clear sky is something like -40 deg.

Rich H

SRLM

About the thermometer to determine rook skyline, take a look at this paper. It mentions that IR temp based horizon sensing doesn't work well in fog or clouds.

http://www.eganfamily.id.au/archive30nov2007/monash/research/papers/TaylorBristol2003.pdf