rangefinding/object detection

[sciguy125]

I'm working on a project that requires rangefinding. I was wondering what people's experiences have been with ultrasonics and IR range finders.

The problem is that I need to be able to detect a wide range of objects over a large field of view. The range doesn't need to be more than a few feet (6' or 7' might be enough), but it does have to be able to detect narrow objects (maybe an upright broom stick, picking up a chainlink fence would be awesome). Essentially, I need to be able to detect objects that I might encounter while walking down the street. Resolution doesn't have to be that great either, a foot or so is probably fine.

My main concern is the field of view. I suspect that ultrasonics can pick up narrow objects, but only near the center. I also have a feeling that the IR sensors have a very narrow field. Ideally, I'd have 360 degrees, but I might be able to settle for 180. Multiple sensors can be an option, provided that they aren't too expensive, or too bulky.

Sorry I can't be more specific, but I can't talk about the project too much at this stage.

[Gdeaver]

Sharp makes a IR distance sensor that is good out to 5 feet. It has a narrow cone of detection. Ultrasonics can sense farther out but have a wide cone of detection. The outdoor environment is a nasty place for IR and ultrasonics. You might consider starting to experiment indoors in a small room. A couple years ago I work with some students on something like this. We built a Vex square bot and mounted a Sharp IR on a servo. The Sharp IR is a simple A to D device. If your ADC readings have allot of bounce, try soldering a 100uf cap across the 5+ and ground at the sensor. Basically you move the servo take an IR reading move again take a reading. Remember it takes time to move the servo so you need a little delay before taking a sample. With this you can make a simple object avoidance robot. After you get this down you could start to play with environment mapping. Add an encoder to measure how far the robot goes and relate that to the sensor data. Its a case of banging a matrix. Take little steps. Tackling the outdoors maybe a little much at first.

[Phalanx]

During our most recent season and during the off season we've been working the Maxbotic Sonic Range Finder and have had excellent success with it.

You may want to check out this thread for some insights as well as a link to the datasheet for it as well.
http://www.chiefdelphi.com/forums/showthread.php?t=46912

[Sparks333]

Our team also used the MaxBotix sensor as well, and actually got a good auto mode working on it using four of them at once. They are extremely reliable, have terrific interfacing characteristics (PWM, Analog, RS232) and have a trigger mechanism that makes it easy to use several at once without reflection from its own sensors. The sensor is great, but, like Gdeaver, I'm dubious about any ultrasonic sensor's ability outdoors, save with lots of postprocessing.

IR is a slippery little beasie, though it does work rather well when used correctly. Some opaque materials are transparent to IR, and IR loves to spread and can be picked up very easily from almost any direction without proper shielding (try holding a TV remote behind your hand. The signal still gets there, by bouncing off everything in sight until it reaches the receiver. It''s magical). It also behaves like light, in that if it hits a reflective object at an angles worse than 90 degrees, you will have a substantially reduced return signal compared to something like ultrasonic. We tested this by bouncing both an Sharp IR beam and a MaxBotix ultrasonic wave off a spider leg cap, and slowly twisting the cap until the sensor could not see it anymore. The IR lost sight after a paltry 15-20 degrees, while the ultrasonic lost it when it was practically on edge.

If this is a project with a lot of resources and computing power, try LADAR - very expensive, but very accurate and 180 (or so) field of vision.

Hope this helps.

[jgannon]

Your absolute best bet is a laser range finder, but they are probably much more pricey than you're interested in. Here's one popular example:

http://www.acroname.com/robotics/parts/R283-HOKUYO-LASER1.html

I have worked with the Polaroid sonar sensor and the Sharp IR sensor. Both are very popular for robotics, but they have their shortcomings. The Polaroid has a field of view of roughly 25º, which makes it pretty difficult to get any worthwhile angular resolution. The Sharp is better in that respect, but I've found it to become mostly worthless beyond even three feet, and that's with the obstacle being a sheet of wood aligned parallel to the plane of the sensor. If what you're trying to detect isn't flat or well-aligned, your readings become much noisier.

In short, reliably detecting things as narrow as a broomstick will be exceptionally difficult without a laser or a whole bunch of redundant sensors. If you go the latter route, keep in mind that trying to read from multiple sensors at the exact same time may cause some interference... it will take some careful coding to avoid this. Also, if you find yourself with a whole bunch of sonar sensors, expect to be driven nuts by the incessant clicking.

Best of luck on your project... keep us posted on your progress.

[Robostang 548]

My team used the sharp sensor this year on our bot. It worked great for what we needed it to do. We mounted it on the front of our bot and for autonomous mode we just pointed the robot in the direction we wanted it to go and went forward at half speed until the ir picked somthing up (usualy a wall or the rack). I know it worked everytime because I could see that while the robot would hit a wall it would stop moving the drive motors. Next year we plan on using the sharp IR's again only for longer range detection.

-Don

[basicxman]

low cst options are
sharp ir
ping ultrasonic
maxbotic ez-1
i think the srf08 is a little pricey but works great!

check www.robotshop.ca, www.hvwtech.com, www.netmedia.com sells maxbotix ez1's

[David Fort]

I presume this info is too late to be useful, but we found that the Vex ultrasonic sensor picks up chain link fence about 90% of the time if you are about 4 feet away from it and the incident angle is 90 degrees.

At 10 inches, it almost always doesn't see it.

We also used the Vex sonar in this year's FTC challenge QuadQuandry to find the post on the goal. We found you could see the post reliably (the cone angle on the sensor is pretty narrow - maybe 15-20 degrees, so we swept the sensor back and forth to find it).

We earlier tried to pick up the distance to the base on the FTC goal, but found it didn't see the surfaces when viewed from the vertex of the hexagonal base.

For surfaces orthogonal to the beam measurements up to 170 inches seem pretty good. (i think it could go further than that, but that is where 16 bits overflows on the Vex timer)

[joshyboy9987]

I was thinking about this a little while myself. We're thinking of having a similar system for our robot in this year's competition. What if you used an Ultrasonic sensor to do a wide sweep, and then if it found anything unusual, you could use an IR or laser rangefinder to do a fine resolution sweep? You could have all of that stuff mounted on servos too, just so that you could get the direction the obstacle is in.

[ubergeek5075]

Our team, for this year's game, is planning on doing obstacle avoidance using distance sensors. The one's we're using are the ultrasonic Maxbotix ones. However, we found problems with using ultrasonics, they have a wide FOV that you cannot tell where exactly an object is in this FOV. (our FOV is about 2 feet wide at worst with a Maxbotix EZ1 unshielded). It just returns the closest distance it finds. Also, the maximum sample rate is 20Hz meaning scanning was ruled out of the question, especially if we're moving.

We did toy with the idea of putting a couple on one servo, offset so reading would be faster, and more fine.

After a little creative coding, and a sensor mounted on last year's robot's camera, I got it turn to a clear path when an object came within range, and in its view. Just a little teaser for anyone paying attention to one of my other posts.

[John Hooper]

I was thinking about this a little while myself. We're thinking of having a similar system for our robot in this year's competition. What if you used an Ultrasonic sensor to do a wide sweep, and then if it found anything unusual, you could use an IR or laser rangefinder to do a fine resolution sweep? You could have all of that stuff mounted on servos too, just so that you could get the direction the obstacle is in.

Are we allowed to use laser rangefinders? I read one rule that said "no lasers" I thought. Then again, a lot of robots use lasers.

[Mark McLeod]

No lasers.


<R02>
ROBOT parts shall not be made from hazardous materials, be unsafe, or cause an unsafe condition. Items specifically PROHIBITED from use on the ROBOT include:
...
• Lasers of any type
...

[mluckham]

Our plan for wall-following calls for two side-mounted Maxbotic sonar sensors - keeping the readings similar in value should mean the robot is parallel to the wall. If one reading is significantly larger than the other, a slight turn is called for.

Another (or a pair) of sonar sensors at the front should do for collision-avoidance.

[Phalanx]

Our plan for wall-following calls for two side-mounted Maxbotic sonar sensors - keeping the readings similar in value should mean the robot is parallel to the wall. If one reading is significantly larger than the other, a slight turn is called for.

Another (or a pair) of sonar sensors at the front should do for collision-avoidance.

Remember that there will be 3 robots on the same side of the field and 3 on the other side if you make it across the side line. You may not always be in the center, and you may or may not have another robot right next to you as you travel. So I think your readings may vary greatly and frequently during the autonomous/hybrid period.

Something to think about.

[joshyboy9987]

Another thing too, is that the outside of the field is that narrow bar. That might be a little difficult to pick up. And you're right, no, we can't use lasers, but I was under the impression it's not for a FIRST competition? Not to mention lasers wouldn't work too well with the divider being plexi-glass or whatever...