Maxbotix Sonar Sensor

[DonRotolo]

After reviewing the specs on the Maxbotix Sonar Sensor, I placed my order. $30 + shipping, I can't wait for it to show up (It'll be here tomorrow!).

Bob from Maxbotix made a post here but I don't believe there were many views.

The short version is that they sell a small (<<1" cube) inexpensive (<$30) sonar sensor that measures from 6* to 254 inches (21 feet!) in one inch increments - and has three outputs: Analog voltage, Serial data, and pulse width. Takes up to 20 reading per second at full speed, or can be easily throttled back.

*Also senses from 0 to 6", but reports 6" and never less.

I will certainly be investigating this sensor over the summer. IF we had only had such a sensor (and the code to utilize it), we might have been able to avoid the defensebot that bumped us, causing us to flip ourselves on the ramp instead of blocking Team 11, and we might be at nationals with 16 and 95 instead of home...

I urge anyone looking to incorporate range data into a design to check into it. Once I get it working on my Vex bot, I'll file another report.

Don

[Gdeaver]

The benefits of this sonar is the choice of interface. I'm thinking the A to D may be best because the TTL serial maybe tied up with the camera. The a to d is also very fast and easy. The cone of detection is also narrower and may benefit from being mounted on a servo and scan for objects. The problem with sonar is that if another robot uses sonar too they can interfere with each other. Maybe next year our team may have to put one on our bot and just let it ping constantly to mess with an opponents sonar auto strategy. For a range < 5' i like the sharp 5' IR sensor. Its great for things like wall following and obstacle avoidance. Put the Sharp Ir on a servo and scan to build up a vector that maps the distance to objects in range. With IR you can detect the difference between a table leg and a wall.

[Donut]

This is one sensor I've been looking at experimenting with in the off season. Possibly the best sonar sensor I've found so far, and the only with an analog input (thank god, I was worried if I wanted to use sonar next year that we might run out of interrupt ports real fast between it and encoders).

Are there any long range (really long range, enough to rival a sonar sensor) IR sensors out there? All I can ever find are Sharp, and though good the best of those is limited to around 6'.

My one worry is if sonar continues to grow in popularity. We're going to start running into problems of messing up each other's sensors, and it'll be interesting to see how we can get around that.

[DonRotolo]

Originally Posted by Donut

Are there any long range (really long range, enough to rival a sonar sensor) IR sensors out there?
....
We're going to start running into problems of messing up each other's sensors, and it'll be interesting to see how we can get around that.

IR sensors: The limiting factor is how much IR light you can throw, and that's generally not much. Ultrasonic is far better for mid-range (to, say, 30-40 feet max). Further than that and you are better off with Radio Frequency (e.g., RADAR). The issue is one of propagation - how far can the (transmitted thing) go?

For these kinds of distance sensors, there is one basic parameter that can be altered to avoid interference: Frequency. IR can use a different carrier (40 kHz is common, but 20-55 is possible), Ultrasonic the same (more limited though), and RADAR, well, there's a lot of MHz out there...

PS: Got my sensor today, will play tomorrow.

Don

[Donut]

Originally Posted by Don Rotolo

For these kinds of distance sensors, there is one basic parameter that can be altered to avoid interference: Frequency

Is the frequency easily changable on sonar sensors though? I know you can vary it, but I don't know how easy it is to do that, or if it will effect the performance of the sensor significantly.

[DonRotolo]

Originally Posted by Donut

Is the frequency easily changable on sonar sensors though? I know you can vary it, but I don't know how easy it is to do that, or if it will effect the performance of the sensor significantly.

Unfortunately, it's only easily varied on some designs - usually those you make yourself. Also, there is a resonant frequency for each transmitter, where you get peak efficiency. You can move off that a bit, but not by much - and in thinking about it more, probably not enough to avoid interference.

However, you might be able to avoid interference by knowing when you command a reading, and only 'listening' in the very small window of time that you expect an 'answer'. If you do that 10 times a second, a few readings should be valid. But, which ones....

Maybe modulate the ultrasonic, and only listen to the signal with the correct modulation?

Sorry, I just don't have the expertise to give a better answer.

Don

[Gdeaver]

The Daventech SRF235 is a higher frequency sonar with very narrow cone of detection. SRF 235 Read the info on it and you'll see the problem with very high frequency sonar. Also it's I2C and there would have to be a coproc involved. There are short range time of flight laser sensors but, First doesn't like lasers and the cost is very high.

[Donut]

The SRF235 isn't any more useful than the Long Range IR Sharp sensors, it only has a range of 1.2 meters.

[Sparks333]

Originally Posted by Don Rotolo

IR sensors: The limiting factor is how much IR light you can throw, and that's generally not much. Ultrasonic is far better for mid-range (to, say, 30-40 feet max). Further than that and you are better off with Radio Frequency (e.g., RADAR). The issue is one of propagation - how far can the (transmitted thing) go?

The other thing about IR is that it bounces really well, plus it penetrates things you'd think would be opaque. In terms of rangefinding, as long as you are looking at something rather large (say, a wall) and had a good outlier-removing algorithim, then, sure. Otherwise, I'd say don't do it.

Aside from that, if we're talking about non-FIRST activities, I've heard that an ultrasonic rangefinding ring can be notoriously difficult to read (on account of so many transmissions coming and going at the same time), so, if you want to delve into some crazy algorithims in OOP, some programmers swear by artificial neural networks. Apparently, they are very adept at going through messy data and pulling reasonable answers from them.

Sparks



Sparks

[Donut]

I'm kind of hoping to use these on an actual robot for competition.

I think we're just going to try to use the Maxbotix ones and hope we don't get interference from others if there are any.

[DonRotolo]

Got the sensor last week, played over the weekend, have some initial findings: (I used a VEX controller for everything, but this is similar to what to expect with the FIRST RC)

I connected it to the controller easily - soldered in a male header and made up a custom cable for the analog input. Took +5v and ground from the RC, and fed the analog signal back. The 10 mV/inch scaling is very convenient, since 2.54 volts = 254 inches. In the RC, a count of 512 is 2.50 volts (not 2.54), so I just scaled the reading by 1/2 and used that as the distance in inches. (Yes, a small error is introduced, but not much). Leaving the RX pin open causes it to read continuously, about 20 times a second.

It detects a pencil out to 24 inches, a metal spray can out well past 10 feet, and larger objects out to the maximum 254 inches (over 21 feet). Accuracy is good, with about a 1/4" error at a foot and 3" error at 12 feet (Measured 144 inches, actual was 141 inches). After I left it on about 40 minutes, the accuracy changed a little, now reading 138 inches (actual 141).

The cone of detection is about 30 degrees, didn't find any dead spots or lumps. It detects the nearest object in that cone, but can miss an object that is particularly small. Near the limits of range, the cone gets a bit narrower, to maybe 20 degrees at 21 feet. Soft, poorly reflective objects (like a cloth-covered chair back) are not detected as accurately, but that's a characteristic of ultrasonics.

When the distance to an object changes rapidly, the output takes a moment to catch up the the reading. Shouldn't be a problem at 20 readings per second. I tested this by putting my hand a foot away while the sensor was reading something at 12 feet.

At 20 readings per second, sometimes the values jumped plus or minus one or two counts - remember, this was the analog signal - so I put in a loop to average 5 readings at a time, which resolved that. Generally readings were steady.

Overall, seems perfect for the kind of stuff FIRST is doing. It sure would have helped us in autonomous defense at Palmetto. I think teams would want to mount it on a servo and scan in seven steps to evaluate 180 degrees, and act on the data accordingly.

One really nice feature is that you can use the RX input with a digital output to turn the 'pings' on and off. If interference is likely, only ping when you need to, increasing the chance of hearing your own ping and ignoring others.

Of course, it has serial data (RS-232 format at TTL levels) and pulse width (read with interrupts) outputs, making it very versatile.

Let me know if there are any specific tests you want me to make. meanwhile I need to develop an algorithm to detect and avoid objects autonomously with my squarebot.

Don

[spears312]

Ive been thinking of talking with some of the programming guys on my team to determine the possibility of something like this and maybe developing a quasi-radar sort of thing. It would certainly be an interesting addition to the sensory capabilities of a robot.

[Sparks333]

If that's for FIRST, then I'd be careful. FIRST doesn't like transmissions that aren't its own (I have heard them exempt wireless cameras, but not to the extent of being able to use them to help with driving) and RADAR would involve lots of transmissions that would directly affect the outcome of the match. It would have to be more 'quasi' and less 'radar' to qualify.

Sparks

[Gdeaver]

Don,
The real test is to have the sonar on a bot executing an autonomous move and have another sonar firing at it. How bad is the interference? I only have 1 SRF04 or I would do some testing. I suspect that there will be problems with multiple sonars on the field. This could be problematic for Vex competitions given the smaller field. Because of the smaller Vex field, the Sharp 5' IR sensor is a natural if only Vexlabs would package it and make it an official VEX sensor.

[Donut]

Originally Posted by Gdeaver

Don,
The real test is to have the sonar on a bot executing an autonomous move and have another sonar firing at it. How bad is the interference? I only have 1 SRF04 or I would do some testing. I suspect that there will be problems with multiple sonars on the field. This could be problematic for Vex competitions given the smaller field. Because of the smaller Vex field, the Sharp 5' IR sensor is a natural if only Vexlabs would package it and make it an official VEX sensor.

I would suspect that since they're all running on the same frequency, the error could be terrible just depending on the luckiness of when their ultrasonic pulses hit your sensor (if you get really unlucky it hits you right after your ping, giving you a reading of 6 inches when they're 20 feet away).

This is part of the reason I'm still considering IR. A newer Sharp sensor has a range of 10' and has a significantly smaller beam (5 degrees), which can be both a plus or minus. The beam is small enough and and has a shorter range that I think interference would be very uncommon from other IR sensors (considering the same type of IR sensors are used on almost all mini-sumo bots without much, if any, interference, I think there won't be much problem on the FIRST fields). If FIRST were to introduce the IR beacon back into the game (from 2004), could this potentially disrupt the Sharp sensors?