Sensors for Mapping

-- Randy M. Dumse

Caution: Objects in mirror are more confused than they appear

Sure. Using Moravec's certainty grids as you move around will produce an image with more resolution than the beam width. It's neat to watch those images appear. I did that back in the 1980s.

It's also possible to go to much higher frequencies, like 1MHz, and get a 2 degree sonar beam width. Cybermotion used to have a device like that. The problem is that with a narrow beam width, you can't collect data very rapidly; you only get one point per round trip time, so scanning takes many seconds. So that's a dead end.

Hobby robotics sonars are still as dumb as they were twenty years ago. One would expect better systems, with multiple sensors and a DSP, but it's not happening.

Vision-based systems are starting to work; look up "Simultaneous Localization and Mapping" (SLAM). Structured vision, like triangulation-based rangefinders, is useful at short range.

True small-scale time of flight systems exist; see "

". There's no reason that thing needs to be expensive, although it currently is.

John Nagle

I am curious to know if there are any low-cost [I.e. under $500] phased-array sonar systems that are designed with a lot of on-board processing (via a DSP). I mean something very roughly comparable to the CMU cam, where a lot of functionality is built in. For sonar, I can see the possibility of a system with 3 microphones and one emitter that will send the user a simple 2D grid showing the intensity of the reflected signal. Such a system may also provide the ability to do very fast scans of an area.

Even if such a device is not available as a kit, perhaps a university has documented such a project enough that an average robotics hobbiest could put it together without know a lot of about digital signal processors.

Joe Dunfee

Hello Vaibhav ,

I am working on a sonar project for my small outdoor bot. I am using

40Khz piezo type transducers available from Jameco. I have not done any measurements on cone angle, at least not quantitative. My system is still in breadboard stage, but when I was testing it a few days ago, I found that at a range of 10 meters point to point (ie, I was aiming the transmitter at the receiver), I could turn the transmitter an estimated 30 degrees or so and the receiver would still respond (I have a beeper on it to alert me).

Now, there's a few ideas I have that I am going to try. First, I am going to use 2 transmitters, instead of one, placed on either side of the receiver but inside a 'waveguide' . This is for rangefinding. I read one of the articles posted earlier and someone talked of putting the transmitter in a tube lined with foil. OK, as sort of a waveguide (or in RF it would be an antenna).

The wavelength of 40Khz ultrasound traveling at, say 340 meters/sec is about 8.5 millimeters. That is the best length of your waveguide, at least according to physics (ie, 1 wavelength) . So that was something else I was going to try. Now, I don't know if it is going to help things any, but just thought I would mention it, since this thread caught my attention.

A waveguide is similar to an antenna but is used on the longer wavelengths (like microwave), but sound is not an electromagnetic wave, it is a pressure wave. It needs a medium to travel in. I will also be experimenting with how far the transmitters need to be from the receiver, and at what angle they should be at (I don't thing pointing straight ahead is the best angle to use, necessarily, especially for 'stereo' sonar) in order to maximize my rangefinding ability.

Have fun with your project,

Best Regards, Joe

Hello Joe,

There is a company here in Massachusetts called Massa Corp, they build phased array sonar imaging systems for the US Navy. Sonar is their specialty. I am pretty sure they have a web site. You may want to google for them and see if they also sell to the commercial market. I am sure they would, just as a matter of doing business.

I know you can use an array (matrix, say 8X8 transducers or so) and get an image. I don't know the exact details how. I figure, if I want an image, I will just use a photo, or video cam (on land that is, I don't know what your application is).

hth, Joe

They didn't show anything about phased arrays on their web page. They didn't have a web store, and obviously deal one-on-one with manufacturers.

I am not sure we are talking about the same thing... the "image" I am talking about is a 3-d map of obsticals. While this is sometimes done with stereo video images, by triangulating common points, it is definitely a challenge to do.

I am working on modifying a Roomba robotic floor vacuum so it can map out its surroundings.

Joe Dunfee