The 'blind' navigator robot

[stik]

WOW! Thank you for your responses!

What do you think the significance of them using the frequency sources as 1:2 frequencies (600Hz and 1200Hz) is? How would you utilize this fact? I also agree that this is probably not a coincidence that they have chosen it to be this way.

Arhi, I must agree with you on the amplitude attenuation, I don't think (according to my knowledge of the physics) that the amplitude will change very much over such a short distance. In your method, would you plan to calculate the phase difference received between the two mic's (placed at a distance apart from each other) by calculating the time difference between peak readings by either mic? I have read up on this and this technique is known as 'interaural time difference'. It is the basis upon which humans are able to localize sound. It seems like a feasible solution.

Do you think though that having the interference pattern it will be possible to adequately separate the two frequencies by filtering? Will the interference pattern be a 'standing' one, having distinct areas of interference, or will the pattern of interference be changing and repeating itself?

Bobledoux, we have been told that we will be able to make measurements at various areas in the test area, however I do not think this will happen very long before actual testing. Therefore I think it may be necessary, if possible, to develop a technique to calibrate the robot each time it begins its 'journey'. Odometry sounds like a good way to confirm distance travelled relative to the robots last point. I think this could be used to supplement the audio measurements. A comparison between the two measurements could be made.

 

[arhi]

stik, don't get me wrong but the whole direction of my posts is dictated by the fact that I'm pretty sure they choose 1:2 ratio and so low frequencies for reason so I am trying to find a way to exploit that option

you can fairly easy make a test environment, 600Hz and 1200Hz can be produced by most speakers, you can make sine generator if you do not have one with few spare parts (ne555 comes to mind ) and there's bunch of people here on forum able to help you with that (don't look at me - audio and me are not on the same page)... set up two microphones connect them to the scope and look what you can see on different positions .. (different orientations of the microphones, different distance between them ... )

all in all, we can theorize for days, make a setup, attach a scope and look at the signal it'l surely give you few good ideas

 

[bobledoux]

I hadn't thought of doppler. That idea has merit. So the change in sound frequency is a function of velocity. The two frequency shifts represent a vector of the robot's direction and speed.

 

[arhi]

bobledoux, Doppler is kinda pushing on as idea when sound and movement are involved ... in theory it should be easy to determine speed and direction using 2 mics and 2 sources .. now, that's theory, one need to do some math, make some tests and see
- is the uC fast enough to measure events with enough accuracy
- is the uC powerful enough to do realtime math that is needed to calculate movement vector
- is the mic and peak detection (or zero cross) sensitive/fast enough to give accurate data

In theory - it sounds easy ... but that's what differentiate theory and practice

what you need to consider is
- you have 2 sound sources, low freq, 90degrees, 1:2
- you have 2 mics
- you have slow moving object
- you cannot move mics -> this in fact means that intended solution was not to rotate object in order to simulate rotating mics
- intensity of the signal can be assumed to be constant all over the "plato"

One fairly important thing you will have to incorporate in the math is temperature that might be crucial for the calculation as sound speed differs with temp. What you can do is calibrate the "speed of sound" initially

 

[bobledoux]

The speed of sound varies by a small percentage by temperature.

Apparently you used a microcontroller board in your earlier studies. Consider that first as you are already familiar with it.

Even the lowest Microchip PIC performs a million instructions per second. Consider how far sound can travel in that time.

Real time math can be performed by microcontrollers. Look at C language software and their supported libraries. While its possible to make up these libraries in assembler language that's getting awfully primitive.

Consider your microphones have to pick up sounds in 360 degrees. How directional are they? Look at the use of a cone reflector in this application:
**broken link removed**

Regarding speed of sound. If this a concern you could move the robot toward a speaker at a known rate and measure the doppler change. That value, compared to expected value could be used to calibrate your measurement.

 

[bobledoux]

Stik,

Would you share with us your approach to navigation? I'd like to know how your project turns out. Also let us know of any other navigation techniques used by other groups in your class.

 

[ronp849]

I seem to remember a marine radio navigation system called Loran that used the interference of two or more radio signals. By counting the peaks and troughs the position could be tracked. I would think the same could be applied to the two audio signals. You might need to find out if the two sources are phase and frequency locked. The fact that one is twice the other might give you a pointer.

Ron