Hi, I was wondering if you could help me with a problem.  I would like to know how far away a robot is from a certain stationary location, specifically my computer.  I would like to do this through walls, so that rules out any line-of-sight methods.  What would be a relatively cheap (sub-$100) way to do this?  GPS is way too expensive, so I was wondering if there could be some way of doing it with RF signals by measuring the time from transmission to reception.  Is that possible, or is the timing too inconsistant?

Unfortunately, it's unlike to find a sub-$100 decision for this.
I even doubt that a sub-$1000 solution exists, considering integration costs.

Well, the robot is going to be controlled by RF signals anyway, wouldn't hurt to give the propagation delay idea a shot.  

But if that doesn't work what about this idea?  Mount a digital compass on the robot to get the direction its facing and use an odometer or infra-red (or something) to get the speed of the wheels.  I could then transmit that data back to the computer and update the position of the robot on a map.  So if its facing south-west, traveling at 0.5 m/s, and I check every 50ms, I would update its position by 0.025m south-west. Does that sound feasible?

That's called dead reckoning, and has many problems.

What spec are you trying to achieve? To be honest you should be able to pick up a PCMIA GPS card for around 100 bucks, on the other hand the accuracy is not going to be enough for room navigation.

I think your original idea will not work unless you use some very expensive synchronization - a round trip of 100 m would take 0.3 microseconds.

How about setting up known targets in each room so that the robot knows where it is? This can be used to correct the estimated position.

Alternatively set up ultrasound emitters in each space and use these for triangulation.

Cheers

Greg Locock

The robot is going to be equipped with a small monochrome camera and relaying the video back to a computer.  So its position does not need to be 100% accurate as that information is not critical to its navigation.  It will also be equipped with infra-red proximity sensors so it does not bump into anything.  I just need to know what its general location is, not its exact location.  A few feet off would be acceptable.

Using known targets or ultrasound emitters would not be possible as I'd like the robot to be able to go into unfamiliar locales, where it would be unpractical to set up such things in advance.

I can imagine dead reckoning has problems, such as the initial setup position would be a factor, and the traction slipping would cause inaccuracies.  I realize the method has problems but unfortunately it might be my best solution (with my budget).  I'll look into the issues further.

$100 on GPS is a little more than I'm willing to spend on this particular aspect of the robot.

Thanks for your input.

If you've got a camera and adequate collison avoidance then dead reckoning is fine. After all, that is basically how we navigate!

The trick will be getting the robot to recognise visually where it is. Once you know that, update the known position and use that for your dead reckoning baseline.

You will still have to deal with anomalous routes, for instance if you can get to a room via a hall with a slippery floor or one with a carpet then your dead reckoning 'map' will be distorted.

Interesting project.

Cheers

Greg Locock