Easy motion detection

[Half Shell]

I've been debating how I can properly do motion detection on a stationary but swiveling sensor turret. I know motion detection is typically done with a stationary sensor that can read a distance and detects changes in that distance. Since the turret is moving and will be picking up constantly changing areas I figured that this would not work.

I have two ideas thusfar.

The first is to have the turret swivel and, for the first two or three back and forths, have it memorize the distance of what its looking at in comparison to the value that the servo is now at. This will create a simple array/lookup table that it will then compare. If there is an error above a certain size recheck it - if the distance is still bad then we have something inbetween us!

The next way was to scan but check multiple times at each stop (for the record, each stop is a short interval of time). If we see a significant change in the values recorded at each stop, we had movement.

Thoughts on how to properly do this? I've never had to work with a motion detecting robotic application before....

[Admin]

If your robot moves between the first and second scan (or during scans), it won't work . . .

At least not with a sharp IR on a servo with a microcontroller . . . Thats why my robot stopped when it did the adaptive mapping scans.

Otherwise, yeap, thats how I would do it.

If there is an error above a certain size recheck it

A threshold check

[SomeSaba]

What if:

you have a servo mounted scanner, and you scan acrross horiztonally, and u use an equation to find the distance based on ur ADC value, and if u know your velocity, u can start a timer and see how far u've gone in the interval between one scan and the other, so you should know how much closer you are to the objects,

....right?

[Half Shell]

Luckily my robot will be stationary all the time. If you've played the game Portal, it will be a recreation of their hilarious turrets there :-D.

As for doing the reading while you're moving, I suppose that would work. I think the only way to get very accurate readings using the second method I mentioned (measuring a change as you're scanning) would require a very accurate sensor, like laser.

[Admin]

yea, the method you guys are talking about is called SLAM. it requires a super faster scanning sensor (like a SICK laser or tons of sonar), and a heck of a lot of processing power to do all the math. state of the art stuff, really . . .