Nov 23, 2007

Why is my robot not perfect? (part 2)

So here is the follow up to the post I made last week. Click here to refresh your memory.

There were a few different points of view with the theory so what better way to proceed than some real live tests.

Test Setup:
NXT robot, 2 wheels on a smoothish table.
2 white lines, 500mm apart
Program was written to wait 2 seconds then drive forward 1024 degrees. (the 2 seconds allowed the robot to settle and let me get my hand out of the way before it started driving)



1st test
75% power 1024 degrees, unloaded. I ran this 3 times and it was spot on in every case.

2nd test
50% power 1024 degrees, unloaded. I wanted to try and eliminate slippage with the surface of the table. If there was slippage at high power, it would be less at low power. Again all 3 runs were spot on.

3rd test
75% power 1023 degrees fully loaded. I found a nice heavy book to place over the top (2.2kg). The wheel does deform by a noticeable amount.


This time when I ran it, the robot came consistantly 30mm short. Vary the power and get the same result, 30mm short.

So if there is no slippage with the table, and changing of the wheel diameter does not affect the distance the only thing left is slippage between the wheel and the tyre.

I got some whiteout and marked both the hub and the tire and ran the experiment again. Turns out there is a small amount of slippage, although not enough to account for all the error.



Conclusion:
I think it's a combination of hub/tyre slippage and slightly deformed wheel diameter. The rubber is soft enough to compress itself slightly changing the diameter.

What do you think? Post in the comments.

--
Damien Kee

8 comments:

Talking Chimp said...

I have noticed the same thing when using the "View" program to measure motor rotational degrees. (E.g., for a 90-degree turn). The number of degrees varies depending on how hard I press down on the robot when rotating the tire (i.e., making the turn manually)

CATpit contractor said...

I assume that when running your tests with the white-out, you ran it with both loaded and unloaded robots?

How did they compare?

Damien Kee said...

Yes I did a test with the whiteout unloaded and got no noticeable deflection over that distance. I might do another test over a much larger distance with it unloaded.

Anonymous said...

Is the robot being dragged or are you using a caster for a third wheel.
-- just wondering how much dragging would contribute to wheel slippage.

Damien Kee said...

There is a castor there and only a straight line track. No dragging that I can see.

AlexD said...

Do you think there may still be energy stored in the deformed tyre? Not enough to move the robot the last 3 cm, but there if you unload it. I've been experimenting with wind-up motors lately, and I notice that they tend to retain some residual energy which is not enough to push the whole car, but if you lift it it can spin the wheels for a while.

Parax said...

1.7mm deflection is the cause!

The Math:
(1023/500)*360=175.95mm Wheel Circumfrence
175.95/(Pi*2)=28.00mm Radius Unloaded

With Load:
(1023/470)*360=165.40mm effective circumfrence
165.40/(pi*2) = 26.32mm Radius Loaded

Meaning the load deflects the wheel by 1.7mm, Measured as axle to table.

note that the wheel is defelected at the bottom only so is not round. The circumfrence of the wheel does not actually change but is distorted so appears smaller. It is the rolling radius that changes.
(see http://www.club80-90syncro.co.uk/Syncro_website/TechnicalPages/TRC%20calculator.htm)

Parax said...

P.S if you noticed that the whiteout marks also rotated in the wrong direction, this would be caused by mechanical precession (http://en.wikipedia.org/wiki/Precession_%28mechanical%29)

Related Posts Plugin for WordPress, Blogger...