DPRG List  

[DPRG] Another encoder counting problem

Subject: [DPRG] Another encoder counting problem
From: Mark Castelluccio markcastelluccio at home.com
Date: Thu Aug 30 00:33:58 CDT 2001

I have not tried the UMBMark test.  I calibrated mine by spinning it in place by
hand about 10 times and then aligning it to reference marks and reading out the
measured angle.  Then I modified the wheel differences accordingly.  To get
straight distance, I just ran the bot and measured the distance with a tape
measure.  I will have to try the UMBMark test sometime.


>-----Original Message-----
>From: dprglist-admin at dprg.org [mailto:dprglist-admin at dprg.org]On Behalf
>Of David P. Anderson
>Sent: Wednesday, August 29, 2001 9:56 PM
>To: dprglist at dprg.org
>Cc: markcastelluccio at home.com
>Subject: [DPRG] Another encoder counting problem
>Mark C. writes:
>> I found that the two wheels are often slightly different in size.  I had two
>> wheels made on a lathe to the same diameter but after adding an
>o-ring, they were
>> two different sizes.  To account for odd inch/tic ratios and still
>avoid floating
>> point, I calibrated everything in micro-inches.  For my wheels, the left wheel
>> was about 28823 micro-inches, the right wheel was 28871 micro-inches.
> With a 4
>> byte int, I could perform arbitrary navigation in a 200 foot square room if I
>> remember correctly.
>Hi Mark!
>I also have found two wheels almost never exactly the same size, and this throws
>off position calculations after a while.  The Borenstein UMBMark benchmark has
>a method for separating out errors related to imprecise wheel base specification
>and miss-matched wheels.  For one of my robots with 5 inch wheels the calibrated
>wheels are not really 80 ticks per inch but (forgiving my floating pointedness)
>#define LEFT_CLICKS_PER_INCH 79.52
>#define RIGHT_CLICKS_PER_INCH 79.48
>#define WHEEL_BASE 9.7325
>For the lego robot with lego "tractor" wheels, it worked out as
>#define CLICKS_PER_INCH 48.42
>#define WHEEL_DIFF -0.010
>#define WHEEL_BASE 7.01675
>Seems like pretty small differences but these were arrived at empirically
>running the UMBMark repeatedly and refining the values until I could see no
>more improvement.
>It's a little bit like sighting in a rifle scope.  There is still an average
>variation for each run but the spread is centered over the "bulls eye" as
>it were.  The difference in wheel sizes shows up as a bias toward one side
>or the other.  Windage and elevation...
>good to hear from you!
>DPRGlist mailing list
>DPRGlist at dprg.org

More information about the DPRG mailing list