Subject:
|
Re: some stuff... robot accuracy
|
Newsgroups:
|
lugnet.org.ca.rtltoronto
|
Date:
|
Sun, 17 Feb 2002 18:56:46 GMT
|
Viewed:
|
585 times
|
| |
|
|
In lugnet.org.ca.rtltoronto, Chris Magno writes:
> any ideas on how to improve accuracy when your accuracy is not
> repeatable????
>
> Half the times the block is perfect but the other times it is always
> different, where do i look? its not like the inaccurate bloc is always
> in the same position the wrong way.
There isn't an answer--if your accuracy isn't tight, it won't ever be
repeatable unless you fix it somewhere else downstream of the system (ie, a
positioner chute like Derek's original idea for example)
The block that is off its position is never in the same place because likely
there's a probability distribution that varies where it will drop. It's
like some days the subway stops exactly at the start of the platform, and
other times it doesn't, so over time, where the people get out and leave
scuff marks on the floor builds a bell shape.
So say there's a tolerance that lets the brick drop into the interlaced
position, maybe under half a brick. If the variability of the EOAT and XY
axes puts you under a half of a brick, then you shouldn't have a problem.
But for most of us can't get that precise positioning, so sometimes we'll be
out of that tolerance required. The percentage we're out of that margin
equals how many times out of 100 we'll be misplacing blocks. It looks like
50% of your distribution curve for your robot is out of the margin for
getting into that location.
Worse off, there's other factors which will give you interactions. Maybe
the XY at one end of the board is off because the light gradient for the
target causes trouble. And this interacts with another sensor because now
you're not on position to get a light reading off the block. Etc etc etc.
So it all works up and it's hard to trace the problem, subsystem by
subsystem because even if you modify one variable, it might inextricably
linked with another variable somewhere else.
How do you reduce variance in the system? Well, to some degree it's
systematic-the tolerances (or lack thereof) build up to make it variable.
For example, the little bit of wobble of your EOAT when one of your axes
stops could contribute. It could also include the randomness that the XY
positioners put in. It could be the variability of the angle the brick is
dropped at. Reduce all the variabilites and you'll get it accurate, while
one variable might yield a good return, there ain't no silver bullet to fix
it all.
Calum
|
|
Message has 1 Reply:
Message is in Reply To:
 | | some stuff... robot accuracy
|
| any ideas on how to improve accuracy when your accuracy is not repeatable???? Half the times the block is perfect but the other times it is always different, where do i look? its not like the inaccurate bloc is always in the same position the wrong (...) (23 years ago, 17-Feb-02, to lugnet.org.ca.rtltoronto)
|
17 Messages in This Thread:
 
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
