Subject:
|
RE: AI and even more exiciting stuff
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Tue, 19 Oct 1999 00:53:42 GMT
|
Viewed:
|
564 times
|
| |
|
|
> 32K is a ton of memory for embedded 8 bit code... if you need data storage
> and processing then that's an entirely different thing -- it could easily
> make AI difficult. As far as fixed point vs. floating point, I
> once worked
> on a project where they did the control loop calculations in fixed point.
> It took literally days for them to rework the code every time you changed
> the control data tables. Since this project was really R&D even
> though they
> wanted it to be pre-production, it was a real pain. When you are
> doing R&D
> type work you should not waste time optimizing to fit your resources you
> should fit your resources to optimize your time.
> In this case (under discussion) it is a big jump in $$$ and convenience to
> go up from the RCX to a more capable controller platform (and no, the
> handyboard does not help here). In my experience fixed point
> math tends to
> dictate fixed control algorithms and fixed control constants whereas
> floating point tends to allow you to make changes just about as
> fast as you
> can recompile. Am I wrong?
Ummm, respectfully, yes. If the team took days to rework the code when
control data tables change, then they were not using tools and methods
appropriate to the task at hand. I'm hoping you weren't "them" :-)
The algorithm is the same whether or not you use fixed point math. The
constants and their REPRESENTATION may change from floats to fixed,
but that should be easy to handle.
In most cases, the conversion from floating point to fixed point should
be handled by an automatic tool. I have used awk, perl, and even TCL
to massage floating point numbers to the appropriate fixed point format
for tables.
Geez, the more we talk about this stuff, the less confident I am that
anyone is paying attention in school. Or maybe they don't have enough
real-world experience put into school environmnents - the common argument
being that they're only trying to get the idea across, and the execution
doesn't really matter.
Oops, time to get off the soapbox again, sorry.
|
|
Message has 1 Reply:
 | | Re: AI and even more exiciting stuff
|
| Ralph Hempel wrote: [ good rebuttal snipped ] (...) Actually, I went through two years of a systems engineering degree. (before finally dropping out) During that time I had many semesters worth of math and CS... however, not one of them went over (...) (25 years ago, 19-Oct-99, to lugnet.robotics)
|
Message is in Reply To:
| | RE: AI and even more exiciting stuff
|
| 32K is a ton of memory for embedded 8 bit code... if you need data storage and processing then that's an entirely different thing -- it could easily make AI difficult. As far as fixed point vs. floating point, I once worked on a project where they (...) (25 years ago, 18-Oct-99, to lugnet.robotics)
|
3 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
|
|
|
|
Active threads in Robotics
|
|
|
|
|
|
