Subject:
|
Re: AI and even more exiciting stuff
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Mon, 18 Oct 1999 06:44:24 GMT
|
Viewed:
|
688 times
|
| |
|
|
Not that I have time, but hey... this is much more fun than actual work :)
On Mon, 18 Oct 1999, Tony Sceats wrote:
> > AI is a huge field, encompassing a variety of algorithms and techniques.
> > Some of them will work OK on the RCX, others take just too much
> > computation, especially when floating point is required.
> >
> > Also, when you talk about functions and all that, well- what OS are you
> > using? the standard lego software? NQC? legOS? Forth?
>
> I am using NQC at the moment. I have just found, and I am downloading as I
> type, legOS.
Good luck. You'll find links to my HOWTO in various places- feel free to
ask if you have problems.
> What is Forth? I have seen numerous people talk about it on
> this mail list - is it an assembly code compiler?
Forth is a language for embedded systems, and since it is designed for
embedded systems, (i.e., very low level) resembles assembly in some ways.
pbForth is Ralph Hempel's system for using Forth on the RCX.
> What I was talking about when I said Programs, are the 5 spaces available on
> the RCX itself for each program.
Those five slots are a relic of the original lego software, which some
systems (notably NQC) use, but which others (notably legOS and forth)
completely ignore.
> I know that NQC has limited the number of routines you can have in each
> program. Surely this is not as limited as it seems (I think 10 routines?!).
Unfortunately, that is the case, since NQC inherits the limitations of
the standard lego software and is merely a different interface to the
same underlying firware. legOS and forth get rid of that firmware, giving
them more flexibility.
> I can understand limits on variables and instructions, as there are only so
> many registars in a CPU, but can we get some RAM management system going, so
> that we can use the extra program spaces as some kind of swap space, like
> windows Virtual Memory?!
Well, no need to get that complicated- a "real" OS like legOS will take
care of that for you.
> Not that I am need of the room (yet...). I have only pretty recently
> actually got this stuff, and I just want to know my limits, as I am a
> curious creature myself.
Well, your main limits (once you get rid of the standard firmware) are:
1) memory. While you'd be surprised how big 32K actually is, it can be a
problem.
2) speed. When you need floating point calculations, the speed issue can be
painful.
> Also, has anybody written, or heard of some generic client/server type
> libraries to work between the PC and the RCX, so that (and I know it defeats
> the purpose of this stuff) the PC can run the software, and just tell the
> RCX basic commands, like left or right or forward, relay sensor reports
> back, and make the PC decide what to do. Can we overcome
> processor/memory/permanent storage problems like this?
The newest legOS should have a very advanced method for this, but you can
also use VB and the standard firmware for something like this. However,
I'm not an expert on this (since I can't run VB.) Others can certainly
fill you in here. Remember, this is inherently kind of limited because of
the range limitations on the IR. Otherwise, there should be few or no
limitations.
> Also (but kinda related) what about addressing indiviual units? I saw
> someone mentioned something about datagrams (but that email is sitting on
> another computer, sorry to whoever sent it). I was thinking of a project
> (actually I've been thinking about many more than this one) whereby I have 2
> RCX units for the one robot - one for the torso, one for leg type things.
> This shouldn't be a serious problem, but expanding upon that idea, why not
> have heaps of RCX units (cost is certainly an issue here) in the one robot.
> You could hook up mirrors internally, essentially acting as a hub - and
> program each unit so that it has an address (or ID if you will), and program
> each RCX unit to accept only those commands with a given ID at the front -
> just like IP.
Many, many discussions similar to this, but no implementations. I have a
flock of units at my disposal (check out the unfortunately large
http://arthurdent.dorm.duke.edu/legos/archives/QL/12.jpg ) and a resident
genius wrote a UDP based protocol for it, but he left for Princeton and I
haven't yet had the time to implement it. But it should be doable.
> Again - is this really feasible?!
It all should be :)
> Sorry to annoy you all with my wide-eyed ideas - but I'm what you might like
> to call a newbie to Robotic Lego.
Ask away- this list occasionally needs a newbie to give it a shot in the arm.
-Luis
#######################################################################
Profanity is the one language that all programmers understand.
-Anonymous
#######################################################################
|
|
Message has 1 Reply:
Message is in Reply To:
 | | AI and even more exiciting stuff
|
| ----- Original Message ----- From: Luis Villa <liv@duke.edu> To: ts <ts@zylotech.com.au> Cc: <lugnet.robotics@lugnet.com> Sent: Monday, 18 October 1999 13:09 Subject: Re: AI (...) I am using NQC at the moment. I have just found, and I am downloading (...) (25 years ago, 18-Oct-99, to lugnet.robotics)
|
16 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
|
|
|
|
