Subject:
|
Re: Mindstorms/reinforcement learning
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Tue, 12 Oct 1999 21:22:40 GMT
|
Original-From:
|
alex wetmore <alex@phred.org=antispam=>
|
Viewed:
|
573 times
|
| |
|
|
I don't think that NQC by itself will handle your needs. I am not an AI
expert (pretty much skipped all AI classes in college), but I expect that
would need a decent amount of memory to remember which behaviors have been
reinforced. NQC works on top of the standard Lego firmware, which limits
one to 32 variables total, with no arrays.
You might consider using LegOS or PBForth, both of which would allow you to
use up to the 32k of memory found in the machine. You could also transfer
data back from the RCX to a host computer, which would perform the
calculations and then send information back to the RCX. If you wanted to do
this all onboard a robot you could use a small computer (such as a Windows
CE machine or Palm Pilot) as the host computer, hooked up directly to the
lego IR tower, which woudl point at the RCX brick.
alex
----- Original Message -----
From: "Erik Zempel" <ezempel@engin.umich.edu>
To: <lego-robotics@crynwr.com>
Sent: Tuesday, October 12, 1999 12:41 PM
Subject: Mindstorms/reinforcement learning
> Hello, allow me introduce myself to the list. My name is Erik Zempel, I'm
> a Junior at the University of Michigan in the Computer Engineering
> department. Two friends and I are hoping to to get some funding for a
> research project involving reinforcement learning techniques for AI
> robots. For this project, we will be using Lego Mindstorms for our
> robots. We chose Mindstorms because the were an inexpensive, quick way of
> building a variety of different robots. For the AI, we are hoping to use
> not-quite-C for the RCX programming.
>
> Currently, we are drafting our proposal. What I want to know is if there
> has been anywork done like this before, and what are the RCX unit's
> capabilites? (i.e. - would it not be a good idea to do this with an RCX)
>
> Any input would be greatly appreciated!
> Thanks!
> Erik
>
> -----BEGIN GEEK CODE BLOCK-----
> Version: 3.1
> GCS/GE d-- s-:- a-- C++++$ UL+ UHPUX+ US+ P+++ L+>+++ E W+++ N++ o? K? w@
> O- M+ V PS+++ PE-- Y PGP-- t(-) 5- X-- R(+) tv+ b+ DI++ D+ G e>++(*) h(+*)
> r++ y?
> ------END GEEK CODE BLOCK------
>
>
>
|
|
Message is in Reply To:
 | | Mindstorms/reinforcement learning
|
| Hello, allow me introduce myself to the list. My name is Erik Zempel, I'm a Junior at the University of Michigan in the Computer Engineering department. Two friends and I are hoping to to get some funding for a research project involving (...) (25 years ago, 12-Oct-99, to lugnet.robotics)
|
9 Messages in This Thread:
  
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
