Subject:
|
RE: Predator/Prey thoughts ...
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Thu, 3 Dec 1998 21:45:02 GMT
|
Original-From:
|
Eric Eilebrecht <ericeil@microsoft!spamcake!.com>
|
Viewed:
|
2400 times
|
| |
|
|
I think the biggest problem with having prey which are not also preditors is
that the prey, by definition, can't win. It can only hold out for a while.
Sure, it's interesting to write the code - but how long do you let it run
before you get tired of it? With a preditor, you know exactly how and when
you have won the game.
Eric
-----Original Message-----
From: Peter Wieland (Exchange) [mailto:peterwie@exchange.microsoft.com]
Sent: Thursday, December 03, 1998 1:16 PM
To: 'tim@ams.co.nz'; lego-robotics@crynwr.com
Subject: RE: Predator/Prey thoughts ...
i think i disagree that sumo wrestling (particularly in this context) is
about who weighs the most and has the most power. It's about how you
arrange the limited set of sensors you've got and how you build your bot. A
light bot with lots of speed could get away from a slow & powerful one
easily but might run over the edge of the circle quickly. A powerful one
without much speed is probably not as well defended against being pushed
from the side. If you've got a blind spot in your sensors or can't tell a
left-hand hit from a right-handed one then you may be in trouble.
but i also don't particularly like the level footing of rock-paper-scissors.
I think the idea of trying to build prey which act like prey is kind of
cool. Prey do silly things like find dark spots or light spots or run in
circles or follow along walls until something "out of the ordinary" happens
in which case they freeze or run like hell. Predators use the same sorts of
criteria (finding dark or light or walls) but either can find the prey
despite their lack of movment, or the flush them out and attack.
has anyone written a neural net program for the RCX yet?
-p
-----Original Message-----
From: Tim McSweeney [mailto:tim@ams.co.nz]
Sent: Thursday, December 03, 1998 12:53 PM
To: lego-robotics@crynwr.com
Subject: RE: Predator/Prey thoughts ...
aha, but that's the beauty of having a human judge as the decider, they can
tell what a bot hit. If it hit a wall then from the bots point of view
nothing changes, the bot backs up and heads off in another direction. If
the bot hits another bot then one of them is going to get picked up and
moved. Probably in either case the bot should back up and take stock of
it's situation, either it hit a wall in which case whatever it was chasing
is probably hiding behind it so it needs to do some navigation, or it hit a
bot in which case the bot got moved and so the hunter should try and
"re-acquire" a new target. Of course similar arguments apply for bots that
are currently avoiding the hunters.
To make things even simpler The "dead" prey bot could be cycled off and on
again to reset it.
The reason that I find paper-Rock-Scissors so interesting is that all of the
bots are on an equal footing, all of them have to find a balance between
hunting and hiding and do both at the same time. It's a lot more
"intellectual" than Sumo Wrestling which basically comes down to who ways
the most, and who has the lowest gear ratio.
There are some really interesting strategies that can arise.
The timid paper sits quietly with it's transmitter of and when it sees a
lone rock wandering round it charges at it screaming
"paper-paper-paper-paper-paper"
The defensive scissors runs around avoiding rocks like the plague, and with
any luck might find the odd paper.
The aggressive rock charges after scissors and completely ignores paper and
just tries to mush scissors, occasionally it gets caught by a paper and done
in.
Which strategy works the best is undefined until the end of a particular
game. it depends on what your opponents do. Of course there is always the
possibility that the bots will modify their behaviour based on what the
others are doing and this is where it gets really interesting. Suddenly how
you program your bot becomes very important.
Cheers
Tim
> -----Original Message-----
> From: news@lugnet.com [mailto:news@lugnet.com]On Behalf Of Matthew
> Miller
> Sent: Friday, December 04, 1998 9:26 AM
> To: lego-robotics@crynwr.com
> Subject: Re: Predator/Prey thoughts ...
>
>
> Tim McSweeney <lego-robotics@crynwr.com> wrote:
> > "Touch 'em and they die" rule appeals, the judgescan do all • the hard work
> > and the bots only have to run in a straight line towards the • the opponent
> > (or away from) and bounce off the walls. Note: there's no • reason why the
> > "arena" can't be arbitrarily complex, say like someones
> living room floor.
>
> Well, yes there is. If there's objects besides robots which
> can be hit,
> robots need some way of telling the difference between "I hit
> you!" and "I
> hit a wall!".
>
> --
> Matthew Miller --->
> mattdm@mattdm.org
> Quotes 'R' Us --->
http://quotes-r-us.org/
|
|
Message has 1 Reply:
 | | Re: Predator/Prey thoughts ...
|
| Well, In real life, if a preditor doesn't eat then it starves. So why not in robot life too? Maybe it can even get more desperate as it gets closer to death. The prey then wins when all of the preditors are dead. -Paul (...) (26 years ago, 3-Dec-98, to lugnet.robotics)
|
2 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
