Subject:
|
RE: Antenna Bumpers/Insects
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Fri, 4 Dec 1998 18:03:54 GMT
|
Original-From:
|
Vaan, Howard <howard.vaan@csfp.co%NoMoreSpam%.uk>
|
Viewed:
|
2746 times
|
| |
|
|
You could train the neural net within range of the IR.
After that, you'd have the node values set and you might be able to fit that
could fit that into the RCX for autonomous behaviour - I'm unsure about the
usefulness of this with three input and three output nodes.
What would we be trying to teach the bug?
> -----Original Message-----
> From: lugnet.robotics@lugnet.com [SMTP:lugnet.robotics@lugnet.com]
> Sent: Friday, December 04, 1998 5:47 PM
> To: lego-robotics@crynwr.com
> Subject: Re: Antenna Bumpers/Insects
>
> Howard,
>
> Since you are into mimicing the behaviour of an insect in some way, and I
> know that you are messing around with the Smalltalk Bot-Kit, you might
> like
> to consider the following for an idea.
>
> One of the interesting things about Bot-Kit is it allows you to drive your
> robots directly from programs written on the PC. This might allow you to
> attach more complex logic (or whatever) to the robots decision-making.
> One
> thing I had in mind was to mimic the behaviour of a cockroach. I believe
> that you can simulate the general behaviour of a cockroach using a neural
> net with approximately 100 neurons. I haven't got a link for this but I'm
> sure that at some time in the past I found a simulation on the Net to do
> this. The computer simulated cockroach would buzz around on the screen
> behaving, to all intents and purposes, exactly as a real cockroach would.
> Now I think this would be interesting; to use a neural network to control
> some sort of physical cockroach built from LEGO. You'd have to build some
> neural net stuff using Smalltalk and then to import (or train) the
> cockroach
> simulation into it. You could then use Bot-Kit interface to give your
> robot
> bug a "real brain".
>
> As far as I can see, the main problem with this idea would be keeping the
> RCX in contact with PC via the IR link at all times. I'm not entirely
> sure
> how this could be done (mirrors?) but some guy in this group did mention
> the
> possibility of some gizmo from Radio Shack being able to extend the IR
> range
> using a radio link extender. I somehow doubt this would give two-way
> communication, though. Of course, using a Cybermaster then would be no
> problem since this uses a radio link anyway.
>
> Just food for thought.
>
> Regards,
>
> Andy Bower
>
> Vaan, Howard wrote in message ...
> > If they really do have such funky antennae they have to have the brain to • go
> > with them. I thought you could accomplish insect behaviour without doing
> > the kind of representational mapping that your answer implies (i.e. size
> > inference).
> >
> > I think this may be in danger of straying away from pure RCX-iness - but
> > biological modelling's a good way to go.
> >
> > > -----Original Message-----
> > > From: Laurentino Martins [SMTP:lmartins@marktest.pt]
> > > Sent: Friday, December 04, 1998 1:11 PM
> > > To: Lego-Robotics
> > > Subject: Re: Antenna Bumpers/Insects
> > >
> > >
> > > I think insects do much better than that.
> > > Their antennas are VERY touch-sensitive and they know which point of • the
> > > antenna was touched.
> > > Better, the also move the antennas to "feel" the object and it's • current
> > > position, size and maybe texture.
> > > I think we are in no position to mimic their behavior with a
> > > RCX/CyberMaster... but it's a great challenge anyway! :-)
> > >
> > > Laurentino Martins
> > >
> > > At 12:57 04-12-1998 Friday , you wrote:
> > > > I've just made a simple bot with the antenna-like config in the
> > > > Constructopedia + implemented the simple touch>stop motor algorithm.
> > > >
> > > > One problem the robot has is with table/chair legs and the like - the • bot
> > > > wanders straight into it and then can't escape because the antennae • make
> > > it
> > > > continually steer back into it.
> > > >
> > > > OK - I know plenty of ways to fix this, from a HW or SW point of view
> .
> .
> > > .
> > > > that's not the question.
> > > >
> > > > How do real insects with long antenna manage this problem? Can they • walk
> > > > backwards? I don't think so . . .
> > > >
> > > > (We don't have many large roaches here in the UK that I can observe)
> > > >
> > > > Cheers,
> > > >
> > > > Howard
> >
|
|
1 Message in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
Active threads in Robotics
|
|
|
|
|
|
