Subject:
|
RE: Line Followers
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Tue, 16 Nov 1999 21:58:31 GMT
|
Original-From:
|
Jim Thomas <jim.thomas@NOSPAMtrw.com>
|
Viewed:
|
706 times
|
| |
|
|
What if the light sensor were mounted on a boom arm (or a sliding track)
which was attached by gears to a motor and a rotation sensor. Then what you
would do is create a fast response control loop using the right on black,
left on white rule. The idea is to always keep the sensor on the edge of
the line and know the robot's offset from the line based on the sensor
position. Now you do whatever line following algorithm you want using the
sensor location input instead of the line detection. I figure that the
sensor loop ought to be about 5 to 10 times faster than the robot loop (or
the response of the robot loop can just be damped or use thresholds). The
total available travel of the sensor would have to be sized according the
slow loop rate and the speed of the robot. This should let your robot run
without wiggling around too much. What do you think, good idea?
JT
> -----Original Message-----
> From: Steve Hassenplug [mailto:SteveH@mailcode.com]
> Sent: Tuesday, November 16, 1999 11:54 AM
> To: lego-robotics@crynwr.com
> Subject: RE: Line Followers
>
>
> I've been working on a line following bot. My original idea
> was to have an
> arm that swings back and forth (to cover about 3 inches) over
> a line and
> would be able to adjust the direction of the bot based on the
> position of
> the line within the arc.
>
> The first problem is that running a rotation sensor directly
> off the arm
> doesn't give me much in the line of feedback (only three of
> four counts)
> and I couldn't get touch sensors to work very reliably,
> either. So it was
> hard to track the position of the arm.
>
> The next issue is that there is a problem if, for example,
> the arm swings to
> the left at the same time the line swings to the right. The
> bot could loose
> the line altogether.
>
> So, I stepped back and made a very simple edge following
> robot. It turns
> right when it sees black, and left when it sees white. It
> still seems to
> wiggle too much, but I'll work on that. It's really my first
> line following
> bot. It made it around the [old] test track in less than
> five seconds, but
> I'm not sure how that compares to other bots.
>
> Does anyone have suggestions on how to build a track everyone
> could use?
> Some way for us all to compare against each other?
>
> My first thought is to make up a 'track' that consists of
> several pieces of
> paper that can be taped together to form a single track, and
> then post these
> pages on the web. Any other ideas?
>
> Steve
>
> -----Original Message-----
> From: Paul Speed [mailto:pspeed@augustschell.com]
> Sent: Tuesday, November 16, 1999 1:27 PM
> To: lego-robotics@crynwr.com
> Subject: Re: Line Followers
>
> Luis Villa wrote:
> >
> > > My idea is to simulate having multiple light sensors by
> > > using one light sensor to sample in several places. • Imagine a light
> > > sensor moving around in a circle sampling discrete points • along this
> > > circle. Standard methods have been discussed for various forms of
> > > rotation sensors but I think a "touch sensor pressed by a gear"
> > > type has some elegance when considering that the touch • sensor could
> > > probably be stacked on the same port with the light sensor.
> >
> > Wow. This sounds like a mechanical mess, but an elegant • solution (in a
> > broader sense) to the depth/breadth issue I was talking • about last night.
> > Good luck.
>
> re: Mechanical mess - agreed. :) Hopefully, and elegant
> solution will present itself there too.
>
> >
> > My two cents:
> > 1) maybe a half or 3/4 circle would be all that is needed? No
> > twisting problems then.
>
> True, but I sort of like the idea behind having a sort of
> blip at four compass points. It makes maze navigation a simple
> problem and then I can focus on maze traversal.
>
> > 2) When you rotate the bot, any idea on how you will realign the
> > sensor to the bot's new direction?
>
> That was the idea behind the synchro platform. The bot will
> always be "facing" the same way which is one of the things that
> necessitates a full circle sampling. Initially I will focus on
> movement restricted to 90 degrees. Eventually it would be nice to
> develop a more general solution that can handle more than four
> directions.
>
> -Paul (pspeed@progeeks.com, http://www.progeeks.com/)
>
> >
> > Just some thoughts-
> > Luis
> >
> > > I will leave solutions to the problem of keeping the light
> > > sensor cable from getting twisted as an exercize for the reader.
> > > (heh) At least until I've built a working example. • Speculation is
> > > always so much easier than proving that it will work. :)
> > >
> > > -Paul (pspeed@progeeks.com, http://www.progeeks.com/)
> > >
> > > Luis Villa wrote:
> > > >
> > > > With just one light sensor, the problem is that you • have to guess
> > > > which direction you just fell off the line. i.e., your sensor
> > > > suddently goes from dark to light. Is the dark to your • left? Your
> > > > right? There's no real way of telling, is there?
> > > >
> > > > If you guess well, you can follow the line well. If you • don't guess
> > > > well, you are in trouble. Frankly, I have no strategies for
> > > > guessing well. However, "go x left, 2x right, 3x left, • 4x right" is
> > > > not unreasonable, and I believe some variation on that • (perhaps x
> > > > left, 2x right, 4x left, 8x right, but I'm not sure) is
> > > > mathematically provable to be the best solution for this type of
> > > > problem. Note: these are not actually go left, but spin • left, spint
> > > > right, etc. with the assumption that you have stopped • when you "lost"
> > > > the line.
> > > >
> > > > With two light sensors, this suddenly becomes much • easier, and with
> > > > three, you can build a great light sensor. However, forks still
> > > > aren't dealt with well- you need depth as well as • breadth to be able
> > > > to analyze those robustly and on the fly (in my opinion). Anyone
> > > > have any thoughts on that one?
> > > > -Luis
> > > >
> > > > On Tue, 16 Nov 1999, David Morgan wrote:
> > > >
> > > > > Date: Tue, 16 Nov 1999 00:08:59 GMT
> > > > > From: David Morgan <lego-robotics@crynwr.com>
> > > > > To: lugnet.robotics@lugnet.com
> > > > > Subject: Line Followers
> > > > >
> > > > > Hi,
> > > > >
> > > > > Could someone please give me a little bit of advice
> on building a
> line
> > > > > follower with the standard lego light sensor? I will • be doing the
> > > > > programming for it with LegOS.
> > > > >
> > > > > Can you achieve decent performance with just one light sensor?
> > > > >
> > > > > Thanks!!
> > > > > Dave Morgan
>
>
>
|
|
1 Message in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
