The "accumulation of error" problem will eat you alive.  People have
been trying to solve that for decades, maybe centuries.  If you could
add a few north/south and east/west buried wires, then occasionally the
mower would cross one and could correct its position.
Wheel turns are really bad for measuring position.  How about using an
optical mouse looking through an (un)magnifying glass at the ground.
That would measure distance better than wheel turns.

Sherpa Doug

> -----Original Message-----
> From: Madhu Annapragada [mailto:madhu@CONECTIV.NET]
> Sent: Friday, August 31, 2001 9:52 AM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: [EE]: - Botboard
>
>
> I have been toying with this idea for a while and I will
> throw it out there
> to see if it even holds up to the piclist scrutiny. What do
> you think about
> teaching the bot the first time? With heading information
> from a compass and
> wheel rotation information from encoders, the bot would build
> a table of
> heading changes vs wheel rotations the first time (starting
> from a home
> position). One would use a joystick to mow the lawn the first
> time. The bot
> records the wheel position every time the heading changes by a certain
> amount. The second time around, the bot would head out in the
> first heading
> direction until it reaches the wheel rotation figure for the
> next heading
> change, change to the new heading and so on.
> Of course, the problems I see right away are wheel slippage
> and  noise from
> the compass. But for arguments sake, if I could solve those
> problems, do you
> see any other potential pitfalls for this method. The merits
> of this method
> are that you do not need to depend on any external markers.
> Teach it one
> lawn ( even multiple lawns using a selector switch) and let
> it go from a
> know position every time. One might have to re-teach it every
> so often if
> the bot starts wandering into the flower patch....
> Anyway, does the method seem feasible?...winter is coming and
> I need to go
> into hibernation in my basement lab...
> Madhu
>
>
> -----Original Message-----
> From: pic microcontroller discussion list
> [mailto:PICLIST@MITVMA.MIT.EDU]On Behalf Of Jinx
> Sent: Friday, August 31, 2001 8:26 AM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: [EE]: - Botboard
>
>
> > How are you planning the cutting path ?
>
> The most successful I've seen is tracking a buried wire. Not
> sure what the signal is but the demo looked pretty good. If
> you edged the area with white (painted bricks ? logs ?) you
> could perhaps use optics. Industrial robots follow a white line
> around their work area
>
> --
> http://www.piclist.com hint: The list server can filter out subtopics
> (like ads or off topics) for you. See http://www.piclist.com/#topics
>
> --
> http://www.piclist.com hint: The list server can filter out subtopics
> (like ads or off topics) for you. See http://www.piclist.com/#topics
>
>
>

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics