Hello,
Has anyone tried to execute a controlled
turn?
I would imagine that one would record an
optimum path sensor profile, and have a controller tune accordingly to that
path for that turn.
Or is the open loop control described
below is the most popular…. (if its going fast enough for a power slide,
it seems pretty darn good)
Thanks,
fred
From:
owner-micromouse@cs.rhul.ac.uk [mailto:owner-micromouse@cs.rhul.ac.uk] On Behalf Of Bill Marshall
Sent: Thursday, November 18, 2004
2:08 AM
To: Micromouse
Subject: Re: Minimal mice
With all this discussion of minimal mice how about this
idea? Forgive me if it's been discussed before, but how about mouse circuit
racing as an intermediate complexity event? I suspect many are put off maze
solving because:
1. You need access to a maze for development.
2. The task is very complicated if anything other than a simple mechanical wall
follower is built.
3. The average micromouse competition is not very exciting to the younger
onlookers who see many robots failing to make much progress.
A rectangular walled circuit 'maze' would be much easier to
construct and could be any size.
The racing mice would need essentially the same hardware as a maze solver
(microcontroller plus side looking wall sensors), but the software would be
greatly simplified. Normally you need the low-level control and navigation
routines with a maze solver on top. The racer still needs the control but no
solver and easy navigation.
Finally, mice screaming round a race track is far more exciting to onlookers.
I built such a race track to test out my PID control algorithms. The PID
control was turned off at the corners where the turn was executed by counting a
fixed number of 'clicks' from the motor tachos, then the PID was turned on for
the next straight. There was no attempt to slow down into the corners and the
result was really great to watch with 'power-slides' round the corners and the
control system visibly fighting to get the mouse back on track on the
straights. This crude non-optimized system still managed 20 laps or more before
some slight variation caused a crash into the side wall. This was like RC cars
to look at, but without the radio link.
It would be a timed competition perhaps requiring 10 laps in one direction,
followed by 10 laps in the other.
The real benefit is that a successful racer only needs the maze solver to be
added, plus some other mods to the software to work in a conventional maze. It
means that someone starting out can get something working and winning at a
lower level. The experience gained and the hardware then form the platform for
the more advanced task.