The competition is a race against the clock. you have to explore the
maze, find the optimum route to the target cells and then run from start
to target as fast as you can. these phases present different challenges
to the mouse.
During exploration you will want to be relatively careful, you can not
afford to lose your location at any time. Extra care may allow you to
detect walls in cells you have not yet visited. Mouse speeds are likely
to be lower and additional wall checking activity will help ensure that
you don’t miss wall data because of variations in the reflectivity of
the walls. Additional, real-time calibration of the sensors may be possible.
After you have found the centre, you have to decide whether to carry
on exploring in the hope of finding a better route or just run with the
one you have. A cunning maze design may penalise you for not exploring
further.
Now you have to perform a speed run. Assuming the rules reward you for
your fastest run, many mice will perform several runs. A fast phase from
start to centre should be followed by a relatively safe run back to the
start. You don’t want to crash on the way back as there are no points
for a fast return to the start. After that, each run can get faster until
you crash. Any mouse is likely to have a limitation, either in terms of
its speed or its abilty to process sensor data on the go.
I suppose you would want to specify you sensor and motor control code
so that you can reliably assert control at any sensible speed. Ensure
that the mouse is limited by its mchanical properties rather than its
computing power. Sensible choice of processor and careful coding should
make this possible.
How fast do you have to be?
At the 17th All Japan Contest in November 1996, the maze had a route
to the centre that was about 75 cells long with plenty of straights. To
be competitive, you needed to be able to run this in less than 15 seconds.
That is an average speed of just under 1m/s. What with turns and acceleration
curves, you will probably want a top speed of about 1.5m/s or better.
That is 1.5mm per millisecond, or 120ms per cell. Within each cell, you
will want to be performing heading corrections frequently enough to stay
on the straight and narrow. This should be within the capabilities of
any modern processor.
The 2009 contest in Japan had a run that was about 65 cells long and you would have needed to run that in under 5 seconds to stand a chance of winning.