Quite a few other sensing schemes have been used.
A simple variation on the top-down sensor might be to have a row of sensors
hanging over the walls and a light source shining up at them from below.
The number of occluded sensors tells you your position. This will, of
course, not allow you to detect walls in the next cell.
Mechanical devices have included wings that sit on the walls and rotate
according to distance. These are coupled to the processor and are robust,
effective and relatively simple. They are also clumsy and prone to collision
problems. An elegant solution though and not to be underestimated. Nick Smith made the first mouse to find the centre of a maze in the UK and that used metal wings.
Laser diodes and position sensitive detectors have been experimented
with although there do not seem to be many mice using them.
A line-scan camera could be constructed. Either a scanning laser spot
or a scanning photosensor could be used to look for the high contrast
between wall and floor. While mechanically relatively complex, the signal
processing should be well within the capabilities of modern microcontrollers.
I recently took apart a hand scanner and found an ideal looking linear optical array with only ten or so connections that should be easy enough to use. So far, I can’t find any data for it though.
A number of attempts have been made at full vision systems. There does
not seem to have been a really successful solution – presumably due to
the relatively high processing load such systems generate. I did read
of a mouse that was supposed to raise a camera on a stalk an examine the
maze from the start. I have no idea what happened to that idea but the
possibilities are particularly intriguing. Imagine a couple of seconds
to raise the camera, allow a (hopeful) five seconds to process the image,
another couple to lower the mast and you are off. Home and dry in, perhaps,
twenty seconds. It will be some time before the micromouse competition
fails to find ways to challenge builders.
The GameBoy camera seems to have potential. It is small, has appropriate optics and can do some form of edge detection inside the device.
Sharp make a position sensitive detector, the GP2D12, which produces
an Alan voltage proportional to distance. This is intended for use over
a distance of 10 to 80 cm and would make an excellent detector of walls
up to 4 cells ahead of the mouse as well as looking down side passages
on the way past openings. I recently saw a calibration curve for one of
these devices which noted that users should be careful not to confuse
distances less than 10cm with longer distances. The reason is that output
falls off below 10cm. It may be possible to use these devices for close-in
wall measuring with some care. The main disadvantage appears to be the
measurement rate of only a couple of hundred samples per second. This
might not be much use in steering a mouse at 2m/s – you may only get a
sample every centimetre or so. Worth experimenting with however.
[NOTE: a new mouse from Derek Hall uses this type of sensor with great results]
Ultrasonic devices can be power hungry and difficult to use over the
short distances in the maze. At 40kHz, the wavelength of the sound is
going to be about 7.5mm. This is going to be too small for phase measurements
and, with time of flight around 160us, tricky to measure. There may be some clever signal processing hardware out there that can do it though
UPDATE: Those clever folk at the Technology Innovation Centre in Birmingham have a micromouse which uses ultrasonic sensors to great effect. It turns out that the time of flight is very measurable with a micro. I hope to have details about their mouse in the Other Mice pages soon.