> I was then trying to say that for a simple temperature control
> scheme like that presented you could probably get some good
> results from using lookup tables. Maths could be avoided in favour
> of experimentation. Although I suspect the best control curve will
> look like a half sine wave, the peak being the normalized desired
> temperature setting.

see motorola 68HC12 docs - "fuzzy logic support"
http://ebus.mot-sps.com/mcu/documentation/pdf/cpu12rmr1.pdf

they discuss (quickly) fuzzy logic, and their support using trapezoidal
control curves (using lookup tables).