Originally Posted by
Phil_gretz
Back in the 1980s, I worked on neural networks that "learned" based on pretty simple rules. I would like to see a system that "rides along" with me for a few weeks. Once it has learned my behavior based on heart rate, power, incline, shift patterns, and range of acceptable rpms, I'd want it to embody my decision processes and take over the controls. Subject to manual override, of course, like cruise control on a car.
Anyway, I wouldn't think this would be too difficult...
True. And it needn't be set up custom to be perfect for each person, but to be acceptable to the majority - just like auto transmissions in cars... my Grand Caravan, even when my foot was to the floor, would shift 2500 rpm before redline, but I would have preferred it to wait... it wasn't perfect for me but it was acceptable. And like high-end modern automobiles there can be user selectable 'modes' - my great grandmother probably doesn't want her legs spinning up to 200 rpm before her bike shifts, but some people might. I could ride an automatic shifting bike, but I wouldn't like it if t kept my cadence at 40 rpm. The user could select an 'optimal cadence' that the system shifts to maintain.