Do you know of any research about the accuracy of forecasts as a function of sensor data input? I.e. what kind of an improvement will we see if some number of Android users install PressureNet?
tail -n 1 $f
Tail seeks backwards so it will only read the one line. Of course, this won't give you a line count. - constant horizontal velocity
- gravity
- flapping implemented via impulse
- collisions are handled with bounding boxes
Maybe someone who knows something about optimal control (ODE) say whether this is analytically solvable? Of course there's still the practical stuff (numerical integration, I/O lag) to deal with but I'm optimistic.