heh hopbeat, cool
the accelerometer part is simple (as we know)
it needs to be able to load in a bitmap and do collision detection/avoidance
sheep must "prefer" to be in the low valleys and will need real tip of the device to get up the hills

ive been doing some testing using this and trying to get the "sheep" to stay in the low (dark) areas and to follow along.

the algo is a bit tricky but my particle subsystem works well (liqflow with just ~10 bigger particles instead of hundreds)