Better navigation through Lookahead

Dr. Camille Goudeseune

Integrated Systems Lab (ISL)

Beckman Institute, U. Illinois

In any task of navigation or manipulation, beginners overcontrol. They notice an error in speed or direction, recover from it, but recover too much. Only with experience can they predict the *future* effect of a command given at this instant, and thereby dampen the oscillation instead of letting it diverge. This is true for any vehicle, real or simulated, military or videogame. Nobody spends thousands of hours navigating in one virtual environment, so everyone's a beginner. Therefore, in VR, lookahead is particularly important. I invite people to augment a navigation task of their choice with such lookahead: a visual representation of where the vehicle will be in a few seconds, if current trends continue. The Navy accelerated the training of submarine operators like this fifty years ago with vacuum tubes, so it shouldn't be too hard with modern tools.