Unscripted Retargeting: Reach Prediction for Haptic Retargeting in Virtual Reality
Aldrich Clarence, Jarrod Knibbe, Maxime Cordeil, Michael Wybrow
Abstract
Research is exploring novel ways of adding haptics to VR. One popular technique is haptic retargeting, where real and virtual hands are decoupled to enable the reuse of physical props. However, this technique requires the system to know the users' intended interaction target, or requires additional hardware for prediction. We explore software-based reach prediction as a means of facilitating responsive, unscripted retargeting. We trained a Long Short-Term Memory network on users' reach trajectories to predict intended targets. We achieved an accuracy of 81.1 % at approximately 65% of movement. This could enable haptic retargeting during the last 35% of movement. We discuss the implications for possible physical proxy locations.