Driver familiarity modeling for generating navigation directions

摘要: Current and future in-vehicle personal assistant systems stand to benefit from knowledge of the level familiarity that driver has with different parts current route. Such could use this information adapt driving directions be more succinct, understandable personalized. However accumulating evidence by direct experience alone take many months still incomplete. Instead, we propose building predictive models road networks generalizing intelligently a small sample data. Inspired psychology studies suggest variety cognitive route in humans, present an ensemble using machine learning methods on GPS time series data collected during normal volunteer pool 22 drivers San Francisco Bay area. We validate models' predictions through extensive questionnaire administered subjects about need for along select routes. Our results indicate significant component can predicted unobtrusively conclude proposed approach integrating such generator directions.