Predictive display for mobile manipulators in unknown environments using online vision-based monocular modeling and localization.