Reduction of User Interaction by Autonomy

Abstract

This paper describes experiments that quantify the improvement that autonomous behaviors enable in the amount of user interaction required to navigate a robot in urban environments. Many papers have discussed various ways to measure the absolute level of autonomy of a system; we measured the relative improvement of autonomous behaviors over teleoperation across multiple traverses of the same course. We performed four runs each on an “easy” course and a “hard ” course, where half the runs were teleoperated and half used more autonomous behaviors. Statistics show 40-70% reductions in the amount of time the user interacts with the control station; however, with the behaviors tested, user attention re-mained on the control station even when he was not interacting. Reducing the need for attention will require better obstacle detec-tion and avoidance and better absolute position estimation.