Using Proxemics as a Corrective Feedback Signal during Robot Navigation

Abstract

This paper presents a novel feedback system based on proxemics. PFbS (Proximity Feedback System) allows a robot to dynamically update its navigational behavior by interpreting a human teacher's distance to the robot as a corrective feedback signal. We demonstrate PFbS on a Pepper robot and evaluate it in a pilot study (N=18) against a joystick interface baseline, in terms of usability, intuitiveness and signal clarity. We also evaluate the learning algorithm for dynamic update of the robot's trajectory. Based on the insights from the pilot, we believe that with proper sensor accuracy, PFbS can be improved to allow for more intuitive and embodied feedback for non-expert users.