Abstract

The problem of active human detection with a mobile robot equipped with an RGB-D camera is considered in this work. Traditional human detection algorithms for indoor mobile robots face several challenges, including occlusions due to cluttered dynamic environments, changing backgrounds, and large variety of human movements. Active human detection aims to improve classic detection systems by actively selecting new and potentially better observation points of the person. In this preliminary work, we present a system that actively guides a mobile robot towards high-confidence human detections, including initial simulation tests that highlight pros and cons of the proposed approach.