Indoor Localization of Flying Robot by Means of Infrared Sensors

Abstract

Autonomous navigation of flying robots in GPSdenied environments such as indoors requires that the flying robot be able to recognize the environment using external sensors. Laser scanners and computer vision are mainly used for indoor mapping and localization in studies on indoor flight. However, such systems require higher payload capacity and processing power for the sensors. In this study, we develop a lightweight flying robot for achieving indoor autonomous flight using four infrared (IR) sensors. As the first stage of this study, we present a localization technique that involves the use of a particle filter. Two problems exist in our system. First, it is difficult to use IR sensors close to a wall, because doing so would yield faulty results when calculating distance using the sensor output voltage. To resolve this problem, we developed a probabilistic output voltage observation model. The particle filter estimates position from voltage information using this model without the use of calculated distance. The second problem is that the spatial resolution is low because only four IR sensors are used. This problem was solved by rotating the robot horizontally at all times to acquire information from various directions. The localization performance was verified experimentally using an electric turntable and a cart. In the first and second experiments, we confirmed that localization is successful even when the robot is in motion and even when the robot is flying near a wall.