Thermal-Inertial Localization for Autonomous Navigation of Aerial Robots through Obscurants

Abstract

In this paper the problem of autonomous navigation of aerial robots through obscurants is considered. As visible spectrum cameras and most LiDAR technologies provide degraded data in such conditions, the problem of localization is approached through the fusion of thermal camera data with inertial sensor cues. In particular, a long-wave infrared camera is employed and combined with an inertial measurement unit. The sensor intrinsic and extrinsic parameters are appropriately calibrated, and an Extended Kalman Filter framework that uses direct photometric feedback is employed in order to achieve robust odometry estimation. This framework is capable of accomplishing real-time localization while navigating though obscurants in GPS-denied conditions. Subsequently, an experimental study of autonomous aerial robotic navigation within a smoke-filled machine-shop environment was conducted. The presented results demonstrate the ability of the proposed solution to ensure reliable navigation in such extreme visually- degraded conditions.