Efficient probabilistic localization for autonomous indoor airships using sonar, air flow, and IMU sensors

Abstract

In recent years, autonomous miniature airships have gained increased interest in the robotics community. Whereas their advantage lies in their abilities to move safely and to hover for extended periods of time, they at the same time are challenging as their payload is strictly limited and as their complex second-order kinematics makes the prediction of their pose and velocity through difficult and imprecise physical simulation. In this paper, we consider the problem of particle-filter-based online localization for a miniature blimp with lightweight ultrasound and air flow sensors as well as an inertial measurement unit (IMU). We present probabilistic models dedicated to the special characteristics of the miniature and lightweight sensors applied on our blimp. Furthermore, we introduce an efficient odometry motion model based on the measurements of air flow sensors and an IMU which is less computationally demanding compared to the standard physical simulation-based control motion model. In experiments with a real blimp in a complex indoor environment, our approach has proven to allow accurate and reliable online localization of a miniature blimp and requires an order of magnitude fewer particles compared to the localization based on the standard control motion model. Furthermore, we demonstrate the substantial improvements in terms of localization accuracy when taking into account the temporal correlation of the air flow measurements in our novel odometry motion model.