Search

Global Navigation Satellite Systems (GNSSs), such as the Global Positioning System (GPS) and Global Navigation Satellite System (GLONASS), are the primary sensors for localization of most unmanned helicopters. However, GNSS signals cannot be tracked reliably due to geographic restrictions or deliberate jamming, and this has been an unsolved problem for a long time to localize the position of helicopters robustly and accurately when the GNSS signals are lost. A new algorithm is presented for this unsolved problem based on the unscented Kalman filter (UKF) using measurement data of inertial sensors. The proposed algorithm localizes the position of an unmanned helicopter using three new techniques. First, it models noises of acceleration measurement of inertial sensors by the white noise bias in addition to the commonly used random walking process. Then, this algorithm prioritizes the propagations of states in the UKF. Third, it leverages the time-varying GNSS dilution of precision in line with adjustments of the measurement noise covariances. The combination of these techniques makes it possible to localize the position of unmanned helicopters that are equipped with two-stroke engines, which generate large vibrations that result in noisy acceleration data. To facilitate automated tuning of the filter parameters, we further develop a population-based tuning method. The proposed algorithm with the auto-tuning method enables the positioning of unmanned helicopters and prompt reaction in the case of a GNSS outage without requiring tedious manual calibrations. The performance of this algorithm is experimentally validated on a fully instrumented model-sized helicopter. [ABSTRACT FROM AUTHOR]