UWB-Aided Hybrid Navigation System in Degraded GNSS Environments

With the advent of urban mobility in urban and semi-urban areas, unmanned aerial vehicles (UAVs) are directly exposed to different global navigation satellite system (GNSS) threats which can severely impact the control and stability of the UAV. To enhance the PNT accuracy, alternative position sensors are required, and a robust sensor fusion framework to cope with all the instabilities introduced by each micro-electromechanical system (MEMS) such as inertial measurement units (IMU) or barometers. A key research question within the PNT community is the capability to provide to each UAV a robust and accurate PNT solution, reducing any risk of collision with other manned or unmanned aviation. The paper is proposing a hybrid federated fusion framework which is fusing data gathered from a GNSS receiver, a low-cost consumer-grade IMU, a MEMS barometer, a monocular camera, and a ultrawide-band (UWB) tag. To enhance the position accuracy of the fusion framework a gated recurrent unit (GRU) is used to predict IMU’s errors. Three extended Kalman filters (EKFs) are used as local filters in a loosely coupled configuration, while a GRU framework is used as master filter. Data is collected in a hardware in the loop simulation (HIL) using a Pixhawk 2.4 as flight control unit (FCU) and Spirent GSS7000 simulator to gather IMU, Barometer and GNSS data. Unreal Engine with the AirSim aid is used to collect visual odometry data using a monocular camera, while UWB position data is gathered using an external MATLAB script, based on a multilateration algorithm. The presented results show an increase in position performance due to the introduction of the UWB aid, which is computing position information using external UWB anchors, leading to a horizontal position error of 0.75 m and a vertical error of 0.15 m.