Navigation and Automatic Recovery of Fixed-wing Unmanned Aerial Vehicles

Today, unmanned aerial vehicles (UAVs) are seeing widespread use as a tool used by the industry, researchers, hobbyists and others around the world. Fixed-wing UAVs, although not in as widespread use a multirotor UAVs, have benefits with longer flight time and range. Low-cost systems typically base their navigation systems on global navigation satellite systems (GNSSs), and are susceptible to natural or artificial interference of GNSS signals. Additionally, while automatic launch of fixed-wing UAVs using a catapult is common, automatic recovery systems are not, which is of special relevance for operating UAVs from ships in the maritime industry. A skilled pilot is typically used for UAV recovery, requiring additional cost and limiting the operation to conditions with sufficient visibility. This thesis considers two main topics: First, UAV navigation both with increased accuracy using GNSS, and local navigation without GNSS for increased resilience. Second, systems for automatic recovery of fixed-wing UAVs. The first contribution explores the use of a low-cost navigation setup using two independent GNSS receivers, aiding an inertial navigation system by using tightly coupled integration of pseudorange, Doppler frequency and carrier phase measurements from two longitudinally separated GNSS antennas on a fixed-wing UAV. The goal is increased navigation accuracy for high-precision applications, such as georeferencing and photogrammetry. The multiplicative extended Kalman filter (MEKF) is used as the estimation algorithm. Measurement models are derived for the raw GNSS measurements based on the MEKF error state, taking into account antenna lever arms and explicitly including the difference in measurement time between the receivers in the measurement model for double differenced carrier phase. The proposed method is verified using data collected from a UAV flight, and different methods for receiver synchronization is compared. In a second contribution, the handling of the covariance matrix in connection to the reset of the MEKF error-state after a measurement correction is considered. This is a topic there is disagreement about in the existing literature. Previous work on this topic is reviewed, and it is argued for one specific covariance estimate modification. The third contribution demonstrates that the recently introduced direction finding feature of Bluetooth, using antenna arrays, is usable for UAV navigation outdoors at a range of up to 700m. It is shown how using repeated measurements from all array elements, instead of only the initial single-element reference samples as often suggested, can contribute to an improved estimate of the signal’s unknown carrier frequency offset, thereby improving the direction estimation performance. To run the direction-of-arrival estimation in real-time with high angular resolution on an embedded computer, we propose a pseudospectrum peak search strategy that combines a coarse search and a local nonlinear optimization for estimate refinement. In an open outdoor environment, using a square antenna array with 12 elements, the azimuthal performance is found to be very consistent with range, with noise standard deviation typically around 1◦. While the elevation is significantly affected by multipath at lower elevation angles, with visible disagreement between frequency channels, it is shown to be consistent with simulations of ground reflection multipath. In a fourth contribution, considerations for choice of Bluetooth array sampling sequence are considered. Bluetooth arrays use a single receiver in combination with software controlled signal switches to receive the incoming signal using all array elements. This implies that measurements are not simultaneous, and that an order of sampling must be chosen. Time-symmetric sampling is considered to reduce the error in estimated direction of arrival due to error in the estimated signal frequency. It is found, using simulation and measurements from an array with 12 antennas in field experiments, that such sampling orders can remove the direct correlation between frequency error and direction error for errors up to approximately 10 kHz. This is at the expense of reduced signal-to-noise ratio as the frequency error increases, as the measurements themselves cancel to remove the effect of the frequency offset. As a fifth navigation contribution, it is demonstrated how the effect of ground reflection multipath for Bluetooth navigation can be reduced by synchronizing measurements from two independent vertically stacked arrays, allowing hardware modularity using low-cost equipment. Using measurements from a 15 × 15cm array in field experiments, we demonstrate a significant reduction in elevation error at elevation angles in the 7◦ to 15◦ range where the error was largest when using a single array. In a sixth contribution, a system for automatic recovery of a fixed-wing UAV on a moving platform at sea is developed and demonstrated. The recovery functionality is added as a modular extension based on non-intrusive additions to a commercial off-the-shelf autopilot. Line-of-sight guidance for the UAV is used for line-following along a virtual runway that guides the UAV into the arrest system. The position of the UAV relative to the arrest system, and the orientation of the arrest system, is determined using GNSS receivers supporting Real-Time Kinematic (RTK) processing. The system is validated using two different UAVs in extensive field testing, first with a stationary recovery net, and then a net mounted on a barge towed by a ship. In the seventh contribution, it is demonstrated that Bluetooth direction finding can be used as a navigation system for automatic recovery of a fixed-wing UAV, guiding the UAV into an arrest system such as a suspended net, independently from GNSSs. The effect of multipath signal interference on the elevation angle estimate is handled by a constant offset calibration. The system is verified in field experiments using a Skywalker X8 fixed-wing UAV and a Bluetooth antenna array. As an eight and final contribution, a concept for planning approach trajectories for automatic recovery on ships is presented, where optimization is used to generate an approach trajectory that enables aborts as late as safely possible. The concept is demonstrated using a simple three dimensional UAV model, for several different scenarios. This is then compared to less complex approach plans. The implemented example demonstrates that optimization can be useful for planning approach trajectories and allow later aborts than the simple methods