Flying an Autonomous Quadcopter on an ARINC424-Like Flight Plan

We designed an experiment to fly an autonomous quadcopter on a flight mission consisting of track-to-fix and radius-to-fix legs. The purpose of this experiment is to test the validity of flying autonomous Vertical Takeoff and Landing vehicles (VTOLs) on flight procedures that were designed for fixed-wing aircraft under the performance-based-navigation (PBN) concept. Using PBN ensures that an aircraft meets the navigation requirements of the airspace without specifying the navigation sensors necessary to meet the requirements. Instead, PBN specifies navigation accuracy thresholds on the drift from desired trajectory. We focus on determining the required navigation performance (RNP), which specifies the navigation accuracy required for 95% of the flight time. Here, we took a practical approach to assess what lateral performance a given VTOL can achieve in following a trajectory conformal to the ARINC424 database standard. To do so, we designed a flight controller that minimized the flight technical error (FTE), or the lateral cross-track error, by means of a proportional-integral-derivative (PID) controller. We conducted repeated flight tests over a known trajectory consisting of track-to-fix (TF), and radius-to-fix (RF) legs, where the flight location was an open field in low wind conditions. During the flight tests we maintained the speed but varied the north and east velocities. The flight controller hardware we used is the pixhawk 6C and the drone’s total mass was approximately 2.4 kg. Additionally, the firmware installed on the pixhawk is the open source PX4. Instead of using PX4 flight controller logic to fly on a mission consisting of waypoints, we designed our own flight controller that uses the position output from the navigation sensors and guides the quadcopter along the trajectory while minimizing the lateral cross-track error. The idea to have this additional flight controller is to ensure that the quadcopter is following a path and correcting its position with respect to that path instead of the quadcopter only adjusting its heading to target the next waypoint. The mean of the FTE recorded from the flight tests was mu = -0.036m and the standard deviation was sigma = 0.25m. To ensure that the 95\% confidence interval includes FTE values that are at the tails of the distribution, we stretched the normal distribution to cover FTE values at the tails but not the outliers. The new sigma value becomes 0.4m (60% increase in the original sigma value). Given the results we obtained with the hardware that was available and the weather conditions obtained, the accuracy requirement for the RNP specification is derived as 2*sigma = 2*0.4 approx 1m.