Approach Methods for Autonomous Precision Aerial Drop from a Small Unmanned Aerial Vehicle

Abstract

Abstract:One of the many basic operations that a fixed-wing unmanned aerial vehicle (UAV) should master autonomously is to deliver an object to a precise position. Among several possible delivery approaches, this paper chooses to focus on releasing the object from the UAV, at a carefully calculated state, and let it fall freely to the selected landing location. In such an operation, the computation of suitable release positions, velocity and attitude is essential. Due to time-varying disturbances like wind, a previously calculated release state may not be suitable once it has been reached by the aircraft. On account of these challenges, this paper suggests a dynamic calculation of the release state with respect to the wind velocity and current state of the UAV, with a sufficient re-optimization frequency. This is tested in simulation and with field tests, and compared with results from a static calculation approach.

Autonomous ballistic airdrop of objects from a small fixed-wing unmanned aerial vehicle

“…Even though the tuning led to a drag coefficient of 0.25, which was found sufficent, it is not perfect. Mathisen et al (2017) shows that the effect of a faulty drag coefficient is between 0.1 and 0.4 m, and that the effect increases with increased speed.…”

“…In turbulent wind conditions, the wind in point p can be subject to a sudden gust and turn out to be very different from the wind in the release point, and an update in point p would not give much improvement. A solution could have been to update the release point continuously, as suggested in Mathisen et al (2017), but more focus here was given on creating a robust system and eliminating other sources of error, see Sect. 5.…”

“…One source of error that would effect the airdrop system is an incorrect wind estimator. However, Mathisen et al (2017) shows that the target error due to 1 m/s increase in wind will be approximately 0.1 m. The same paper shows that the target error due to 1 m error in altitude is about 0.4 m, and that the target error due to 1 m/s error in the ground speed is about 3.4 m. The main source of error affecting the along-track error of the precision drop system seems to be the release velocity error. It arises when the UAV does not have the ideal velocity at the release point, and is caused by a combination of the UAV's longitudinal speed controller not functioning perfectly, and the UAV losing speed when the motor is turned off.…”

“…However, this altitude error was assumed constant throughout the test. The circular error probability (CEP) velocity and horizontal position accuracies of the autopilot's GPS receiver are 0.1 m/s and 2.5 m respectively (U-blox 2017), which means that the horizontal position inaccuracy will be dominating according to the error analysis in Mathisen et al (2017). The RMS accuracy for the timepulse signal of the GPS is 20 ns (U-blox 2017), which is negligible, but there is also a time delay from when the GPS solution is valid to it gets to the computer.…”

“…The solution builds on the publications in Mathisen et al (2017) and Leira et al (2015), where this paper's novel contribution is the combination of a target detection system, a target position estimation system, an airdrop planning system and a guidance and control system. The airdrop planning system is simpler and more precise than the one described in Mathisen et al (2017): The airdrop planning system presented in this present paper does not optimize its path frequently, like the one in Mathisen et al (2017), but it contains a more finely tuned ballistic model and presents more reliable results where the ability to deliver is higher. Moreover, the detection and georeferencing systems are implemented on the on-board computer and used in real-time, which was not the case in Leira et al (2015).…”