Nonlinear Autopilot Design for Extended Flight Envelopes and Operation of Fixed-Wing UAVs in Extreme Conditions

The project has developed autopilot algorithms for fixed-wing unmanned aerial vehicles (UAVs). Autopilots have evolved significantly over time, from early autopilots that merely held an attitude to modern autopilots capable of performing automatic landings and complex maneuvers. However, conventional autopilots operate within a relatively small flight envelope and the dynamic behavior of the vehicle is thus restricted such that linear control design methods can be applied. This puts unnecessary limitations on aircraft capabilities as well as wind conditions in which aircraft can be safely operated. UAVs can also operate in autonomous modes, which requires that the autopilot is fault-tolerant, reconfigurable, and intelligent in order to handle unforeseen events. Hence, we have developed nonlinear autopilot algorithms where the main goal was to increase the flight envelope of fixed-wing UAVs such that they can perform more advanced maneuvers and operate in harsh weather conditions. This functionality is needed in many commercial applications. In the first years of the project several new results have been published. This includes a nonlinear model predictive attitude controller for fixed-wing UAVs. Deep reinforcement learning attitude control algorithms have also been developed using proximal policy optimization. Finally, a complete aerodynamic model for the Skywalker X8 fixed-wing UAV have been developed. The aerodynamic derivatives in the model are calibrated by wind-tunnel tests. The mathematical model allows researchers to simulate agile maneuvers and test advanced nonlinear control algorithms.

The overall objective of the project was to address fundamental research questions related to the nonlinear control of unmanned aerial vehicles (UAVs). More than 10 master's students have graduated so far and the two doctoral candidates will graduate in 2021. The project has demonstrated that: - Nonlinear autopilots can increase the flight envelopes of UAVs significantly compared to conventional linear designs. - Nonlinear autopilots can operate in extreme wind conditions and thus increase the number of days UAVs can be operated safely. - Nonlinear designs can recover control after severe failures and unforeseen events. The main results have been published at international conferences and in journals.

The main motivation for the proposal is summarized below: - The main goal is to increase the flight envelop of fixed-wing UAVs such that they can perform more advanced maneuvers and operate in harsh weather conditions. This functionality is needed in many commercial applications. - Mathematically prove global asymptotic/exponential stability of the closed-loop system using nonlinear control theory. - Achieve fault tolerance and robustness of the nonlinear autopilot system. - Experimentally verify and demonstrate the nonlinear autopilot algorithms and their real-time implementation in challenging applications and case studies at NTNU's UAV Lab in cooperation with NTNU AMOS (NTNU Center of Excellence). Demonstrate the following UAV operations and maneuvers: 1. Agile flight: Track a path with large curvature (course- and climb-rates) using nonlinear control. 2. Operation in turbulent wind conditions: In strongly turbulent winds, the autopilot must be designed to handle large roll and pitch angles that could result from these disturbances also when not attempting any high-curvature maneuvers. 3. Autonomous precision landing using large angles of attack (AOA): We intend to land an aircraft autonomously at a given landing target using deep stall to reduce the impact speed. 4. Stabilization of camera: The goal is to operate a small UAV with camera without using a gimbal to stabilize the camera. The aircraft autopilot should stabilize the camera to avoid large bank angles during turning. The project targets education of two PhD and eight MSc candidates as well as training of one postdoctoral researcher through Working Packages 1-3 dealing with: WP1: Nonlinear control allocation and attitude control WP2: Nonlinear speed and path-following control WP3: System integration, testing and case studies The results will be published at conferences and in the top ranked journals in the field.