This paper presents a novel method for fast and robust detection of actuator failures on quadrotors. The proposed algorithm has very little model dependency. A Kalman estimator estimates a stochastic effectiveness factor for every actuator, using only onboard RPM, gyro and accelerometer measurements. Then, a hypothesis test identifies the failed actuator. This algorithm is validated online in real-time, also as part of an active fault tolerant control system. Loss of actuator effectiveness is induced by ejecting the propellers from the motors. The robustness of this algorithm is further investigated offline over a range of parameter settings by replaying real flight data containing 26 propeller ejections. The detection delays are found to be in the 30∼130 ms range, without missed detections or false alarms occurring.@en

As Multi-rotor Unmanned Aerial Vehicles, or drones, are gradually becoming more popular in civilian applications, the safety of these flying machines becomes a significant concern. Such drones are powered by multiple rotors to generate lift and control torques. Hence, the failure of rotors can severely threaten their flying safety. Direct consequences of rotor failures are loss-of-control and a subsequent crash if no ad-hoc flight control method can take over. Such a method, built on the principles of Fault Tolerant Control (FTC), is thus essential to improving the safety of multi-rotor drones. Fixed-pitch quadrotors are the simplest type of multi-rotor drones and have been extensively used in various applications thanks to their simplicity and higher energy efficiency. However, they suffer most from rotor failures since it requires a minimum of four fixed-pitch rotors to achieve full attitude control. Therefore, devising FTC algorithms for quadrotors presents a significant challenge. As there have been many efforts to develop FTC for quadrotors flying in nearhover conditions, a primary objective of this thesis is further expanding the capability of FTC methods to high-speed conditions where significant aerodynamic effects arise that brings large model uncertainties to the control algorithm. The high-speed flight conditions can be, for instance, the cruising phase of a quadrotor (e.g., delivery drone). Once rotor failure occurs, these aerodynamic effects can adversely impact the performance of FTC methods, and even drive the damaged quadrotor into upset conditions with abnormal attitude and angular rates. On the one hand, it is essential to improve state-of-art FTC methods withstanding significant aerodynamic effects as well as possible large initial disturbances. On the other hand, these aerodynamic effects need to be further investigated and modeled to facilitate the development of FTC in high-speed conditions. These two aspects constitute the two major parts of this thesis...@en

To explore the aerodynamic effects on a quadrotor in the high-speed flight regime and establish an accurate nonlinear model, free-flight tests with a quadrotor are carried out in a large-scale wind tunnel. The flight data reveal that complex aerodynamic interactions could appear and significantly influence the forces and moments acting on the quadrotor, which indicate the inaccuracy of state-of-art models established based on helicopter aerodynamic theory. To cope with this problem, gray-box models considering these effects are identified from flight data using a stepwise system identification approach, which combines both prior knowledge of rotorcraft aerodynamic properties as well as data observations. Previous models introduced in the literature are compared with the gray-box models. Validation results show an 80% reduction of moment model residuals and a 20% reduction of force model residuals. @en

This paper proposes an Incremental Sliding Mode Control driven by Sliding Mode Disturbance Observers (INDI-SMC/SMDO), with application to a quadrotor fault tolerant control problem. By designing the SMC/SMDO based on the control structure of the sensor-based Incremental Nonlinear Dynamic Inversion (INDI), instead of the model-based Nonlinear Dynamic Inversion (NDI) in the literature, the model dependency of the controller and the uncertainties in the closed-loop system are simultaneously reduced. This allows INDI-SMC/SMDO to passively resist a wider variety of faults and external disturbances using continuous control inputs with lower control and observer gains. When applied to a quadrotor, both numerical simulations and real-world flight tests demonstrate that INDI based SMC/SMDO has better performance and robustness over NDI based SMC/SMDO, in the presence of model uncertainties, wind disturbances, and sudden actuator faults. Moreover, the implementation process is simplified because of the reduced model dependency and smaller uncertainty variations of INDI-SMC/SMDO. Therefore, the proposed control method can be easily implemented to improve the performance and survivability of quadrotors in real life.

@en

The Safe Flight Envelope (SFE) is a prerequisite for flight envelope protection and essential for preventing Loss of Control (LoC) of a flying vehicle. Reachability analysis has been proposed for defining a SFE considering the dynamic characteristics of a system. However, the conventional Level Set approach for conducting reachability analysis is computational inefficient and impractical to solve problems having a state space with more than 4 dimensions. For this, we have proposed a computational efficient Monte-Carlo (MC) based approach. As an application, the SFE of an off-the-shelf quadrotor during the high-speed forward flight are estimated, based on the aerodynamic model identified from the high-speed flight data (V < 16 m/s). Taking into account the actuator dynamics, the state space of the problem has 6 dimensions in excess of the computational capabilities of the Level Set approach. By contrast, the result shows that the Monte-Carlo simulation based approach is able to solve this high dimension problem in a matter of seconds.

@en

In order to fill the gap of knowledge about the aerodynamic effects of a quadrotor in aggressive flight conditions, identification of the aerodynamic forces and moments occurring during high speed flight are performed. A gray-box model is established as a mapping from states, rotor speeds to resultant forces and moments. Prior knowledge about quadrotor aerodynamics together with a stepwise regressor selector is applied to identify the model structure. Data for this research is obtained from free flight tests in the Open Jet Facility (wind tunnel) of the TU Delft, with wind speeds up to 14m/s. State information is inferred from inertial on-board sensors and Motion Capture cameras. Moments produced by aerodynamic effects have been observed and precisely modeled. During validation, the new model showed superior prediction performance when compared to current models that neglect fast flight aerodynamic effects.@en