Fault Tolerant Control of a Quadrotor Helicopter

Official URL: https://risc01.sabanciuniv.edu/record=b2583865_(Table of contents)

Application areas of Unmanned Aerial Vehicles (UAVs) are widened signi cantly over the last decade. O - the-shelf components such as low-cost sensors and actuators have broaden their usability. UAVs can be used in various missions from logistics to surveillance where ongoing research keeps encouraging new developments. Substituting a human operator with an on-board computer proposes a very appealing solution to improve the operation productivity and the cost. However, this replacement raises some safety concerns and it might be harder for a ight computer to recover from hazardous situations. Especially, UAVs that operate over crowded areas and high safety demanding environments introduce new constraints on their design process. Fault Tolerant Controllers (FTC) serve to reduce this safety gap by modeling and recovering faults during a mission. Fault recovery is a highly sought-after research topic where it is aimed to increase robustness and immunity to possible fault scenarios. This thesis deals with developing a fault tolerant controller for a quadrotor helicopter. A high- delity nonlinear model of a quadrotor is constructed using Newton-Euler formulation where Dryden wind e ects and sensor noise are included to simulate real-world ight conditions. A hierarchical control algorithm is employed for outer and inner control loops where PID-LQG controllers are designed to control position and attitude dynamics. For full state feedback, rst a linear Two-Stage Kalman Filter (TSKF) is implemented to detect and estimate the faults and provide state estimates. Second, an Extended Kalman Filter (EKF) is used to provide more accurate state estimates. In order to increase robustness to external disturbances and uncertainties in the plant dynamics, a disturbance observer is designed and integrated to the control system. Simulations carried out with the high delity model have shown that the proposed fault tolerant control algorithms successfully detect and compensate for actuator and/or sensor failures in a trajectory tracking task, and hence provide good tracking performance with reasonable control effort.