Sensor fault tolerant control of a quadrotor uav

Abstract

This dissertation handles the issue of fault tolerant control (FTC) of a quadrotor unmanned aerial vehicle (UAV) in the existence of sensor faults. First of all, equations of motion are obtained using Newton's and Euler's laws. After that, the complete nonlinear model of vehicle is acquired. Altitude and attitude of the vehicle are controlled by PID controllers and the performance of these controllers are shown. Coefficients of the controllers are selected by trial and error method for best performance. Kalman filter is an optimal mathematical tool to estimate the states. The use of standard Kalman filter is restricted to linear models. Since the model of the quadrotor that is used in this thesis, is nonlinear, standard Kalman filter cannot be used. Therefore, nonlinear Kalman Filters, the EKF, the UKF and the CKF are used for estimating the states in the quadrotor. Also, performance comparison of these filters are made. It is seen that the CKF has less estimation error in the executed flight scenarios. Since the control of the quadrotor heavily depends on the measured values that receives from sensors, proper operating of the sensors is very important. However, small quadrotors and UAVs are mostly equipped with low-cost and low quality sensors. Measurements of these kind of sensors suffer from bias and external noise due to temperature variations, vibration and other external conditions. Since the bias is a very common fault for these sensors, in this study, a sensor bias is taken into consideration as a fault and happens abruptly at a certain time and continues throughout the scenario. Although it is not realistic, for convenience, it is assumed that the value of each state variable is measured by a dedicated sensor. To see the effect of the sensor bias fault on the system responses, two different flight situations are tested. It is noticed that the sensor bias causes a deviation between the actual value and the measured value of a state variable. Faulty measured values are fed back to the controller and as a consequence of this, quadrotor diverges from its reference command. By using the residual signals, sensor faults are detected and isolated. Then, two methods are proposed for removing the effects of faults and achieving active fault tolerant control (AFTC). The size of a fault can be obtained by analyzing the magnitude of the residual signals at the fault occurrence time. In the first method, rectification block uses this information and fixes the measured faulty sensor values and sends the corrected values to the controller. In the second method, faulty sensor value is ignored and measurement model of the filter is reconfigured. Thus, the quadrotor UAV follows its reference commands in the existence of sensor faults and the effectiveness of the presented two techniques are shown.