A study on robust control system of a constant spinning quadrotor

Abstract

This thesis proposes a unique quadrotor system for 3D omnidirectional scanning with a laser line sensor and a robust control system. For 3D omnidirectional recognition, we propose the new approach to operate the quadrotor, which constantly rotates in the horizontal direction with a vertically wide field of view (FOV) laser line sensor. This method covers over 97$\%$ area of its surrounding sphere through a single sensor without an additional actuator. To maximize the FOV of the sensor, the sensor needs to be attached off-centered because a body of the quadrotor might block the sight of the sensor. This leads to several control issues, such as a variation of a center of mass (CoM) and uncertainty on an inertia matrix. To overcome these problems and achieve a stable attitude control with a constant rotation of the quadrotor, we propose the robust control strategies with a calibration method and a nonlinear disturbance observer. By obtaining the location of CoM offline, we can compensate for the effect of the CoM variation without straining the controller. With the CoM-compensated system, we designed the nonlinear disturbance observer to observe the model uncertainty and attenuate it. We also develop an integrated navigation system for the spinning quadrotor by fusing an inertial measurement unit (IMU), magnetometer sensor, and the laser line sensor. We developed the odometry algorithm based on an iterative closest point method to obtain relative motions of the quadrotor from the laser line sensor. The laser line odometry system has low sampling rates and significant time latencies because their data processing efforts consume considerable time; hence, by employing a precomputing Kalman filter, we propose a real-time optimal estimation scheme for integration of the IMU sensor and delayed measurements of the laser line odometry. Moreover, to generate 3D depth information from the line depth of the laser line sensor, the estimation of yaw angles of the quadrotor is necessary. To obtain fast and accurate estimates of the yaw angle, we propose an interference-compensating magnetometer calibration scheme and an integrated navigation system with calibrated measurement of magnetometer and gyroscope sensor. The actual flight test of the constant spinning quadrotor is carried out to examine the feasibility of the concept of the 3D omnidirectional scanning system and the performance of the proposed control and navigation system for the spinning quadrotor.