Methods for Control and State Estimation of Unmanned Aerial Vehicles

Unmanned Air Vehicles (UAVs) have many advantages for military and civil usage, such as aerial photography, surveillance, agricultural applications, to name a few. Even though they have many benefits, there are still problems that need to be solved in terms of control, observation and path planning of UAVs. This thesis investigates each problem separately and proposes novel methods: Maximum Correntropy Kalman Filter (MCStF) and Fuzzy Interacting Multiple Velocity Obstacle Avoidance Method (FIMVO). For comparisons and tests, DJI Tello and Crazyflie 2.0 UAVs have been used as real-time applications. In the control part, three different controller methods, proportional integral derivative (PID), type-1 fuzzy PID (T1-FPID) and interval type-2 fuzzy PID (IT2-FPID), have been compared with each other. For comparison, two different scenarios have been used under the unknown payload connected with a flexible cable. Robustness, stability criteria, and task performance have been investigated, and simulation and real-time results have been shown. The state estimation methods part has been the other main focus topic. A comparison of the maximum correntropy Kalman filter (MCKF) with the conventional Kalman filter is given. Then, a novel method, MCStF, has been proposed and compared with KF and MCKF. For comparison, interacting multiple method has been applied for each filter. The proposed and compared methods have been tested on the real-time system; in addition, the computation time and root-mean-square error have been defined as performance criteria. The last part of the thesis focuses on path planning. Fuzzy controllers and improved state estimation algorithms have been applied for multi-UAV usage. Instead of global mapping, local mapping algorithms, geometric-based Velocity Obstacle (VO), Reciprocal Velocity Obstacle (RVO), and Hybrid Reciprocal Velocity Obstacle (HRVO) avoidance methods have been focused on. These methods are compared with histogram-based collision avoidance methods in the simulation environment. Then, geometric approached velocity obstacle avoidance methods and the state-of-art velocity obstacle method, FIMVO, have been compared and tested on simulation and real-time systems. For comparison, the collision count and computation time are shown as performance criteria. Moreover, the added performance graph shows reliability, trajectory smoothness, task performance and algorithm simplicity.