Design and Testing of Indoor UAS Control Techniques

Unmanned Aerial Vehicles (UAVs) represent an ideal platform for testing advanced control techniques. In recent years, many researches based on modern control theories and design of UAV autopilot algorithms have been completed. Most of the modern autopilots incorporate controller algorithms to meet the always more demanding requirements of flight maneuvers and mission accomplishment. However, due to the complexity and the computational need of the control algorithms, most of the commercial autopilots are based on Proportional Derivative Integrative (PID) philosophy, in which trial and error methods are used for the definition of the control gains. In this thesis, a PID controller system is proposed and it is compared with a variable structure method (that is a Sliding Mode Controller (SMC)). The controllers are designed and implemented for quadrotor indoor applications, facing the main challenges of this particular application. The overall structure of the proposed controllers is composed by two loops: (i) an outer loop related to the slow dynamics of the UAV, i.e. position, and (ii) an inner loop related to the fast dynamics of the system, i.e. attitude and angular velocities. Firstly, a nonlinear mathematical model of the plant is derived, starting from the well-known equations of motions for an aircraft. Then, the two controllers are designed and tested in a suitable simulation environment reproducing as close as possible a real-world situation so that testing can lead to a meaningful outcome. In this regard, a model of the sensors, an Extended Kalman Filter (EKF) and a Trajectory Planner are included in the simulation environment, which are the main elements of a quadrotor. Different types of trajectories are tested, starting from a hover maneuver and waypoints-following paths. Test results show better performance using the SMC, which is able to follow the proposed trajectories accurately and satisfying the given time constraints, while the PID exhibits larger position errors due to a slower inner loop resulting in slower response times in attitude variations. Furthermore, inertial parameters are varied within a limited range to mimic the presence of different payloads. The quadrotor autopilot is based on the Pixhawk 2.1 Cube and it is able of flying indoor thanks to an infrared-based motion capture sensor (the Otus Tracker), replacing the GPS as it is highly unreliable for indoor applications. Future work includes Software-In-The-Loop testing using code generation supported by the “Embedded coder support package for PX4 autopilot” in Matlab, and eventually code deployment on the target hardware to carry out Processor-In-The-Loop testing.