Design, Implementation and Testing of Advanced Control Laws for Fixed-wing UAVs

The present PhD thesis addresses the problem of the control of small fixed-wing Unmanned Aerial Vehicles (UAVs). In the scientific community much research is dedicated to the study of suitable control laws for this category of aircraft. This interest is motivated by the several applications that these platforms can perform and by their peculiarities as dynamical systems. In fact, small UAVs are characterized by highly nonlinear behavior, strong coupling between longitudinal and latero-directional planes, and high sensitivity to external disturbances and to parametric uncertainties. Furthermore, the challenge is increased by the limited space and weight available for the onboard electronics. The aim of this PhD thesis is to provide a valid confrontation among three different control techniques and to introduce an innovative autopilot configuration suitable for the unmanned aircraft field. Three advanced controllers for fixed-wing unmanned aircraft vehicles are designed and implemented: PID with H1 robust approach, L1 adaptive controller and nonlinear backstepping controller. All of them are analyzed from the theoretical point of view and validated through numerical simulations with a mathematical UAV model. One is implemented on a microcontroller board, validated through hardware simulations and tested in flight. The PID with H1 robust approach is used for the definition of the gains of a commercial autopilot. The proposed technique combines traditional PID control with an H1 loop shaping method to assess the robustness characteristics achievable with simple PID gains. It is demonstrated that this hybrid approach provides a promising solution to the problem of tuning commercial autopilots for UAVs. Nevertheless, it is clear that a tradeoff between robustness and performance is necessary when dealing with this standard control technique. The robustness problem is effectively solved by the adoption of an L1 adaptive controller for complete aircraft control. In particular, the L1 logic here adopted is based on piecewise constant adaptive laws with an adaptation rate compatible with the sampling rate of an autopilot board CPU. The control scheme includes an L1 adaptive controller for the inner loop, while PID gains take care of the outer loop. The global controller is tuned on a linear decoupled aircraft model. It is demonstrated that the achieved configuration guarantees satisfying performance also when applied to a complete nonlinear model affected by uncertainties and parametric perturbations. The third controller implemented is based on an existing nonlinear backstepping technique. A scheme for longitudinal and latero-directional control based on the combination of PID for the outer loop and backstepping for the inner loop is proposed. Satisfying results are achieved also when the nonlinear aircraft model is perturbed by parametric uncertainties. A confrontation among the three controllers shows that L1 and backstepping are comparable in terms of nominal and robust performance, with an advantage for L1, while the PID is always inferior. The backstepping controller is chosen for being implemented and tested on a real fixed-wing RC aircraft. Hardware-in-the-loop simulations validate its real-time control capability on the complete nonlinear model of the aircraft adopted for the tests, inclusive of sensors noise. An innovative microcontroller technology is employed as core of the autopilot system, it interfaces with sensors and servos in order to handle input/output operations and it performs the control law computation. Preliminary ground tests validate the suitability of the autopilot configuration. A limited number of flight tests is performed. Promising results are obtained for the control of longitudinal states, while latero-directional control still needs major improvements.