This thesis describes the development of robust discrete time sliding mode controllers where only output information is available. A connection between discrete time sliding mode controllers and so-called min-max controllers is described. New conditions for the existence of stabilizing output feedback discrete time sliding mode controllers are given for non-square systems with bounded matched uncertainties. A novel sliding surface is described; this in itself is not realizable through outputs alone, but it gives rise to a control law which depends only on outputs. An explicit procedure is also described which shows how a Lyapunov matrix, which satisfies both a discrete Riccati inequality and a structural constraint, can be obtained using LMI optimization. This Lyapunov matrix is used to calculate the robustness bounds associated with the closed-loop system.;For systems which are not static output feedback stabilisable, a compensation scheme is proposed and a dynamic output feedback discrete time sliding mode controller is described with a simple parameterisation of the available design freedom.;Initially, a regulation problem, to drive all plant states to zero, is considered. Then a new scheme which incorporates tracking control using integral action is proposed for both the static and dynamic output feedback discrete time sliding mode controller. The scheme requires only that the plant has no poles or zeros at the origin and therefore with an appropriate choice of surface, the controller can be applied to non-minimum phase systems.;The theory described is demonstrated for various engineering systems including implementation on a DC-motor rig in real-time and simulations on a nonlinear, non-minimum phase model of a Planar Vertical Take-Off and Landing aircraft. The effectiveness of the controller is further proven by its application for control of the longitudinal dynamics of a detailed combat aircraft model call the high Incidence Research Model, a benchmark problem used by the Group for Aeronautical Research and Technology in Europe. Simulations with real-time pilot input commands have been carried out on a Real Time All Vehicle Simulator and good results obtained.

Structural Coupling is the interaction between an aircraft's Flight Control System (FCS) and its structural and aero-dynamics. These interactions have the potential to cause significant problems to the aircraft, possibly through structural fatigue failure, or by corruption of sensor readings leading to degradation of the FCS loop. Presently notch filters are used to attenuate the sensor signals at the structural mode frequencies which cause significant problems. However, each of these notch filters adds a small amount of phase lag to the control loop which limits the performance of the FCS. This research has looked at using Kalman filters to provide an alternative approach to overcoming the structural coupling problem.

High-Altitude Long-Endurance Unmanned Aerial Vehicles (HALE UAVs) have been the focus for many researchers in the past decades, and are becoming more and more attainable as technology is improving. Recent interest into solar-powered HALE UAVs is leading the way to achieving flight-times which are not limited by a requirement for refuelling, and could continue for months at a time. Such aircraft have huge potential in providing reconnaissance or communications services, and could potentially take on the usual roles of satellites as a much cheaper alternative. One area of HALE UAV design which requires further technological developments is automated control of trajectory, and active alleviation of gusts. The latter concern is of particular interest to this thesis. In this work, a coupled flight dynamics and aeroelastic methodology will be introduced. This approach features a geometrically-exact beam model, which is required to capture the large deformations expected in such flexible aircraft, along with unsteady aerodynamics. The intrinsically nonlinear system is then linearised to provide an approximation to the system dynamics which is more approachable for control synthesis. With a linearisation of the system, robust linear control methods are applied to derive a controller which can reduce the loading onto the system, while simultaneously stabilising it and providing some degree of robustness which can still perform given the unmodelled dynamics that appear in the nonlinear system. This linear control approach will be investigated on two types of HALE UAV with very different dynamic responses to determine how well such control methods deal with nonlinearities, and what the limitations of such an approach may be to apply in real-life. In addition, various novel control effectors will be then considered for gust alleviation of a very flexible aircraft, using the closed-loop control as a method to fairly assess their effectiveness.

The future of aircraft lies in highly intelligent vehicles which are able to adapt themselves to different missions and take-off and land anywhere. Tailsitters, aircraft capable of controlled transitions between horizontal flight (like a fixed-wing aircraft) and vertical flight (like a rotary-wing aircraft), are one such form of vehicle. The focus of this research stems from a collaboration between the Surrey Space Centre and QinetiQ involving a new class of VTOL tailsitter - the QinetiQ Eye-On™ UAV - which offers uniqueness from all other known tailsitters in its use of twin helicopter rotors, elevons and a rudderless flying-wing design. A core objective of this research is to develop a control and navigation architecture capa.ble of handling the transitional flight regime in tailsitters and to understand the mechanism by which a transition can be controlled. Very little past research has addressed either of these in detail, often including only cursory modelling and simulation, no regard to how transitions can be 'shaped', and with no consideration of optimisation of transitions or whether their control laws are even robust. Another shortcoming of past literature is the minimal amount of experimental work undertaken which mostly uses only generic, simple single-propeller aircraft and does not consider transitional flight. Furthermore, examination of real-world applications where a tailsitter can be used has also been barely considered in literature. The limitations are addressed in three research divisions: I, II, III. {I} Development of a novel six degree of freedom (6-DOF) non-linear model with custom-designed numerical fluid dynamics, has allowed. high-fidelity simulation of all flight regimes to be performed. The developed control and navigation architecture is the first all-encompassing control architecture for the class of twin rotor tailsitters; it uses rudimentary low-level controllers and is capable of performing three different transitions: vertical to horizontal (V to H) , H to V (altitude elevation) and H to V (altitude invariant). The last of these is a ground-breaking discovery; transitions from H to V can be performed with virtually no increase in altitude. One improvement from past literature is that transitioning is undertaken in a closed-loop manner by commanding the vehicle to follow velocity and pitch setpoints. A carefully selected set of parameters has been devised to allow transitions to be shaped by transition time period, flight speed, sample size and smoothness of the control setpoint command. For the first time optimisation is applied to obtain ideal parameters for the transitions and robustness simulations stochastically consider environmental disturbances and variation of vehicle mass. [II] A comprehensive experimental framework has been developed tha.t uses various advanced testbed configurations to validate the control architecture, requiring the fusion of both aeroplane and helicopter technologies. Initially, an indoor motion capture testbed uses a series of precursory vehicles (including both quadrotor and Chinook) to pioneer taiisitter technology. A thrust testbed was also developed to explore thrust curve relationships and obtain optimal thrust zones for differing flight regimes. The outdoor testbed required the development of a complete self-contained autopilot system, including telemetry and ground station, which was tested in a progressive fl ight campaign spanning four flight locations across the UK. The experimentation forms the first demonstration of &-DOF untethered flight for the class of twin rotor tailsitters in VTOL, manual transitional and semi-autonomous transitiona.l modes. [III] In terms of systems analysis, two distinct civil and military scenarios are evaluated: linear asset monitoring, and perch and stare (which includes an innovative miniaturisation analysis) . The first thorough and realistic consideration is also given to the use of a reusable tailsitter v.'ith a docking station for staged exploration in extra-terrestrial environments. Industrially, the research programme extensively develops the technology necessary for autonomous flight of the UAV and extends from Technology Readiness Level (TRL) 2 to 6. Academically, significant contributions have been made to the field including: tailsitter modelling, transitioning methodology, control architecture, optimisation, testbed design, flight experimentation, systems design.

From the perspective control practitioners, control of dynamics systems subjected to varying parameters is not a new topic. However systematic methods to accommodate such problems are relatively few and recent. This thesis addresses a subset of such problems falling under the nomenclature of Reconfigurable Control Systems in relation to flight control applications. A survey in the initial phase of research indicated a wide range of ad-hoc solutions with relatively brittle or non-existent theoretical guarantees towards to the stability of the entire system. Often the reconfigurable architecture consists of multiple conceptual components performing task of identifying system parameter changes, monitoring degradation in system performance and eventually finding some corrective action to regain lost performance. The change in system parameters if attributed to faults and damages to system, then the task of the control system is to achieve fault tolerance. Such fault tolerance is of high interest for flight control community since such a control system adaptation may lead to accommodation of real life faults during aircraft operation such as control surface damages, hydraulic actuation failures etc. The thesis work aims towards developing online control redesign methods capable of taking into account realistic requirements. The goals are 1) To find control input values in presence of faults. 2) Accommodate changes in performance criterion in presence of faults and, 3) Incorporate actuator limitations such as rate and position bounds. The research work is divided in three subparts. The initial phase consists of a study of existing solutions and methods capable of providing reconfigurable flight control architectures. This phase also covers some flight control literature relevant in the context of faults. Though the conclusions of this initial phase seem theoretically simple and straight forward, it is interesting to understand the amount of time and efforts invested by real world flight control practitioners to deduce these results. Essentially the work flow of this research work stems from practical requirement eventually leading to theoretical developments that can approximate the requirements often demanded by the people in field. The second phase consists of study and application of existing Model Predictive Control methods to the field of reconfigurable flight control. MPC has been successful in major complex control problems due to its online constrained optimisation methodology. Along-with certain theoretical extensions it is well capable of providing a successful means to redesign control action online in presence of failures. Simulation studies of sufficient fidelity and complexity on a full envelop fighter aircraft nonlinear model prove such control reconfiguration capabilities of MPC. Some new extensions of MPC have been developed to show it performing in a superior manner to conventional nominal formulations. The third phase of the research work focuses on further theoretical developments in the field robust adaptive control in MPC frame of operation. A new MPC formulation is derived which can accommodate constraints, uncertainty and constant disturbances (due to failed inputs). The novelty lies in the theoretical properties of this MPC as under certain conditions it is guaranteed to be asymptotically stable. This setup implements an optimization problem more complex than that of the nominal case. Typically, when disturbances and uncertainties are incorporated into the performance measure within MPC formulations, mini-max (worst-case) NP-hard problems can arise. The thesis contributes to the theory of robust synthesis by proposing a convex relaxation of a mini-max based MPC controller by adopting a Linear Matrix Inequality (LMI) optimisation formulation.