Decentralized Robust Nonlinear Model Predictive Control For Uas

The nonlinear and unsteady nature of aircraft aerodynamics and limited range of controls and states make the use of linear control theory inadequate. For unmanned aerial systems in particular, control technology must evolve to a point where autonomy is extended to the entire mission flight envelope. This requires advanced controllers that have sufficient robustness, track complex trajectories, and use all the vehicle's control capabilities at higher levels of accuracy. In this work, a robust nonlinear model predictive controller is designed to command and control an unmanned aerial system to track complex tight trajectories in the presence of perturbances. The flight control system developed achieves the above performance by using a nonlinear guidance algorithm that enables the vehicle to follow an arbitrary trajectory; a formulation that embeds the guidance logic and trajectory information in the aircraft model, avoiding cross coupling; an artificial neural network, designed to adaptively estimate aerodynamic and propulsive forces; a mixed sensitivity approach that enhances the robustness for an adaptive nonlinear model predictive controller.

Robust Adaptive Model Predictive Control of Nonlinear Systems.

This book offers a novel approach to adaptive control and provides a sound theoretical background to designing robust adaptive control systems with guaranteed transient performance. It focuses on the more typical role of adaptation as a means of coping with uncertainties in the system model.

The problem of plasma vertical stabilization based on the model predictive control has been considered. It is shown that MPC algorithms are superior compared to the LQR-optimal controller, because they allow taking constraints into account and provide high-performance control. It is also shown that in the case of the traditional MPC-scheme it is possible to reduce.

The multi-objective trade-off paradigm has become a very valuable design tool in engineering problems that have conflicting objectives. Recently, many control designers have worked on the design methods which satisfy multiple design specifications called multi-objective control design. However,the main challenge posed for the MPC design lies in the high computation load preventing its application to the fast dynamic system control in real-time. To meet this challenge, this thesis has proposed several methods covering nonlinear system modeling, on-line MPC design and multi-objective optimization. First, the thesis has proposed a robust MPC to control the shimmy vibration of the landing gear with probabilistic uncertainty. Then, an on-line MPC method has been proposed for image-based visual servoing control of a 6 DOF Denso robot. Finally, a multi-objective MPC has been introduced to allow the designers consider multiple objectives in MPC design. In this thesis, Tensor Product (TP) model transformation as a powerful tool in the modeling of the complex nonlinear systems is used to find the linear parameter-varying (LPV) models of the nonlinear systems. Higher-order singular value decomposition (HOSVD) technique is used to obtain a minimal order of the model tensor. Furthermore, to design a robust MPC for nonlinear systems in the presence of uncertainties which degrades the system performance and can lead to instability, we consider the parameters of the nonlinear systems with probabilistic uncertainties in the modeling using TP transformation. In this thesis, a computationally efficient methods for MPC design of image-based visual servoing, i.e. a fast dynamic system has been proposed. The controller is designed considering the robotic visual servoing system's input and output constraints, such as robot physical limitations and visibility constraints. The main contributions of this thesis are: (i) design MPC for nonlinear systems with probabilistic uncertainties that guarantees robust stability and performance of the systems; (ii) develop a real-time MPC method for a fast dynamical system; (iii) to propose a new multi-objective MPC for nonlinear systems using game theory. A diverse range of systems with nonlinearities and uncertainties including landing gear system, 6 DOF Denso robot are studied in this thesis. The simulation and real-time experimental results are presented and discussed in this thesis to verify the effectiveness of the proposed methods.

Modeling error and external disturbances can severely degrade the performance of Model Predictive Control (MPC) in real-world scenarios. Robust MPC (RMPC) addresses this limitation by optimizing over control policies but at the expense of computational complexity. An alternative strategy, known as tube MPC, uses a robust controller (designed offline) to keep the system in an invariant tube centered around a desired nominal trajectory (generated online). While tube MPC regains tractability, there are several theoretical and practical problems that must be solved for it to be used in real-world scenarios. First, the decoupled trajectory and control design is inherently suboptimal, especially for systems with changing objectives or operating conditions. Second, no existing tube MPC framework is able to capture state-dependent uncertainty due to the complexity of calculating invariant tubes, resulting in overly-conservative approximations. And third, the inability to reduce state-dependent uncertainty through online parameter adaptation/estimation leads to systematic error in the trajectory design. This thesis aims to address these limitations by developing a computationally tractable nonlinear tube MPC framework that is applicable to a broad class of nonlinear systems.