Hamilton-Jacobi (HJ) reachability is commonly used to determine Safe Flight Envelopes (SFEs), yet it is often limited by the "curse of dimensionality". While Deep Neural Network (DNN)-based solvers like DeepReach mitigate scaling issues, they remain computationally and memory-intensive.

This paper introduces KANReach, a novel solver that leverages Kolmogorov-Arnold Networks (KANs) with B-spline activation functions for grid-free reachable set estimation. The architecture is evaluated using three case studies: a first-order regulator, a double integrator, and a 3D Dubins car. Benchmarking against DeepReach reveals that at comparable model sizes, KANReach achieves superior accuracy via Value Function Regression (VFR). However, it currently faces difficulties in Physics-Informed Learning (PIL) that limit its training efficiency relative to traditional DNNs. Additionally, this paper proposes Absolute Maximum Error Bounding (AMEB), a technique that exploits the unique convex-hull property of B-splines to derive formal safety guarantees that verify the computed SFE. ... Read more

Modern commercial aircraft rely on hardware redundancy and median-voting logic to ensure air data system integrity, but these mechanisms remain critically vulnerable to Common-Cause Failures (CCFs). During common-cause events, whether environmental (icing, volcanic ash, bird strikes), physical (insect nesting, drain blockage), or systemic (maintenance errors, manufacturing defects), identical faults across multiple sensors can evade detection, feeding catastrophic misinformation to flight control laws. This paper presents the development and validation of a distributed analytical redundancy architecture for detecting and isolating CCFs when traditional hardware consensus fails. The proposed solution employs a Kinematic Double-Model Adaptive Estimation (DMAE) framework utilizing Unscented Kalman Filters (UKFs). To mitigate simultaneous faults, the architecture deploys three independent DMAE estimators in parallel across the Captain, First Officer, and Standby sensor channels, synthesizing a single, high-integrity flight reference via median-consolidation logic. The framework's performance is rigorously evaluated against a comprehensive spectrum of complex, Airbus-defined fault scenarios, including simultaneous blockages, signal freezes, and biases. Furthermore, the algorithm is successfully transitioned from offline environments to real-time Simulink models for validation within the high-fidelity Airbus SESAME simulation environment. Experimental results confirm that the distributed DMAE successfully isolates faults and reconstructs true airspeed and angle of attack trajectories even when all physical sensors are differently compromised, demonstrating its industrial viability against common-cause anomalies.
... Read more

To reduce emissions and improve efficiency, modern aircraft designs are moving towards higher aspect ratios and lighter materials. While these design choices enhance performance, they also result in more flexible aircraft structures. This flexibility leads to greater interaction between rigid body motion and structural dynamics. Accurate modelling of these interactions is critical for evaluating aircraft handling qualities, predicting structural loads, tuning control laws for stability and performance, and developing simulators for pilot training.
System identification techniques provide a means to derive these models from flight test measurements. State-of-the-art system identification methods successfully capture the effects of structural dynamics. However, they rely on the assumption of (quasi-) steady aerodynamics. In steady aerodynamic models, changes in parameters such as angle of attack or control surface deflections are assumed to result in instantaneous changes in aerodynamic forces and moments. In reality, due to wake effects from unsteady aerodynamics, these forces and moments take time to develop, introducing delays in the response. Accurately capturing these delays is crucial for correctly predicting and modelling the aircraft’s dynamic behaviour. Failure to account for unsteady aerodynamics can lead to errors in load predictions, degraded handling quality assessments, and suboptimal control law design.
This dissertation develops a methodology for identifying a parametric flight dynamics and loads model from flight test measurements for a flexible aircraft that also include the effects of structural dynamics and unsteady aerodynamics. A two-step approach was adopted where the identification procedure consists of separate state estimation and parameter estimation steps. This allowed to perform model parameter estimation using a linear least squares approach. In contrast, alternative methods such as the output-error approach perform state estimation and model parameter estimation in a single nonlinear optimisation process. While this method can provide accurate results, it requires accurate initial parameter estimates to achieve convergence and a good fit, and it imposes a significantly higher computational load, making it less efficient for larger and more complex models.
A scaled Diana 2 glider unmanned aerial vehicle (UAV) was used as the flight test platform in this research. Using a UAV allowed to conduct flight testing with much fewer rules and regulations compared to full-scale aircraft testing, while also significantly lowering costs. A glider configuration was selected due to its high aspect ratio and flexible structure, making it well-suited for studying aeroelastic effects. Furthermore, the flight tests could be conducted at airspeeds and reduced frequencies corresponding to unsteady aerodynamic conditions... ... Read more

This dissertation develops a unified framework for defining, computing, and certifying safe operating regions for autonomous dynamical systems operating under deterministic, robust, and stochastic dynamics. As these systems are increasingly deployed in safety-critical environments—ranging from autonomous vehicles and air traffic management to robotics and automated decision-making—ensuring that their behaviour remains within safe bounds is of vital importance. Classical reachability methods offer strong theoretical guarantees, but they often rely on idealised models, struggle with high-dimensional systems, and cannot formally certify learned approximations. This work addresses these limitations through new safety definitions, robust generalisations of reachability theory, and a learning-verification pipeline for scalable certification.

The first contribution focuses on the notion of safe sets themselves. \Cref{ch:maneuverable} introduces the concept of the maneuverable control-invariant set—a novel refinement of traditional invariant sets. Unlike classical safe sets, which guarantee that the system can remain within a given region, maneuverable sets also ensure the system's ability to move between any two states within the set in a given time horizon. Although such maneuverability and invariance cannot always be achieved simultaneously, this definition adds a valuable dimension to safety characterisation. The set is computed using a combination of forward and backward reachable sets derived from Hamilton-Jacobi (HJ) reachability analysis. Building on this, \Cref{ch:safeset} presents an improved definition of a safe set that does guarantee both viability and full internal maneuverability simultaneously. It is obtained through a different combination of forward and backward reachable sets. While the first two chapters establish refined definitions of safe sets under idealised dynamics, real-world systems rarely operate with perfect models. To address this, the next contribution turns to the challenge of ensuring safety under model uncertainty and disturbances.

\Cref{ch:RobustBRT} addresses robustness under model uncertainty. It extends the HJ reachability framework by formulating a three-player differential game that includes unmodelled dynamics as an additional adversary alongside external disturbances. The resulting Robust Avoid Backward Reachable Tube (RABRT) generalises previous formulations by ensuring safety in the presence of both disturbances and uncertain system dynamics. The associated value function is shown to be the viscosity solution to a well-defined Hamilton-Jacobi-Isaacs (HJI) partial differential equation (PDE), and the method integrates with existing level-set tools, preserving compatibility with standard computational frameworks. While this formulation provides robustness to unmodelled dynamics, its computational demands grow quickly with system dimensionality. As classical level-set methods struggle in high-dimensional settings, recent work has turned to neural network–based approximations. The next chapter explores how such approximations can be made verifiable and trustworthy through formal certification.

To overcome the curse of dimensionality inherent in classical reachability methods, \Cref{ch:Care} introduces the Certified Approximate Reachability (CARe) framework. Instead of solving PDEs directly on grids, CARe leverages neural networks to approximate the value function and then uses formal verification tools—specifically Satisfiability Modulo Theory (SMT) solvers and Counterexample-Guided Inductive Synthesis (CEGIS)—to certify the learned approximation within a bounded error. This framework bridges the gap between deep learning and formal safety verification, offering a scalable method for high-dimensional systems while preserving correctness guarantees. While CARe provides a scalable and certifiable solution for deterministic systems, many real-world applications involve stochastic dynamics and probabilistic safety requirements. To address this, the final part of the thesis extends the reachability framework to systems governed by uncertainty and noise.

Chapters \ref{ch:StochasticABRT} and \ref{ch:stochasticsafeBRT} extend the thesis into the stochastic domain. \Cref{ch:StochasticABRT} introduces the Probabilistic Forced Backward Reachable Set, which identifies states from which the system enters an undesirable set with high probability, regardless of the control policy. The chapter develops convergent algorithms for computing this set in both discrete and continuous settings and provides error bounds under discretisation. \Cref{ch:stochasticsafeBRT} then presents the Safe Probable Backward Reachable Set (SPBRS), which guarantees not only that the system will reach a target with a given probability but also that it will remain inside a certified safe set throughout the trajectory. This is especially important in stochastic systems where intermediate safety violations can occur despite a safe terminal state. Two algorithms are developed—one computationally efficient and the other optimal—and their convergence and tractability are formally analysed.

Each chapter is supported by theoretical proofs, algorithmic implementation, and numerical validation on benchmark systems such as the Dubins car and double integrator. This dissertation offers a cohesive and extensible framework for analysing and certifying safety in both deterministic and stochastic systems. By introducing new definitions of safety, developing scalable algorithms, and integrating formal verification with deep learning, the work makes contributions that are both foundational and practical. It addresses longstanding limitations in the scalability, conservatism, and verifiability of reachability-based safety analysis. The results have direct implications for real-world autonomous systems, including autonomous vehicles, robotics, air traffic management, and safety-critical embedded systems. More broadly, the methods contribute to the growing effort to make machine learning and autonomy formally trustworthy, especially in domains where safety cannot be left to statistical performance alone. Looking ahead, this work lays the groundwork for future research in adaptive safety verification, stochastic control synthesis, and trustworthy decision-making in uncertain environments—pushing the boundaries of what is possible at the intersection of formal methods, control theory, and autonomy. ... Read more

Linear Parameter-Varying (LPV) models provide means to approximate complex, nonlinear, and time-varying system dynamics using a set of Linear Time-Invariant (LTI) models, interpolated by a scheduling function to ensure smooth transitions across the system’s operating envelope. This study demonstrates that multivariate simplex B-splines can serve as such function, evaluated for State-Space quasi-LPV (SS-qLPV) models by providing a global approximation using local basis functions. The Inverted Pendulum on a Cart Model (IPCM) is used as a demonstrator in an open-loop setting, with an affine LPV representation based on cart velocity and pendulum angle as scheduling parameters. Several scheduling function estimation methods: piecewise constant Zero-Order Hold (ZOH), polynomial uni and multi-variate Ordinary Least Squares (OLS), and multivariate simplex B-splines are evaluated. Results indicate that, at the same polynomial order, B-splines show higher approximation capabilities compared to polynomial methods, as shown by the root mean squared error (RMSE) of the residuals. However, under broader simulation conditions, LPV-ZOH can be computationally less expensive and can achieve lower RMSE, although piecewise constant methods have discontinuities at the switching points, which can have an impact to closed-loop performance. The study highlights trade-offs in scheduling
function selection and suggests future research in optimizing simplices for improved performance. Applying B-spline scheduling functions with gain scheduled controllers in closed-loop control is the next direction for increasing control performance in complex, high-dimensional systems. ... Read more

The thesis, conducted in collaboration with INNIO Jenbacher GmbH & Co OG (INNIO), presents two investigations advancing industrial gas engine control systems. The first investigation develops and compares four control strategies for engine speed regulation: iterative Linear Quadratic Regulator (iLQR), pole placement, Nonlinear Dynamic Inversion (NDI), and Incremental Nonlinear Dynamic Inversion (INDI). Each controller was derived from INNIO nonlinear Jenbacher engine models and evaluated through simulation under consistent tuning methodology. The analysis emphasized multi-variable control coordination, cross-coupling effects, robustness against model mismatch, and practical implementation considerations for commercial deployment.
The second investigation establishes a comprehensive methodology for identifying and characterizing fuel transport delays using GT-SUITE/MATLAB co-simulation calibrated with Jenbacher engine data. Virtual sensors at multiple locations quantified delay behavior across load acceptance, rejection, and steady-state operating conditions. Quadratic models were developed to enable real-time delay prediction based on boost pressure and operating mode. Both investigations provide control-theoretic frameworks and empirical models directly applicable to INNIO’s current and future engine platforms.
... Read more

Loss of Control In-flight (LOC-I) remains a leading cause of fatal accidents in commercial aviation, with aerodynamic stall identified as a frequent precursor. Effective Upset Prevention and Recovery Training (UPRT) requires representative Flight Simulation Training Devices (FSTDs), which in turn depend on accurate stall models. Current models are typically derived through in-flight system identification, but rely solely on global aircraft states and thus depend on assumptions about the onset and progression of flow separation. This study introduces a methodology to derive quantitative flow-separation maps from in-flight recordings of tufts and flow cones on the PH-LAB research aircraft during stall maneuvers.
High-speed video was analyzed through a dedicated image-processing pipeline to classify tuft states and reconstruct separation patterns. The resulting maps provide direct, spatially resolved evidence of stall onset and progression. This allows for a direct validation of existing Kirchhoff-inspired formulations, while highlighting their shortcomings in capturing spanwise variations and local effects of separation. Furthermore, synchronization with flight-test data allows for correlation of tuft measurements with global aircraft states and supports the development of a logistic model to describe local separation behavior. Beyond validating existing models, the proposed approach establishes a robust experimental framework for integrating flow visualization into stall model identification, with direct implications for the improvement of stall models and, ultimately, flight safety. ... Read more

Loss of control inflight is the most common cause of fatal accidents in aviation. Aerodynamic stall models are utilized in pilot training to enhance safety and prevent accidents. This research presents an advanced longitudinal stall model for the Cessna Citation II, achieved through innovations in modeling methodologies and experimental design. By introducing dynamic stall maneuvers with step inputs, the study mitigated $\tau_1$ and $\tau_2$ parameter correlation, enabling more reliable parameter identification. A separable nonlinear least squares method significantly reduced computational time for nonlinear stall model estimation, decreasing it from hours to seconds. This approach revealed two minimally correlated flow separation states, offering deeper insights into wing flow characteristics and improving model accuracy. The lift model was refined to incorporate pitch rate and elevator deflection effects, while the drag model was enhanced with a lift-induced drag component. Additionally, a center of pressure model was derived from pitching moment data, advancing the understanding of stability during stall. A novel structure for characterizing degraded elevator control effectiveness was also developed. These advancements resulted in substantial performance improvements, with mean squared errors for lift, drag, and pitch moment coefficients reduced by 32\%, 29\%, and 27\%, respectively. The models also demonstrated greater consistency across diverse maneuvers, evidenced by reduced variability in $R^2$ values. This work contributes to more accurate stall modeling, enhancing both aerodynamic understanding and aviation safety.
... Read more

One of the most widely applied identification methods for stall modelling from flight test data is based on Kirchoff’s method of flow separation. This approach has not lead to a satisfactory aerodynamic pitching moment model. The introduction of the so-called X-variable, representing the point of flow separation on the wing, interferes with identification of a pitch damping term, that is needed for dynamic stability. Moreover, flow separation is only a small contributor to the pitching moment, leading to a lack of physical interpretability. In general, Kirchoff methods lead to models incompatible with the nominal flight envelope. This paper presents a nonlinear unsteady model of the pitching moment using lag states of the angle of attack measurements, identified from flight test data collected with a Cessna Citation II. The model is formulated in terms of well-known stability derivatives and is a one-on-one extension of the nominal envelope model. Model regressors are selected from a large pool of candidates using Multivariate Orthogonal Function Modeling. The candidate pool is based on a newly formulated mathematical model, such that each model contribution has a clear physical interpretation. This has lead to a 𝐶𝑚𝛼 contribution depending on various lag states of the angle of attack, and pitch and downwash lag damping contributions as univariate splines. The model has good predictive abilities and can report a reduction of 55.9% in validation MSE compared to a Kirchoff based pitching moment model by van Ingen et al.. ... Read more

Loss-of-Control (LoC) is the primary cause of drone crashes, necessitating efficient onboard prevention systems that are effective in terms of sensor requirements, computing power, and memory. This study introduces a data-driven approach for detecting LoC in quadrotors, using Critical Slowing Down (CSD) theory as an Early Warning Signal (EWS) of approaching a critical transition. This paper employs a Fuzzy Logic Inference System (FLIS) to aggregate the CSD metrics alongside other EWS indicators, such as actuator phase delay, to provide a fuzzy indicator that quantifies the quadrotor’s stability. The proposed FLIS is applied to two LoC modes: the first is a yaw-induced LoC event during free-flight of the quadrotor in which growing off-axis instabilities during the maneuver culminate in LoC. The second is a roll-induced LoC event during a gimballed flight of the quadrotor in which growing off-axis instabilities during the maneuver also culminate in LoC. This approach proposes novel EWS indicators and a LoC detector and is generalizable across varying mass/size without needing precise state estimation of the quadrotor, instead only relying on onboard gyro and rotor speed data. Using real flight data from a GEPRO quadrotor, and a custom-built drone mounted on a 3-axis quadrotor gimbal testing rig, this paper demonstrates that various EWS indicators inferred with a FLIS can provide accurate, and timely detection of an upcoming LoC event, regardless of their specific causes or the maneuvers involved. This novel approach significantly enhances LoC detection rates relative to previous studies, and improves detection times, providing crucial additional seconds for corrective action. ... Read more

Neglecting actuator dynamics in nonlinear control and control allocation can lead to performance degradation, especially when considering fast dynamic systems. This thesis provides a novel method to account for actuator dynamics in the control allocation solution, dynamic incremental nonlinear control allocation, or D-INCA. The incremental approach allows for the implementation of a first order discrete-time actuator dynamics model in the quadratic programming (QP) solver. This model is used to find the optimal command inputs in addition to the desired physical actuator deflections, hereby compensating for actuator dynamics delays. Whereas, the baseline incremental nonlinear control allocation (INCA) approach requires pseudo-control hedging of the outer loop reference to increase closed loop stability margins under actuator dynamics delays. To its advantage, D-INCA does not require feedback of higher order output derivatives than INCA and can be used with nonlinear non-control affine systems. Furthermore, with adaptive D-INCA, or AD-INCA, an actuator dynamics parameter estimator is introduced to adapt the actuator model online, minimizing actuator tracking errors after actuator failures. The proposed methods are applied to a fighter aircraft model with an over-actuated innovative control effectors suite and results are compared to the baseline INCA controller. ... Read more

Aerodynamic stall has been a critical factor in recent aircraft crashes, leading to revisions in the regulations on the fidelity of stall models in flight simulation training devices. However, the updated regulations still lack a clearly defined accuracy required for effective pilot training. To determine the required accuracy, this research investigates how the Just Noticeable Difference (JND) thresholds for the stall abruptness parameter translate from a passive, observer task to an active flying task. An experiment was performed in the SIMONA Research Simulator with 16 active pilots, who performed two separate experiments. In one experiment, a stall autopilot flew the maneuver during which a staircase procedure was used to determine the passive JND threshold. The JND thresholds of a1 = 0.11 ± 0.094 found in this experiment were lower than the JND thresholds of a1 = 0.16 ± 0.14 found in a similar experiment by previous research. In the other experiment, the method of constant stimuli was used to determine the JND threshold for the active flying scenario. A psychometric curve, based on the Gaussian cumulative distribution function, was fitted using the combined responses of the participants. The resulting psychometric function of the active experiment lies entirely to the right of the passive psychometric function, and, when comparing the 75% thresholds for both experiments, the active threshold was found to be five times higher than the passive threshold. This indicates a decreased sensitivity to changes in stall abruptness when pilots are flying a stall themselves. ... Read more

This research evaluates various linear quadratic control techniques, with a particular focus on loop transfer recovery methods, to enhance the safety and robustness of flight control systems. It aims to address the limitations of classical control strategies in managing complex, multivariable systems by implementing advanced recovery mechanisms. The study utilizes a simulation model of the Cessna Citation PH-LAB aircraft to explore the effectiveness of linear quadratic regulator (LQR), linear quadratic Gaussian (LQG), and loop transfer recovery (LTR) control methods. It examines their time and frequency-domain characteristics, robustness against uncertainties, and performance in tracking normal acceleration commands. The research identifies the strengths and weaknesses of each method, contributing to the development of more reliable and efficient control systems for aviation.
... Read more

The rise in high-speed quadrotor applications demands aerodynamic models that explain the flight characteristics at these broader flight regimes. For developing these models using system identification methods, having information-rich data at these flight conditions is crucial. While there have been sufficient works in the input design for identification procedures in quadrotors, there are limited guidelines for high-speed flights. This work focuses on two key contributions to aid the high-speed quadrotor model identification process. First, an automated flight testing pipeline using various software tools has been developed to support the iterative input design process. Second, various inputs such as position tracking, trajectory tracking, and attitude injection while trajectory tracking have been designed. The goal of these input designs was to push the quadrotor velocities while trying to maximize the excitation in forces and moments. The HITL simulation tool, being part of the testing pipeline, was used to iteratively design and verify inputs before actual flight tests. Finally, automated high-speed flight tests of the designed inputs were conducted at the Cyberzoo indoor facility, reaching maximum velocities of up to 7 m/s. The results show that attitude injection while trajectory tracking provides a method to excite forces and moments during high-speed flights. This work is an initial attempt at understanding the effects of input design at high-speed flights with the goal of obtaining data for quadrotor model identification. The methodology and pipeline developed here serve as a platform to build and explore various other identification inputs with the scope to push the quadrotor flight speeds even further. ... Read more

This research presents a novel method to define the Barrier Function (BF) used for the synthesis of a Control Barrier Function (CBF). The proposed method uses multivariate B-spline estimators as the BF which allows for the creation of a more data-driven CBF. Additionally usage of B-splines offer a local basis, linearity in parameters and allow for additional constraints to be added to the regression problem. New theory is derived in order to utilize the power of multivariate B-splines for the creation of CBFs. The proposed methodology is implemented on Dubin's car. In the experiments two distinct multivariate B-splines are used as BFs. The first of these is created based on data sampled from a more conventional continuous reference function. The second one is based on data sampled from a reference function which is based on a combination of step functions. Numerical simulations using the created multivariate B-splines as BFs show that the the proposed methodology can be used to create a functional CBF. This is under the condition that the used multivariate B-spline has a degree of continuity equal to the relative degree of the system state which is to be controlled. In addition, the spline BFs estimated from discontinuous data show that the proposed methodology offers significant flexibility with regards to the types of safe sets which can be turned into functional CBFs using the proposed method. ... Read more

In recent years, quadrotors have emerged as versatile aerial vehicles, serving both recreational and critical industrial purposes in fields such as surveillance, search and rescue, agriculture, and transportation. However, ensuring their reliability and safety in the presence of operational faults or disturbances remains a significant challenge. This research aims to address these concerns by developing a robust controller that integrates Incremental Nonlinear Dynamic Inversion (INDI) with H-infinity loop shaping design process (LSDP), resulting in enhanced disturbance rejection and improved robustness to model uncertainty. A nonlinear aerodynamic model, based on flight test data, is used to enhance the fidelity of the quadrotor’s motion equations. The proposed controller is tested and compared with a traditional PD controller. Simulations of attitude control were conducted under different compensators to evaluate the impact of various robust control strategies, particularly in terms of disturbance rejection when combined with the INDI implementation in a 6-DOF nonlinear quadrotor simulator. Results demonstrate the superior disturbance rejection capabilities of the robust controlled system, highlighting the effectiveness of the proposed control strategy in improving performance and reliability under uncertain conditions. ... Read more

In this article a novel Control Barrier Function (CBF) named the fault tolerant Control Barrier Function (ftCBF) is introduced. The ftCBF is able to keep a vehicle within a predefined safe set with changing control bounds and changing system dynamics. The ftCBF is shown to be feasible in fault tolerant control applications, as opposed to existing CBF methods. This novel constraint is tested on a double integrator system, and on a non-linear Dubin’s Car system with changing system dynamics and changing control bounds. In the simulations it is shown that the ftCBF is able to keep the vehicle in the safe set with failure events occurring at any place in the timeline. The ftCBF contains design parameters that allow a trade-off between safety and performance. ... Read more

Digital fly-by-wire (FBW) control technology has had - and continuous to make - a great impact on modern-day aviation. In particular, it enables stability and control augmentation of the dynamic characteristics of the bare airframe and brings opportunities for substantial automation of tasks related to flight. At the heart of FBW control technology are the control laws, which are embedded in a flight control computer. The so-called divide-and-conquer paradigm long prevailed in the development of flight control laws, which is based on trim-and-linearization of the nonlinear dynamics over a wide range of conditions over the flight envelope. This process enables the use of powerful Linear Time-Invariant (LTI) control design and synthesis tools. Gain scheduling is performed in the subsequent nonlinear implementation step. This is a critical task, for which different techniques can be used to arrive at successful designs. However, the process can be intensive and time-consuming, especially in the case of highly nonlinear aircraft.

Nonlinear Dynamic Inversion (NDI) was developed as an alternative to the divide-and-conquer strategy. Instead of subdividing the operating domain into many different local regions, NDI brings the notional benefit of automatic gain scheduling. This drastically simplifies the nonlinear implementation step. Moreover, as it enables a decoupling of different parts of the control design, NDI brings advantages in terms of design modularity. In its classical form, these aspects are achieved through the use of an on-board model embedded in the control law. Alternatively, in an effort to reduce this model-dependency, a sensor-based incremental variant of NDI (INDI) was proposed in the past. This form aims to increase control law robustness in the face of parametric modeling offsets by relying more directly on sensor measurements instead. Accordingly, different inversion strategies (model-based, sensor-based, or combinations of these) lead to vastly different robust stability and performance characteristics. However, a systematic understanding of these robustness implications has long been missing. In this thesis, this problem is approached using H_∞-based multivariable analysis and synthesis techniques.

In addition to the question of robustness, this thesis also focuses on multi-objective control design in the context of control allocation for input-redundant plants. This concerns over-determined control problems, for which secondary performance criteria can be addressed in addition to the primary motion control task. In particular, it is investigated how the framework of incremental control allocation (INCA) can be used in such multi-objective design scenarios. ... Read more

Unsteady numerical simulation has been proven to be an essential tool for research. The quality of the results can be improved by using mesh adaptation. Mesh adaptation uses error indicators to refine the mesh in regions with high errors. The error indicators used are output errors with the most accurate output error estimation method being adjoint-based error estimation. However, for this method, the primal solution needs to be stored, which is storage intensive, especially for large unsteady simulations. The method proposed in this thesis uses a neural network autoencoder to compress and reconstruct the primal solution. This solution is compared to a reconstructed solution using Proper Orthogonal Decomposition (POD).

The one-dimensional unsteady Burgers equation is used as validation for the methods using a manufactured solution while the lid-driven cavity flow is investigated using the proposed method. The manufactured solution of the one-dimensional Burgers case could be exactly reconstructed using two POD modes. For the autoencoder a small latent space was used. For low resolutions, the small latent space did not prove to be a problem as the primal and residual could be captured accurately. However, for higher resolutions, the reconstruction error of the autoencoder became dominant for the residuals and resulted in erroneous adjoint-based error estimates while the primal remained qualitatively similar.

For the lid-driven cavity flow, the POD was still able to capture the solution using a low number of modes due to the smoothness of the solution. This resulted in an unfair comparison between the POD and autoencoder reconstructed solutions. The reconstructed autoencoder error estimates for lower resolutions were more accurate due to the latent space being large enough to capture the residual of the discrete primal accurately enough. When moving to higher resolutions, the autoencoder was not able to reconstruct the residual accurately enough leading to erroneous error estimates. Therefore, the latent space of autoencoders should be sufficiently large in order to gain an accurate reconstruction of the residual. If the latent space is large enough, the error estimate is accurate and the local error estimates can be used as a first iteration error indicator for mesh refinement. ... Read more