Aircraft with disruptive designs have no high-fidelity and accurate flight models. At the same time, developing models for stochastic phenomena for traditional aircraft configurations are costly, and classical control methods cannot operate beyond the predefined operation points or adapt to unexpected changes to the aircraft. The Proximal Policy Option Critic (PPOC) is an end-to-end hierarchical reinforcement learning method that alleviates the need for a high-fidelity flight model and allows for adaptive flight control. This research contributes to the development and analysis of online adaptive flight control by comparing PPOC against a non-hierarchical method called Proximal Policy Optimization (PPO) and PPOC with a single Option (PPOC-1). The methods are tested on an extendable mass-spring-damper system and aircraft model. Subsequently, the agents are evaluated by their sample efficiency, reference tracking capability and adaptivity. The results show, unexpectedly, that PPO and PPOC-1 are more sample efficient than PPOC. Furthermore, both PPOC agents are able to successfully track the height profile, though the agents learn a policy that results in noisy actuator inputs. Finally, PPOC with multiple learned Options has the best adaptivity, as it is able to adapt to structural failure of the horizontal tailplane, sign change of pitch damping, and generalize to different aircraft.

In this thesis, we consider investment optimization for airport infrastructure which is required charge and refuel electric and hydrogen powered aircraft using battery and hydrogen canister swaps respectively. The task at hand is to determine the most cost-effective infrastructure, consisting of spare batteries, and battery chargers for electric aircraft, and spare hydrogen canisters and fueling points for hydrogen aircraft. Previously developed models are expanded upon in this study by introducing the possibility of slot allocation, where flights that are not yet in possession of landing and take-off rights are assigned to them in such a way that requires the smallest extra infrastructure to be acquired. We derive several (mixed) integer linear programming formulations to solve this problem and develop heuristics which are able to approximate the optimal solution using only open-source resources. These are expanded upon by the introduction of instances where more than one battery type is allowed, electricity pricing becomes dependent on the time-of-use, and storage of electricity at the airport is allowed such that the peak demand can be as low as possible. Finally, we incorporate a distinction between the long-, medium-, and short-term decisions which have to be made by the airport operator into the model analysis. This allows the user to determine the most cost-effective infrastructure combination which can meet a required level of service. When testing these methods, we found that exact solutions can be found within reasonable time for cases with up to 200 batteries. Furthermore, a first-in-first-out policy heuristic has shown to be capable of generating promising results while being applicable to larger instances. The models have been illustrated in a case study at the airports Schiphol (Amsterdam) and Zestienhoven (Rotterdam - The Hague), where they have proven to be able to solve all daily instances to optimality.

This research applied a Dual Extended Kalman Filter (DEKF) to time-varying human operator (HO) parameter estimation in preview tracking tasks. The preview time parameter was of particular interest, as the amount of preview can be highly variable in practical situations. The filter was centered around a linear cybernetic HO model with six identifiable parameters for single-integrator and eight for double-integrator control tasks. The DEKF was applied to a range of time-varying simulations, both remnant-free as well as with realistic remnant based on an experimentally identified first-order model. In addition, the tests were validated using existing experimental tracking data. By keeping the HO physical limitation parameters constant (i.e. neuromuscular parameters ω_nms and ζ_nms and time delay τ_v), good estimation results were obtained, particularly for tracking of SI systems. General guidelines for the sensitive tuning process of the filter are proposed. The presence of coloured remnant noise in double-integrator tracking was found to greatly affect the quality of the estimates. Future versions of the filter should explicitly include models of remnant noise, to reduce its sensitivity and ease the tuning process. The obtained algorithm is applicable to single sets of measurement data without a priori assumptions on time-variance or the need for averaging. With some additional development, this makes the DEKF a suitable candidate for practical applications, such as driver monitoring and advanced driver assistance systems in the automotive industry.

With increasing UAV operations, tailored measures need to be implemented to integrate the massive employment of these vehicles in day-to-day operations. One of the current issues of UAVs is the noise emitted by these vehicles. The purpose of this research is to offer a doorway towards understanding the behavior of drone noise with changing operational and design parameters.
The radiated noise from four selected drones was experimentally measured under realistic environmental settings in an open field. The study dives through spectral analysis of a series of maneuvers, assesses the suitability of acoustic imaging algorithms for drone noise sources, explores adequate noise metrics, and lastly, identifies influential drone parameters essential for the modeling of drone noise in the future. Spectrogram and power spectral density analyses unravel the frequency range of interest between 200 Hz to 3000 Hz with dominant harmonics of the rotors’ blade passage frequencies. Above this range, the ground reflection hinders the proper identification of noise sources. Further investigation is done with 4 beamforming algorithms. As drone noise may exhibit complex radiation patterns due to interference of multiple rotors, sophisticated beamforming techniques (i.e. Functional (FB), Dipole (DB), and Quadrupole Beamforming (QB)) were implemented which are potentially superior to Conventional Beamforming (CB) in the setting of drone noise. The results of the beamforming algorithms have been verified with synthetic data. Notably, FB was found to perform best on a source if a correction for the expected source type is applied first, however, it is outperformed by other methods if the source type is not known a priori. While all implemented algorithms yielded excellent results on synthetic data, their accuracy greatly decreases when applied to real-life data. This can be attributed to the multitude of external factors and imperfection in vehicle design, as well as the dependency of DB and QB on prior information regarding the pole orientation on the rotors. In conclusion, only CB could be applied reliably. Analysis of CB over snapshots of drones during different maneuvers shows that the rotors prove to be the main sources of noise in hovering and increasing height from Hover. The source identification, however, becomes unreliable with the change in height. To yield a complementary view on the sound characteristics of drones in specific maneuvers, the sound pressure levels and various sound quality metrics have been computed. The metrics reveal the effects of drone parameters (i.e. rotor size, distance to the observer, thrust settings, etc.), thus leading to a selection of possible candidates for a future noise model. This may be useful to assess the human psychoacoustic annoyance of drone noise in an environment where drones become a part of our everyday life.

Potential usage of the workspace of driving simulators is restricted by a lack of knowledge on oncoming maneuvers, due to the unpredictable nature of human drivers. This research aims to explore the possibility of using information from driver inputs, dynamic vehicle states, and features of the road environment to improve motion cueing through simulator prepositioning. Probabilistic models were established to predict longitudinal maneuvering for a short scenario consisting of a drive on a two-lane road through a rural area. By combining the accelerator deflection, the vehicle’s velocity and the future speed limit in a logistic regression model, the area under the receiver operating characteristic curve was 0.84 for acceleration prediction and 0.77 for deceleration prediction, using a lookahead time of 5 seconds. In order to use these predictions for motion cueing, the prediction model was extended to a prepositioning module. The proposed design was tested in combination with a classical washout algorithm on a small hexapod-based driving simulator in a human-in-the-loop experiment. No distinguishable results were obtained in objective and subjective evaluation of the motion cueing quality, as possibilities to improve motion cueing were limited with the experimental setup used. Nevertheless, the workspace usage was improved significantly. The average distance that could be maintained from the edge of the simulator’s workspace, could be increased from 0.025 meter to 0.084 meter, for equivalent motion cueing.

Increased public awareness and political concern about the environmental footprint of aviation have created very ambitious goals to reduce emissions in the future. This objective of this research project is to develop a multidisciplinary turbofan engine design and sizing tool for conceptual engine design. The tool will be used in a parametric analysis to determine how the main design parameters impact engine performance and sizing. Methods to improve the propulsive and thermal efficiency have been analyzed to assess the effect on engine characteristics. The bypass ratio, fan pressure ratio, overall pressure ratio and turbine inlet temperature have been analyzed in a parametric analysis. It was found that higher values of these design parameters improved engine performance at the cost of engine size and weight.

Online Reinforcement Learning is a possible solution for adaptive nonlinear flight control. In this research an Adaptive Critic Design (ACD) based on Dual Heuristic Dynamic Programming (DHP) is developed and implemented on a simulated Cessna Citation 550 aircraft. Using an online identified system model approximation, the method is independent of prior model knowledge. The agent consists of two Artificial Neural Networks (ANNs) which form the Adaptive Critic Design and is supplemented with a Recursive Least Squares (RLS) online model estimation. The implemented agent is demonstrated to learn a near optimal control policy for different operating points, which is capable of tracking pitch and roll rate while actively minimizing the sideslip angle in a faster than real-time simulation. Providing limited model knowledge is shown to increase the learning, performance and robustness of the controller.

The popularity of Remotely Piloted Aircraft Systems (RPAS) is growing at an increasing rate as an alternative to manned aviation for different purposes and applications. So far, most of their operations have been of military character, but the demand for civil and commercial applications is growing exponentially. RPAS have been allowed to operate in segregated airspace by restricting other users from entering the volume associated to the operation of the RPAS. This temporary solution based on accommodating their operations on a case-by-case basis is not feasible on the long-term. Moreover, the latest forecast carried out by SESAR Joint Undertaking (SJU) reveals that by 2050 RPAS will represent 20% of the fleet. The available airspace is a scarce resource and is already saturated, which means that additional unmanned operations would not be possible
simultaneously. Integration in non-segregated airspace is the only manner in which the full benefits and capabilities of RPAS will be achieved, while maintaining the same levels of safety and efficiency as manned aviation.

In order to achieve a safe, efficient and transparent integration, a common regulatory framework and new technologies are currently under investigation. However, neither operational nor performance standards of RPAS integration have been addressed properly. For that reason, this project is focused on the development of a methodology to determine minimum operational and performance requirements by assessing the impact of RPAS operations on the Air Traffic Management (ATM) Network. The scope is limited to the en-route phase as a starting point since none of the flight phases have been analysed yet. This project is carried out in collaboration with EUROCONTROL.

RPAS are well known for presenting a wide range of performance characteristics, especially in terms of cruise speed and rate of climb. Thus, two different unmanned performance models have been chosen from the available performance models in the EUROCONTROL database: RP01 (MQ-9), which presents a similar performance to commercial aviation aircraft, and RP02 (RQ4A), which is considered a low-performance
RPAS in terms of cruise speed and rate of climb. Building on a base scenario in the Paris Control Terminal Area (CTA),Monte Carlo simulations of the air traffic have been performed in order to randomly vary the main input variable: the set of en-route flights that is substituted by unmanned aircraft. Two different scenarios have been analysed separately in order to establish the requirements: one for same-performance RPAS, and the
other for low-performance RPAS. In order to assess the impact, three different Key Performance Areas (KPAs) have been selected: capacity, efficiency and safety. These KPAs are in turn characterized by their corresponding Key Performance Indicators (KPIs), namely sector capacity, sector overload, flight time efficiency, flight path efficiency and number of potential conflicts.

Based on an analysis of the current occupancy of the flight levels, operational requirements are defined in terms of altitude segregation. RPAS are not allowed to operate in the most occupied flight levels (FL): from FL300 to FL400. Therefore, depending on their operation and performance characteristics (i.e. ceiling), they will adapt their cruise FL to operate below FL300 or above FL400. The optimal value of performance requirements for same-performance RPAS has been found to be a minimum cruise speed of 390 kts, while for low-performance RPAS this minimum cruise speed is of 280 kts. Additionally, for low-performance RPAS, it is found that the rate of climb of 1,000 fpm results in the best performing case when changes in FL are required.

The results show that the values of the key performance indicators for the base scenario are difficult to reach when RPAS are sharing the airspace. However, the establishment of minimum operational and performance requirements contributes to a significant reduction of the negative impact that the integration of RPAS operations implies.

The development of this methodology and its application to the characteristics of each airspace sector will contribute to the full integration of RPAS in non-segregated airspace.