The advent and rise of low-cost, high-frequency short-haul flights in Europe increasingly necessitates the need of more sustainable travel methods for trips with a distance under 1000 km. Many different proposals and policies have been put in place, including full flight bans for trips under 500 km, but the research space still has not embraced a possible co-existence of major air and ground transport options in this segment.

This research addresses this by developing a reproducible method using openly accessible data to assess air and rail network capacities of three of the busiest air transport routes in Europe. A capacity analysis is conducted and the modelling of travel time, travel cost, change in carbon dioxide emissions and passenger experience is performed, in order to investigate the feasibility and logistics of shifting passengers between air and rail.

The study finds that 30-50% of passengers could shift completely to trains, given sufficient rail network capacity, significantly reducing total carbon dioxide emissions and improving passenger experience throughout.

Predicting aircraft Take-Off Weight (TOW) has been a long-sought task by aviation stakeholders, especially for operational and regulatory bodies involved in flight planning. Unfortunately, TOW being a sensitive parameter to operational trends and cost indices, aircraft operators tend to keep it confidential. In recent years, Machine Learning (ML) algorithms have achieved increased prediction accuracy and capabilities in the field, provided the availability of TOW data. This paper studies the implementation of gradient boosting algorithms as well as Random Forests to better understand which algorithm is best-suited for aircraft TOW prediction (prior to take-off) solely based on Flight PLan (FPL) and Terminal Aerodrome Forecast (TAF) parameters. The study focused on flights at Amsterdam Airport Schiphol (AMS) for training the algorithms, using an 80-20% train-test split. Between Gradient Boosting Decision Trees (GBDTs), LightGBM, XGBoost, and Random Forests, GBDTs achieved the smallest Mean Absolute Percentage Error (MAPE) with 1.71 and 2.17% on the training and testing datasets, respectively. The most influencing feature proved to be the requested cruise speed, followed by great circle distance between airports, and aircraft type. The model was validated on Paris - Charles de Gaulle Airport (CDG) and Brussels South Charleroi Airport (CRL), proving its independence from airport type. However, the distribution of flights in the training dataset, especially that of aircraft and airline types, proved to be an influencing factor for the model's applicability to other airports. Future work includes expanding the training dataset to all flights in the European network, and introducing trajectory-based features such as aircraft speed intent. With a larger training dataset, neural network algorithms could also be explored. Finally, regarding the improvement of trajectory predictions, it was found that better accuracy of TOW predictions does not suffice and that other operational parameters' effect should be investigated, especially speed profiles.

To verify the aircraft position provided by Automatic Dependent Surveillance-Broadcast (ADS-B)
transponders, multilateration (MLAT) technique incorporates time difference of arrival (TDOA) measurements at multiple ground-based receivers to estimate the corresponding distances between those and the aircraft. This approach requires precise time synchronization among receivers that can not always be guaranteed. Alternatively, received signal strength (RSS) measurements can be utilized to derive these distances. In this paper, crowdsourced RSS measurements from 43 receivers were used to construct parameterized signal propagation models that capture the relationship between RSS and distance. The quality of these models
was evaluated by examination of model parameter and estimated distance errors in both 2D and 3D. The results show that at most 26.3% of available RSS measurements could be represented by the models given the cut-off criteria for model parameter errors. Moreover, the models with higher parameter errors demonstrated poor ability to capture RSS measurements at greater distances. The localization errors in MLAT with TDOA were compared to MLAT with RSS where the later resulted in more accurate position estimation in cases where the receiver clocks were not synchronized. However, MLAT with TDOA generally produced significantly more accurate position estimation given the reliable timestamps of signal arrival. The assessment of localization accuracy using crowdsourced data resulted in root mean square errors of 118.1 meters in MLAT with TDOA and 9858.6 meters in MLAT with RSS in 2D, representing the best results obtained.

Addressing the increasingly urgent need for sustainable aviation solutions, this study explores operational innovations as a quicker and more scalable addition to novel zero-emission propulsion systems. Through the use of regression-based causal inference methods, this study aims to understand the relationship between flight fuelburn inefficiency and the factors causing these inefficiencies. Such an approach allows for the attribution of inefficiencies to factors on an overall scale, requiring less specific domain knowledge for initial results. A case study, involving a sample of 100,000 flights, representative of European operations, reveals that airspace structure (3.2% increase in inefficiency) and turbulence along the flight plan (2.5% increase) are the leading causes, while variations in average airspeed, congestion, and crosswind contribute the least to flight inefficiency. A compilation of the results shows that the performed analysis leaves 61% of the observed flight inefficiency unaccounted for. Future work would include the exploration of different metrics even closer to actual climate and air quality effects, as well as detailed uncertainty quantification. The developed flight inefficiency prediction model allows experimentation with counterfactual scenarios, contributing to the global transition towards more sustainable air transport networks.

Air traffic delays have a major impact on the aviation industry, affecting airlines, passengers, and the broader ecosystem. With increasing regulatory and sustainability pressures, accurate delay predictions are now critical, as they enable reductions in contingency and discretionary fuel on flights, lowering total fuel usage. This research aims to develop an explainable supervised learning model to improve existing en route delay predictions, focusing on intercontinental flights from North America to Amsterdam Schiphol Airport. While prior studies have explored flight delay prediction, they have not addressed two critical research gaps identified in this research: the inclusion of day-of-operations features, such as passenger information, aircraft weights, and cost index, and the use of transatlantic flight data for
predictions 90 minutes before departure. To address these gaps, two Gradient-Boosted models, CatBoost and LightGBM, were trained using internal airline, airport, and METAR data. Both models outperformed the airline’s current in-use statistical model, with CatBoost achieving an MAE of 3.44 minutes and RMSE of 4.61 minutes and LightGBM achieving an MAE of 3.43 minutes and RMSE of 4.56 minutes. However, the model’s overall predictive accuracy, as
indicated by the R2 scores, remained relatively low, reflecting the inherent challenges of en route delay forecasting. The most significant performance increase over the current model was observed under adverse weather conditions. Despite these improvements, the test run also showed that the performance of the models deteriorates quickly as significant
differences exist in TAFs and actual weather. An explainability framework was developed to aid end users of the model, providing insights into the model’s decision-making process and helping to build user confidence in its predictions. This research advances en route delay prediction by providing more accurate delay forecasts, particularly in critical weather conditions, and proposes practical improvements to support future studies focused on enhancing model adaptability across diverse operational contexts.

Recent developments in language models have created new opportunities in air traffic control studies. The current focus is primarily on text and language-based use cases. However, these language models may offer a higher potential impact in the air traffic control domain, thanks to their ability to interact with air traffic environments in an embodied agent form. They also provide a language-like reasoning capability to explain their decisions, which has been a significant roadblock for the implementation of automatic air traffic control. This paper investigates the application of a language model-based agent with function-calling and learning capabilities to resolve air traffic conflicts without human intervention. The main components of this research are foundational large language models, tools that allow the agent to interact with the simulator, and a new concept, the experience library. An innovative part of this research, the experience library, is a vector database that stores synthesized knowledge that agents have learned from interactions with the simulations and language models. To evaluate the performance of our language model-based agent, both open-source and closed-source models were tested. The results of our study reveal significant differences in performance across various configurations of the language model-based agents. The best-performing configuration was able to solve almost all 120 but one imminent conflict scenario, including up to four aircraft at the same time. Most importantly, the agents are able to provide human-level text explanations on traffic situations and conflict resolution strategies.

The emissions of the transport sector inside the EU-27 have risen by 33 % between 1990 and 2019. A modal shift from unsustainable transport towards more environmentally friendly transport modes can be taken as one solution to mitigate the overall emission of the transport sector. In this paper, multiple open-source models and databases are utilized to compare travel emissions and time of air travel and various ground transport options, including car, bus, and rail. Compared with previous research that relies on closed-sourced or hand-collected data to extract public ground transportation routes information, this paper utilizes the openly accessible General Transit Feed Specification(GTFS) database, facilitating the calculation at a large scale inside EU-27. 820 pairs of routes between popular 41 city centers inside EU-27 are selected for comparison. The results consistently demonstrate that air travel always produces higher emissions per passenger than rail and bus travel for all routes. Emissions from cars are significantly influenced by occupancy rates and the type of vehicle fuel. The emissions from a single person in a petrol/diesel car can exceed those from air travel. However, if four people travel in a hybrid electric or electric vehicle, the per-passenger emission can be similar to rail. Among all public transport, the rail is the most competitive one to replace air travel by offering passengers similar travel time and reducing emissions. The trade-off factor between emissions and time is also investigated on its effect on the passenger route choice decision. In addition, this paper offers insights into the development of emission models and provides recommendations for various stakeholders.

Air Traffic Control (ATC) is tasked with ensuring safe separation between aircraft in a given Controlled Traffic Region (CTR). To achieve this an Air Traffic Controller (ATCo) verbally gives clearances using over the air communication. These clearances are kept track of by the ATCo using so-called ‘flight-strips’, which in modern systems are often digital. The allocation of an ATCo’s time is an important factor in the achievable traffic density within a CTA, which makes ATC an interesting domain to use Automatic Speech Recognition (ASR) models to allow a computer system to ‘listen in’ to the conversation of the ATCo. Although previous research has been done to create such models, few of these result in open available models or domain specific corpora for the creation of such a model. This study will therefore use two open in-domain and one out-of-domain corpora to create such a model and in this process identify domain specific challenges and how these challenges can, in certain cases, be mitigated.

Passenger transportation in Europe is often duplicated using modes of transportation which are environmentally inefficient. Quantifying the carbon dioxide emission inefficiencies of flights versus transit is beneficial to understand the potential savings of a modal shift. In this paper, we analyze the emissions in Europe from multi-stop flights using flight data from March 2019. The excess emissions are quantified by comparing each multi-stop flight with an intermodal journey that does not exceed 60 minutes of extra travel time. We find that on average, transfer passengers using intermodality can reduce their journey’s total
(segment) well-to-wheel and life-cycle assessment emissions by 33% (80%) and 30% (72%), respectively. 840 thousand (19 % of total) transfer passengers starting or ending their journey in Europe can skip the feeder flight while saving an average of 28 minutes of door-to-door travel time. For air travellers taking intra-European multi-stop flights, 157 thousand transfer passengers (10% of the total) do not have to even enter an airport. Further insights regarding the European mobility vision are made, with recommendations for various stakeholders.

Charting hyperlocal wind using a drone is a challenge of increased attention as it unlocks potential in a variety of fields. In context of the METeo Sensors In the Sky project, this study proposes a method to estimate the magnitude and direction of wind using a quadcopter in hover and cruise without a dedicated wind sensor. Only on-board sensors are used, with no knowledge of thrust and rpm. A deterministic method models drag experienced by the drone classically as a quadratic function of true airspeed, and estimates wind by deducting the estimated true airspeed with the GPS ground speed. Additionally, a particle filter is implemented and compared to the deterministic method. To validate the proposed methods, a series of verification flights is conducted in which the drone is flown straight into the wind, perpendicular to, and away from the wind. The results show that the proposed method can estimate wind for various ground speeds and altitudes. The root mean square error ranges between 0.3-2.0 m/s and 5-35 degrees in most scenarios with high true airspeeds. In most cases, the particle filter shows a slight improvement over the deterministic method, at the cost of reduced adaptivity to wind changes (gusts).

Punctuality is a key performance indicator for any airline. Hub-and-spoke airlines are particularly committed to on-time arrivals to guarantee passenger connections. Flights that are delayed at departure need to compensate for the lost time whilst airborne. Because fueling takes place well before scheduled departure, predicted departure delays determine the planned fuel amounts for en-route speed optimization. To prevent unnecessary fuel burn, airlines benefit from highly accurate departure delay predictions. This study aims to extend previous work on airline departure delay forecasting to a dynamic and probabilistic domain, whilst incorporating novel day-of-operations airline information to further minimize prediction errors. Random Forest, CatBoost, and Deep Neural Network models were proposed for a case study on KLM departures from Amsterdam Airport Schiphol between 1 January 2020 and 1 August 2023. The Random Forest model was selected for its superior probabilistic performance and high accuracy in predicting delays between 5 and 25 minutes, for which en-route speed optimization has the largest effect. The departure delay probability distribution forecasts are made at six distinct prediction moments: 90, 75, 60, 45, 30, and 15 minutes before scheduled departure time. At the 90-minute prediction horizon, the model reaches a Mean Absolute Error (MAE) of 8.46 minutes and a Root Mean Square Error (RMSE) of 11.91 minutes. Simultaneously, for 76% of flights, the actual delay is within the predicted probability distribution range. Considering the costs and emissions associated with the decision-making following the departure delay prediction model, this study puts strong emphasis on explainability. Flight dispatchers are therefore provided not only the predicted departure delay but also the main factors impacting the prediction, explaining the context of the flight. The versatility of the model was demonstrated in two shadow runs, where delays caused by familiar and unfamiliar factors were successfully predicted.

The application of automatic speech recognition in the air traffic control domain has been researched extensively. However, its primary application remains in the training and simulation of air traffic controllers. This is due to the insufficient performance of automatic speech recognition in specific environments, such as air traffic control, where strong performance and safety requirements are paramount. This study demonstrates how a large-scale, weakly supervised automatic speech recognition model, Whisper, could meet these performance requirements and establish a new approach to air traffic control communication. Fine-tuning Whisper in the air traffic control domain resulted in a word error rate of 13.5% on the ATCO2 dataset and 1.17% on the ATCOSIM dataset. Furthermore, the study reveals that fine-tuning with region-specific data can enhance performance by up to 60% in real-world scenarios.

Fuel-efficient flight operations and improved Air Traffic Management (ATM) operations are identified as one of the main pillars in achieving net-zero CO2 emissions by 2050. While considerable research has focused on airspace management and ATM operations, flight operations as managed by airlines have received little attention.

Accurate aircraft/tail-specific performance modeling is crucial for identifying savings while existing models such as Euro-Control’s BADA and manufacturers’ book models remain too generic. Additionally, trajectories including optimal routes, altitudes, and airspeeds, must be determined to minimize fuel consumption. Emerging solutions leverage in-flight data connectivity and Machine Learning (ML) methods to provide pilots with real-time decision support. However, quantifying and validating saving potentials present challenges due to unpredictable variables and performance modeling complexities.

This thesis aims to address these challenges by developing a tail-specific performance modeling framework using high-fidelity flight data and ML methods. The framework identifies and corrects tail- and flight-specific biases from the flight data, allowing fuel savings to be identified on a per-flight basis in post-flight analysis. The tail-specific performance model shows different Maximum Range Cruise (MRC) speeds than generic values determined by the aircraft. The benefits emerging from these optimal speeds are determined by high-accuracy simulations of different cost index strategies on flight-specific and network-wide levels. Three cost index strategies are evaluated and compared to generic MRC operations. Savings in both fuel and time are observed of 75 kg and 93 s, 96 kg and 111 s, and 127 kg and 107 s, on an average per-flight basis. In conclusion, this research demonstrates the existence and magnitude of fuel and time savings by flying tail-specific cruise speeds compared to generic values determined by the aircraft.

Wind estimation plays an important role in many aspects of our world, both for nowcasting and forecasting. This is also true for aircraft flight routes; where wind field knowledge can contribute to flight planning and safety. Aircraft can relay flight information back to ground-based receivers, from which local wind measurements can be derived. Such measurements can be used in interpolation or a meteo-particle model to produce estimates elsewhere; yet those means do not take into account wind dynamics. A physically inspired neural network approach is applied in this work. Flight paths are derived from the aircraft measurements, and used to simulate training data from ECMWF ERA5 reanalysis data. Next, actual aircraft measurements serve as input, with the goal of predicting an entire wind field over the Netherlands and surrounding areas at cruise altitudes. The network can be leveraged by introducing physical losses, which are found to smooth the predicted flows. The network is able to predict flow fields on both simulated and real measurements, given sufficient input data. It improves over the existing methods of relying on 6-hour forecasts or using the meteo-particle model. Magnitude error of wind is reduced to 2.85 m/s, coming from 3.88 m/s for 6-hour GFS forecasts and 4.78 m/s for the Meteo-Particle model. Directional, physically inspired networks have an error of 11.2 degrees, compared to 14.4 and 17.3 degrees for 6-hour GFS forecasts and Meteo-Particle model, respectively. However, both error metrics fluctuate significantly depending on whether flow is uniform or of divergent nature at a particular day. The network is fit for nowcasting and future work can be done for longer-term forecasting.

The consumer drone sector is expected to grow rapidly in the coming decades. In Europe alone, it is predicted that in excess of seven million such machines will be flying by 2050. This poses a risk of conflict in dense airspaces, with both aircraft and other drones. Such a growing market provides a need to make drones visible to ATC and other airspace users. While several passive surveillance methods exist, such as primary drone radar, a cooperative surveillance system would provide more data to airspace users and other drones to allow for features such as automatic separation. An Automatic Dependent Surveillance system concept is presented in this paper, allowing the drone to broadcast information about itself without external input. This is akin to ADS-B, from which the system inherits its format for the time being.
The study's main contents are threefold. The first consists of recommendations made on the basis of literature. Then, a simulation approach to examine system capacity and related constraints through a sensitivity study is done. Finally, a hardware proof-of-concept, consisting of inexpensive and simple off-the-shelf components is built and tested.
Overall, the paper demonstrates that such a system is indeed feasible. Through the literature, it was found that direct integration of the system with current ADS-B on the 1090 MHz frequency is possible, but may cause performance degradation for existing aircraft. Therefore, the carrier frequency and code allocation are changed. The simulation and capacity study shows that the system works in high-density scenarios (in excess of one drone per square kilometer), but will require additional work on hardware, format and modulation techniques to enable this.
Finally, the hardware demonstrator shows that an inexpensive COTS implementation with a range of approximately 200 meters is possible, on hardware drawing less than five Watts of power.

Radio frequency fingerprinting has been identified as a method to increase integrity in aircraft surveillance while retaining its openness. One way to uniquely determine transmitting devices is to distill the device its radio frequency (RF) fingerprint by looking at the physical features of the message signal it transmits. This physical layer fingerprint is the unique trace the transmitter leaves in the signals. This research proposes a method to RF fingerprint ADS-B and VDL2 messages to identify the transmitting aircraft using a complex-valued convolutional neural network model. Raw data from ADS-B and VDL2 messages are collected over multiple days using low-cost RTLSDR hardware. Results show that the model can identify ADS-B and VDL2 messages from up to 200 different aircraft based on the raw IQ preamble and bit synchronization samples of both signal protocols. Further analysis of the robustness of the model shows that the model accuracy can be highly affected by changing channel conditions during training and testing. This research shows that testing the RF fingerprinting model’s robustness to channel conditions is necessary since the models are prone to mistakenly considering channel information as transponder RF features.

Testing novel concepts in the allocation and use of aircraft is commonly done in air traffic simulators. Current day simulators rely on proprietary models and often propose questionable solutions. This paper uses BlueSky as simulator, in cooperation with performance models OpenAP and BADA. BlueSky aims to provide the user with full freedom to simulate any scenario. Improving BlueSky and adding functionality further aids in this goal. By expanding logic of the simulator, the user should attain more functionality without compromising the user-friendliness of BlueSky. This paper treats the implementation of thrust setting and flight path angle in BlueSky. This enables continuous descent operations to become part of the possibilities within BlueSky. The challenge was to prepare a set of ADS-B data using Fuzzy logic to validate the new model. Its robustness was then be demonstrated using a large number of flights.

The aviation industry faces the challenge of reducing its climate eects. This paper aims to determine whether optimizing flight trajectories can reduce the environmental impact of aircraft during the climb and descent phases. It addresses the research gap in sustainable flight trajectory optimization by combining state-of-the-art research on optimization techniques with research on aviation’s climate eects. A genetic algorithm is created for the climb and descent phases to minimize the impact of gaseous emissions; the objective function entails the climate and air quality costs of the emission of carbon dioxide, carbon monoxide, nitrogen oxides, sulfur oxides, and water vapor. OpenAP is used to evaluate the emissions of aircraft. The optimization includes variable mass and speed, wind, departure and arrival procedures, and airspace constraints. The optimization considers both the vertical and horizontal flight paths simultaneously. The model was tested for more than 22,000 flights in June 2018 at Amsterdam Schiphol Airport. It has been found that an average reduction in environmental costs of 8.6% and 18.6% is achievable for departing and arriving flights, respectively. The reduction depends on the aircraft type, weather conditions, and whether the aircraft departs or arrives. The open-source optimization model can be used for further research, for instance, evaluating the reduction possibilities at dierent locations. The research shows that the aviation industry can reduce its environmental impact considerably by optimizing departing and arriving trac trajectories.

Accurate 4D trajectory predictions are required for the implementation of Trajectory Based Operations. In addition, decentralized, free routing can make medium- to long-term flight trajectories more difficult to predict. Novel trajectory prediction techniques are needed, independent of waypoint-to-waypoint navigation and air traffic control operator behaviour. This research aims to improve the accuracy of medium- to long-term 4D flight trajectory predictions by incorporating a model that encompasses the dynamics of the air traffic situation. Data-driven techniques are well-suited to trajectory prediction purposes as high-fidelity air traffic and environmental data are widely available. A statistical analysis is first conducted to select the most suitable air traffic dynamics features for trajectory prediction purposes. The selected air traffic dynamics features are then translated to a spatiotemporal map. This paper proposes a composite, deep neural network to predict individual trajectories, merging a LSTM network with a 2D Convolutional LSTM based network.

Several initiatives are being developed to shift the current paradigm in Air Traffic Management (ATM) from the tactical-based approach to more strategic-based coordination of flights. This transformation of the ATM system relies on the improvement of predictive models that predict the 4D-trajectory of an aircraft. Previous studies primarily applied deterministic models that compute a single predicted trajectory. These models were assessed on their predictive accuracy. However, the accuracy of the predictions is highly impacted by uncertainties that affect the progression of a flight. These uncertainties are commonly related to the lack of detailed information concerning the flight intent, or the inaccuracy of positional and weather-related data. This study applied two probabilistic techniques: the model-based particle filtering model and the data-driven Gaussian Process Regression. Both approaches model the uncertainties and provide a predictive distribution of trajectories that allows for the evaluation of both the accuracy and the uncertainty of the predictions. These models were applied to predict the descent trajectories of aircraft arriving at Amsterdam Airport Schiphol. The results showed that the uncertainty of the predictions could be reduced by incorporating flight-plan data and meteorological data in the predictive models. Also, the accuracy was improved which demonstrates the importance of these sources of data in the predictions of aircraft trajectories. The proposed models have been able to quantify the uncertainty in trajectory predictions that could be used to further develop and improve the management and prediction of 4D-trajectories.