With growing air traffic demand, infrastructure-constrained airports face increasing pressure to maximise runway capacity while maintaining high levels of safety and operational efficiency. Improving operational procedures plays a key role in this, as it aims to reduce delays, improve traffic flow, and make more effective use of available airspace.
This study presents a Mixed-Integer Linear Programming (MILP) model for the runway departure scheduling problem at Eindhoven Airbase, a single-runway airport with mixed civil and military operations. The model encodes the airport’s current Instrument Flight Rules (IFR) departure logic, including wake turbulence and radar separation, runway occupancy time, SID-dependent restrictions, and local release conditions required to maintain radar and border separation. Using data-driven scenario inputs that capture a wide range of traffic mix, runway direction, SID flown, and meteorological conditions, the model evaluates the reduction of release condition thresholds while optimising under three different objectives, namely maximising departure throughput, minimising total system delay, and minimising the maximum individual aircraft delay.
The resulting safe, regulation-compliant schedules are benchmarked against First-Come First-Served (FCFS) procedures and the optimised schedule using the current procedures. Results indicate an increase in capacity of about 20%, and a 38% reduction in delay during peak hours relative to the current procedure. The trade-off results between the three objectives align with those reported by Gupta et al. These findings show that minimising system delay leads to the best overall results, and that the current release conditions offer room for optimisation. ... Read more

Aircraft arriving at an airport are routed over surrounding areas. This creates a trade-off: aircraft expose population below to noise, whereas routing aircraft around densely populated areas requires additional fuel. Air traffic control (ATC) optimizes it. Convolutional neural networks (CNNs) and reinforcement learning (RL) have been used successfully for robot path planning, but aircraft have more inertia, and their cost is expressed continuously, instead of discrete, hard-boundary cost maps typical in robotics. It is therefore unclear to what extent the combination of a CNN and spatial maps transfers to ATC. This work presents a Soft Actor-Critic (SAC) agent that balances noise and fuel, and observes spatial population density through a CNN. Three different methods of representing the observation are explored: a sweep over resolutions and geometries, a combination of multiple scales in one observation, and domain randomization. Adding a population density map at 4 km/px already reduces the noise by 32.5% against an agent without a map observation, at only 13.6% more fuel. The best configurations are a multi-resolution observation of 4 km/px and 16 km/px and a 4 km/px observation with zoom augmentation. An observation with multiple resolutions make performance robust to the choice of resolution rather than substantially increasing it. On a held-out airport, the best configurations trade a small increase in noise (9.93% and 7.31%) for a decrease in fuel (1.15% and 1.64%) compared to a specialist model trained on that airport. The CNN and spatial map architecture provides flexibility: it generalizes across airports, where existing models need per-airport retraining. Moreover, the noise-fuel trade-off can be modified post hoc by changing the observed population density map. ... Read more

Air navigation service providers currently rely on deterministic demand forecasts for Air Traffic Flow Management (ATFM), which inherently fail to quantify forecast uncertainty. This study presents a top-down probabilistic forecasting framework for inbound air traffic demand using Air Traffic Control the Netherlands (LVNL) as a case study. The framework models the uncertainty of an existing deterministic forecasting system using quantile regression, thereby capturing heteroscedasticity directly at the aggregate demand level. To improve empirical coverage under non-exchangeable conditions, Conformalized Quantile Regression is combined with Adaptive Conformal Inference. The framework was applied using approximately two years of operational ATFM data. Compared to a statistical error-margin baseline, the proposed machine learning approach produced substantially narrower prediction intervals and achieved empirical coverage closer to the target, leading to a 25.1% lower Mean Winkler Interval Score (MWIS). Furthermore, Adaptive Conformal Inference more closely tracked the nominal 90% target coverage level over time than static conformal calibration, while also slightly reducing average interval width and MWIS. A retrospective operational analysis indicates that the probabilistic framework enables more flexible risk-based decision-making compared to deterministic forecasting alone. For a representative 150-minute prediction horizon, specific probabilistic decision thresholds simultaneously reduced both the false positive and false negative rates of capacity breaches relative to the deterministic baseline across the evaluated test dataset. These results demonstrate the potential of adaptive conformal prediction techniques to support probabilistic inbound demand forecasting within ATFM operations. ... Read more

Accurate identification of aircraft noise abatement procedures (NAPs) is essential for reliable environmental impact assessment and noise modelling. However, current analyses of the implementation of noise abatement procedures are mainly based on interviews with airlines. Previous studies establish methods to model NADP procedures and estimate flap settings. However, only limited studies have proposed methods for classifying noise abatement procedures. This study presents a data-driven approach for the classification of Noise Abatement Departure Procedure (NADP), Continuous Descent Approaches (CDA) and Reduced Flap Setting Approach using flight trajectory data. The framework integrates energy-based performance, flight performance indicators and machine learning techniques to classify flight trajectories into specific noise abatement procedures. The methodology is applied to operations at Amsterdam Airport Schiphol, evaluated on a dataset comprising over 19,000 inbound and over 19,000 outbound flight data in May 2025, alongside a machine learning training dataset of over 7,000 inbound flights. The results indicate that the developed framework is capable of classifying flight trajectories from enhanced mode-S and radar data to specific noise abatement procedures. However, the developed framework is constrained mainly by data resolution and feature sensitivity. Future work should focus on improving temporal resolution, expanding the dataset and incorporating event-based validation for enhancing model reliability. ... Read more

This paper investigates how descent wind forecast uncertainty affects VNAV-PATH idle descents and loss-of-separation (LoS) risk in dense arrival traffic at Amsterdam Airport Schiphol. A 3-DOF aircraft model, BADA performance data, KNMI wind forecasts, reconstructed FMS wind profiles, and arrival scenarios are used to assess operational consequences. The results show that wind forecast errors increase with forecast horizon and fewer uplink levels, affecting aircraft energy management, speed deviations, and the likelihood of thrust or spoiler intervention. Although aircraft pairs with LoS events exhibit larger speed and wind errors, increasing the share of older forecasts does not measurably increase the mean number of LoS cases per scenario. This suggests that traffic geometry and planned spacing primarily govern aggregate LoS occurrence. The findings support refreshing descent wind forecasts before top of descent (TOD), while highlighting the importance of wind profile reconstruction accuracy when evaluating the benefits of more recent forecasts. ... Read more

Extended Arrival Management (E-AMAN) enables earlier sequencing and upstream delay absorption by extending the freeze horizon, but increases exposure to prediction uncertainty and pop-up flights, which can degrade planning stability. This study presents the BlueSky AMAN Simulator (BAMS) and evaluates how stochastic uncertainty and pop-up mitigation strategies affect E-AMAN performance at Amsterdam Airport Schiphol.

Paired Monte-Carlo experiments are performed for freeze horizons of 14, 20, and 25 minutes under stochastic take-off, departure-route, and en-route uncertainty. Multiple mitigation strategies are evaluated, including Back-of-the-Line (BOL) scheduling, delayed-slot scheduling, and enabling planning at take-off.

The results show a clear trade-off between earlier planning and stability. Pop-up flights primarily drive sequence disruptions, while trajectory prediction uncertainty mainly increases temporal instability through repeated Expected Approach Time (EAT) revisions. Among the evaluated strategies, BOL scheduling at a 20-minute freeze horizon provides a balance between stability and delay performance, while longer horizons show diminishing returns due to uncertainty propagation. In addition, planning at take-off is shown to improve performance in both current and extended AMAN operations.

The findings indicate that feasible E-AMAN implementation at Schiphol requires either reduced uncertainty or stability-preserving scheduling strategies. ... Read more

The anticipated growth in air traffic is expected to increase demand on air traffic control capacity beyond its current limits, motivating the development of autonomous conflict detection and resolution (CD\&R) methods for self-separating airspace. While both state-based and intent-based conflict resolution methods have been studied extensively, a direct comparison between the two approaches under identical conditions is lacking. This paper addresses that gap by comparing the intent-based A3 CD\&R model to the state-based Modified Voltage Potential (MVP) resolution method. Performance is evaluated through verification scenarios and random traffic simulations at three density levels, using simulations in the BlueSky ATM simulator. Results show that the A3 method achieves substantially fewer conflicts and losses of separation than MVP across all density levels, with a markedly lower domino effect parameter indicating better airspace stability. However, when losses of separation do occur under the A3 model, the intrusions are larger than those observed with MVP. Regarding efficiency, MVP incurs large flight time increase due to speed-based resolutions, whereas the A3 model resolutions result in minimal flight time increases at the cost of larger route deviations. These findings demonstrate that intent information provides meaningful advantages for conflict resolution performance. ... Read more

Departure management at major airports relies on Target Off-Block Time (TOBT), a human-declared readiness estimate that is prone to inaccuracy. Sensor-derived predictions from computer-vision turnaround monitoring offer a complementary signal, but their higher update frequency and distinct error profile risk destabilising the departure sequence.

This study evaluates whether Predicted End of Ground handling Time (PEGT) predictions can improve departure sequencing at Amsterdam Airport Schiphol without sacrificing schedule stability. A characterisation of operational PEGT data shows that PEGT becomes more accurate than TOBT within approximately 27 minutes of departure, but produces nearly twice as many updates and exhibits pessimistic bias in the final minutes before off-block. These properties motivate the design of selective acceptance filters.

Using a reconstructed rule-based Departure Manager and counterfactual replay of 21,152 departures across 31 operating days (August 2024), 230 configurations of five conjunctive, interpretable acceptance filters were evaluated via Latin hypercube sampling. Results show that unrestricted PEGT adoption reduces vacated slots by 22.6% but increases late resequencing by 18.6%, confirming that improved accuracy alone does not guarantee operational improvement.

However, selective filtering, predominantly through suppression of frequent and late-stage updates, identifies a regime of 55 configurations (24% of those tested) that simultaneously improve all five metrics relative to the TOBT-only baseline: resequencing (-0.6%), late resequencing (-6.6%), vacated slots (-13.3%), TSAT delay (-1.6%), and on-time performance (+0.2%). These configurations improve both the TOBT-only and naive unrestricted-PEGT baselines on every tested metric, demonstrating that composite use of TOBT and selectively filtered PEGT can transcend the baseline stability–slot adherence trade-off.

The results are based on one month of nominal operations at Amsterdam Airport Schiphol; generalisation to disrupted conditions and other departure management architectures requires further investigation.
... Read more

ICAO projects a near-tripling of passenger volumes to 12.4 billion by 2050, yet the resulting bottleneck sits not in the sky but at the runway. Reinforcement learning agents can cut mid-air separation violations by over $99\%$, but treat the airport as a single infinite-capacity sink: traffic concentrates on one runway while others idle, a failure mode termed Runway Overload. This work closes that gap with a Goal-Conditioned Reinforcement Learning hierarchical framework coupling spatial and temporal arrival management. A high-level Manager uses Constrained Position Shifting to allocate runways and assign Required Times of Arrival. A low-level 4D-Worker, trained with Soft Actor-Critic, treats time as a state variable and learns path-stretching manoeuvres to meet arrival times without predefined templates. An Extra Trees regressor bridges the layers with fast arrival-time estimates, avoiding full trajectory rollouts. Evaluated in BlueSky-gym across two training phases and a coordination-evaluation phase, the selected policy reaches success rates above 99.8\%, on-time rates near 99.9\% and tracking error below 0.2 minutes. A heading-augmented observation, intended to preserve the Markov property during holding patterns, instead causes catastrophic failure on closely spaced parallel runways when trained without hindsight relabelling, possibly reflecting a spatial-reachability constraint; the selected policy omits both heading augmentation and hindsight experience relabelling without performance loss. Coupling the Manager with this policy no longer reproduces the runway-overload signature of prior single-sink models under either runway-assignment mode: static assignment splits traffic evenly by construction, whereas dynamic assignment leaves a small residual load imbalance and a separation-compliance cost that widens sharply at low $k$ before plateauing, without a measured benefit in this uniform-demand setting. Across 600,000 simulated aircraft, stalling is effectively absent and no delay propagation between arrivals is observed; under homogeneous-fleet, single-airport conditions the framework therefore addresses the spatio-temporal integration gap in trajectory-based arrival management. ... Read more

European hub airports face persistent demand-capacity imbalances generating 22.4 million minutes of delays in 2024, costing €2.8 billion. Traditional reactive approaches prove inefficient for long-haul transatlantic flights. Moreover, long-haul flights have historically been granted exemptions from ground delay programs, meaning short-haul flights disproportionately absorb delays during capacity constraints. This research develops an uncertainty-aware cruise speed control strategy for managing transatlantic arrivals to Amsterdam Schiphol Airport. The methodology decomposes arrival time uncertainty using Johnson distributions conditioned on flight status and temporal horizon, continuously monitoring probability that demand at Initial Approach Fixes exceeds capacity. One-time strategic speed adjustments trigger only when exceedance probability surpasses a predetermined threshold of 70 \%. Fast-time simulation of 30 days using BlueSky and EUROCONTROL trajectory data validates the approach. Across 79 interventions, the strategy achieved 35 \% capacity exceedance certainty reductions on average. These required only 8.72 minutes average delay through conservative Mach reductions of 0.01-0.05. The approach offers a practical, environmentally beneficial pathway for long-range air traffic flow management at capacity-constrained hubs. ... Read more

There is increasing interest in deploying autonomous air vehicles or drones in urban environments for missions such as package delivery to emergency medical transport. These missions have the potential to ease ground congestion and reduce greenhouse gas emissions in cities.

Operating in an urban environment poses challenges to air vehicles that are distinct from traditional air traffic management. Mainly, drones will need to avoid both dynamic (other drones) and static (buildings and city infrastructure) obstacles during flight. Additionally, the expected densities will be orders of magnitude larger than what is currently seen in conventional airspace.

However, this thesis limits the analysis to constrained airspace, where drones operate in urban areas between tall buildings and/or other infrastructure. This means that drones are restricted to fly along a constrained network that is above the existing street network or any other pre-defined network with a fixed route topology. In constrained airspace, drones can no longer fly directly to their destination and have points of convergence at the intersections of the network.

This thesis focuses on addressing challenges and risks of high-density air operations in constrained urban environments via two research goals. Thesis goal 1 analyses how airspace designs and rules affect the safety and efficiency of the urban airspace at varying traffic density. Thesis goal 2 develops and evaluates a method for analysing the operational feasibility of urban air missions considering local wind conditions.
... Read more

Traffic flows are highly dynamic. Traffic densities vary locally throughout the day, indicating potential for idle descents outside of night hours. Alternatively, fixed-FPA descents can be flown with lower uncertainty. This thesis presents the development of novel per-flight criteria for executing idle and fixed-FPA descents from cruise to the Initial Approach Fix (IAF), using associated trajectory uncertainties. The influence of the criteria on the number of successful idle descents is assessed throughout a 24-hour operation. Trajectory uncertainty models were identified from simulations under various wind conditions. Using conflict probabilities, nominal aircraft spacing, and IAF arrival times, descent criteria sets were developed. The number of successful idle descents was evaluated for various reference scenarios with execution restrictions in time frames, developed criteria sets, and trajectory uncertainties. The results show conditional inverse relationships between the trajectory uncertainty, reference scenario and criteria strictness, and the number of successful idle descents. Scenarios with stricter time restrictions contain fewer successful idle descents. This also holds for stricter criteria, provided that aircraft spacing does not conceal the effect of the different criteria. Similarly, higher trajectory uncertainty reduces the number of successful idle descents, provided that, additionally, the criteria set was not too strict to conceal the effect. At least 50% of the aircraft descending outside of peak hours could complete an idle descent, regardless of scenario, criteria set, and uncertainty set. Including peak hours, this changes to 40% of all descending aircraft. The difference between high and low uncertainty remained below 2% of all flights for all explicitly developed criteria and scenarios. The theoretical maximum is found when all aircraft fly idle descents in the allocated time frames. This is 76% outside peak hours and 68% overall. This research provides a foundation for assigning idle descents and demonstrates their potential by allowing them outside night hours. ... Read more

Urban air mobility (UAM) is presented as a potential solution to urban congestion. By utilising aerial vehicles for tasks like parcel delivery, public transport, and surveillance, pressure on traditional ground-based transportation infrastructure can be alleviated. This is particularly important with the rise of e-commerce and the increasing demand for fast and efficient delivery methods. UAM has the potential to revolutionise urban travel, offering faster commutes and enhancing surveillance capabilities for improved traffic management and emergency response.

The U-space concept, developed within the European Union, provides a framework for the safe integration of drones and small unmanned aircraft systems (sUAS) into urban airspace. It focuses on establishing services, regulations, and procedures to manage UAM operations effectively. An important component of this concept is Type Zu airspace, designated for high-density urban operations. This airspace requires strict regulations and safety-critical services like dynamic capacity management, conflict resolution, and continuous monitoring to ensure safe and efficient U-space operations.

Conflict detection and resolution (CD&R) of air traffic is required to ensure the safety of such operations, and VLL urban airspace presents unique challenges compared to conventional air traffic management. Buildings and other obstacles restrict aircraft movement, making manoeuvring and conflict avoidance more difficult. Unpredictable urban wind patterns further complicate flight planning and trajectory prediction. These factors, combined with the inherent complexity of urban environments, necessitate the development of robust CD&R algorithms and rules specifically tailored to the challenges of VLL urban airspace.

The core research objective of this dissertation is to identify and develop effective CD&R algorithms and rules for safe and efficient UAM operations in VLL urban airspace. This involves evaluating the limitations of existing CD&R methods, designing new algorithms that address the specific challenges of urban environments, and defining clear rules and procedures for aircraft navigation and conflict resolution…
... Read more

This thesis investigates future air traffic growth projections for an en-route environment and its impact on airspace complexity. The study compares Free Route Airspace, FRA, with the traditional ATS Routes Network, assessing airspace efficiency through various complexity metrics, including flight interactions, air traffic controller workload, and traffic patterns. These complexity metrics present the dependent variables of the study. The research builds a simulation model where the independent variables are the size of airspace, the type of demand, and the operational environment. With help of BlueSky ATM Simulator, the simulation model shows how independent variables can affect the complexity metrics. Additionally to this, the study evaluates the environmental impact by measuring CO2 emissions, which are found to be significantly lower under the FRA due to more direct routing. Also, the findings of this research suggest that FRA offers more efficient traffic distribution, particularly in larger airspace areas. In smaller airspace areas, the controller workload increases due to less predictable flight paths and the more flight interactions under Free Route airspace environment. The study concludes that FRA is advantageous for larger airspace areas, enhancing efficiency and sustainability, but it also introduces challenges in managing air traffic, particularly in smaller or highly concentrated airspace environments. ... Read more

With its 500.000 flights per year, Schiphol Airport is one of the busiest airports in Europe. Efficient runway use is vital for a smooth day-to-day operation. This paper investigates the new runway scheduler at Schiphol Airport, also referred to as the Departure Manager (DMAN). The new DMAN is built for a better utilisation of the outbound capacity with a higher predictability. This study aims to verify whether positive behaviour regarding earlier Target Off-Block Time (TOBT) updates leads to the desired results in capacity, predictability and reduced delay, based on historical data. An experimental model was used to simulate a new Departure Manager (DMAN), which uses a set of priority rules to assign flights to the runway slot in which they depart. The simulation did not find an overall significant result for all cases, runways and months, due to the averages influenced by the size of the data. However, a positive trend can be seen in the presented results, indicating that earlier TOBT updates lead to a better runway schedule. One cannot schedule better than on time, but late TOBT updates accumulate like a snowball effect throughout the planning. ... Read more

The usage of drones in urban environments is expected to grow rapidly in the coming decades. To ensure the safe operations of drones, conflict detection and resolution are vital. Currently, a lot of research has gone into state-based CD&R, which has proven effective in unconstrained airspace but suffers from a large number of false positive conflicts in constrained airspace. The use of intent in constrained CD&R has the potential to reduce the number of false positive conflicts and improve the safety of drone operations significantly. In this paper, an intent-based detection and resolution method for orthogonal constrained very low-level urban airspace is presented and evaluated against a state-based method. The intent-based method calculates the future position along the trajectory at a time interval of 3 seconds for each aircraft, and conflicts are then detected by comparing these positions. The conflicts are solved utilizing a rule-based algorithm. The results show that the intent-based method has a much lower false positive rate for all traffic densities, as well as a higher average detection time before conflict for larger look-ahead times compared to the state-based method. The resolution of the state-based method, however, shows better performance with fewer losses of separation occurrences. With improvements, the intent-based method's low false positive rate, combined with the use of a larger look-ahead time, allows conflicts to be detected more reliably and earlier than the state-based method, thereby facilitating earlier conflict resolution and enhancing safety. ... Read more

The use of drones in combination with a delivery truck can have a significant impact in improving the efficiency of last-mile delivery. Drones can be dispatched to customers from the truck, allowing the truck to continue delivering packages at the same time. This approach gives rise to the widely researched Traveling Salesman Problem with multiple Drones (TSP-mD). Numerous heuristic models have been developed to solve the problem in a near-optimal manner. However, these optimization strategies do not account for disruptions, which are common in delivery networks and can negatively impact their performance. While existing literature usually considers static models, a more dynamic approach could address these disruptions by adapting to real-time circumstances. To explore this, a dynamic method is developed in this paper for solving the TSP-mD. Its efficiency is compared to an existing static heuristic model from the literature. The comparison is performed in the BlueSky Open Air Traffic simulator, in which disruptions are introduced, such as truck delays and drone speed variations. Experiments in this environment demonstrate that the existing algorithm consistently achieves shorter mission completion times across all uncertainty settings. However, the newly developed method shows a significant improvement in performance under uncertain conditions. Therefore, the use of global optimization for the TSP-mD should be reconsidered. ... Read more

This paper explores the impact of Automatic Dependent Surveillance - Contract (ADS-C) on air traffic control (ATC) procedures, focusing on operational efficiency and safety margins in lower airspace. Through 64 simulation configurations, the study evaluates how varying airspace density, separation buffer size, and vertical error in ADS-C data influence operational metrics, such as fuel burn, track miles, and flight time. The simulations utilize synthetic ADS-C data with a 100% equipage rate, providing insights into how ADS-C can be applied to manage intersecting flight trajectories. Results indicate that separation buffer size is the most influential factor. Smaller buffers lead to significant reductions in fuel burn, track miles, and flight time compared to the baseline, though this comes at the expense of increased conflict risks. Airspace density demonstrated trends where higher densities showed the greatest fuel savings but more conflicts, highlighting a trade-off between operational efficiency and safety. These findings support the role of ADS-C in increasing predictability and improving trajectory management, both of which are key to Trajectory-Based Operations (TBO). By improving the accuracy of aircraft intent and trajectory data, ADS-C can optimize flight paths and enable more efficient air traffic management. However, carefully considering separation buffers and airspace density is essential to balance efficiency with safety. ... Read more

Event cameras are novel bio-inspired sensors that capture motion dynamics with much higher temporal resolution than traditional cameras, since pixels react asynchronously to brightness changes. They are therefore better suited for tasks involving motion such as motion segmentation. However, training event-based networks still represents a difficult challenge, as obtaining ground truth is very expensive and error-prone. In this article, we introduce EV-LayerSegNet, the first self-supervised CNN for event-based motion segmentation. Inspired by a layered representation of the scene dynamics, we show that it is possible to learn affine optical flow and segmentation masks separately, and use them to deblur the input events. The deblurring quality is then measured and used as self-supervised learning loss. ... Read more