Improving the efficiency of commercial aviation is an important economic and environmental factor. Jet fuel costs account for at least 30% of airlines’ operating costs, which are consequently passed on to passengers after all. By analyzing flight trajectories for events, nonconformities, and anomalies, it is possible to gain a better understanding of how these factors contribute to increases in the cost of air operations and associate carbon dioxide emissions. Studies should also be undertaken in order to understand the correlation between these nonconformities and other situational factors, such as adverse weather conditions, restrictive traffic management measures, or events of capacity degradation, should also be undertaken. This paper presents a rule-based algorithm and the validation process to detect the in-flight holding pattern in 4D trajectory data. In addition to the method, a near-real-time holding pattern monitor has been evaluated and applied by the Brazilian Air Navigation Management Center (CGNA), which enhances situational awareness in air traffic flow management and creates a historical database of the Brazilian Airspace Control System (SISCEAB). With this database, future work is being proposed to increase the predictability of aircraft arrival times, the correlation with the weather forecast, and the prediction of occurrences of holdings for flight plans.
As electric vertical take-off and landing (eVTOL) technology advances, planning for data-driven and safe urban airspaces becomes increasingly essential. This study employs artificial intelligence clustering techniques to analyze aircraft trajectories in São Paulo’s urban airspace – one of the world’s busiest – to characterize the current airspace usage against official regulations and provide a foundational framework for future eVTOL trajectory studies. Using radar data from helicopter flights, this research applies unsupervised learning algorithms, specifically DBSCAN and hierarchical agglomerative clustering, to group and analyze trajectories. Data preprocessing steps, including filtering and dimensionality reduction through Principal Component Analysis, prepared the dataset for meaningful clustering. Results indicate that the models effectively identified daily and hourly flight patterns, offering insights into air traffic behavior relative to regulated air corridors. A comparison of cluster centroids with established visual corridors for helicopters in TMA-SP 2 highlights the potential of these methods to support safe and efficient airspace management and monitor compliance with designated routes. Future work will focus on refining models, expanding data sources, and performing analyses of airspace use with a focus on emerging urban air mobility vehicles.
The complexity and criticality of airspace control, combined with the increasing demand for flights, underscore the need for advanced tools to enhance the optimization and safety of operations. Given the numerous variables that influence air traffic and the difficulty of analytically assessing the impact of each, Artificial Intelligence (AI) techniques provide a viable strategy for more accurately weighting each factor. Considering that runway configuration change is one of the most influential variables in air traffic management, this work presents the development of a system to predict runway configuration at Congonhas Airport (SBSP) using Gradient Boosting techniques, specifically LightGBM. The developed LightGBM-based classification model utilizes data from the Meteorological Aerodrome Report (METAR), Terminal Aerodrome Forecast (TAF), and Weather Research and Forecasting (WRF) to predict potential runway reconfigurations, determining which side of the runway will be in use. Predictions are made for 12 future time points, spaced 15 minutes apart, providing forecasts from the next 15 minutes up to 3 hours ahead. The model achieved a 98.6% accuracy on the test set and, when tested on an unseen data period outside the training and test set, maintained a high accuracy of 92.0% with an F1-Score of 91.8%. These results highlight the effectiveness of the developed model, outperforming traditional rule-based methods currently in use at ICEA (Institute of Airspace Control), which achieve approximately 79% accuracy. The objective is to provide a more precise quantification of meteorological variables’ influence on airport operations, contributing to more efficient and safer airport activity planning.
The rapid growth in global air travel demand has significantly increased air traffic, leading to heightened delays and environmental concerns. This surge underscores the critical need for optimized sequencing techniques in air traffic management to maintain operational efficiency. Operational Research provides powerful methods to tackle these complex issues by offering structured approaches for optimization. In this study, we develop a Mixed-Integer Programming model to address the Aircraft Landing Problem, incorporating operational constraints specific to Guarulhos International Airport. Our model aims to enhance the efficiency of aircraft landing sequences by minimizing cumulative deviations from unimpeded times, thereby reducing fuel consumption and improving overall airport performance. To assess the model’s viability and efficacy, we employed a data-driven approach, collecting real operational data from Guarulhos International Airport during select days in 2023. This approach enabled us to evaluate the model under various traffic scenarios, including those involving threshold changes. The results indicate that our model’s implementation consistently achieved viable solutions with practical execution times and outperformed both actual operational data and the First-Come, First-Served (FCFS) baseline regarding deviations from unimpeded times. Under clear visibility and moderate traffic conditions, the model showed an average improvement of 93.33% over real data and 29.91% over the FCFS baseline. In peak traffic scenarios, it demonstrated a maximum improvement of 60.15% compared to FCFS when optimization began at 100 nautical miles (NM) from the airport, with an average improvement of 45.37% across the analyzed days. When optimization began at 40 NM from the airport, the model achieved an average improvement of 31.64% over FCFS. Similar gains were observed in scenarios involving threshold changes. These findings demonstrate the model’s potential as a robust tool for optimizing landing sequences, contributing to reduced deviations from unimpeded times and enhanced airport efficiency.
A operação dos aeroportos está diretamente relacionada às condições meteorológicas, especialmente à direção e intensidade dos ventos. Fatores como mudança na direção e intensidade do vento, podem gerar riscos para as operações de pousos, que desencadeiam eventos de reconfiguração de pista (ou troca de cabeceira). Essa reconfiguração pode resultar em atrasos, acúmulo de voos em fase de aproximação e, por consequência, o aumento no consumo de combustível, ampliando as emissões de CO2, uma vez que manobras nessa fase são menos eficientes. Esse acúmulo de voos esperando autorização para pouso é denominado Efeito de Fila. Neste cenário, até que o voo receba autorização para pouso, diversas medidas ATFM podem ser aplicadas, como a Espera em voo. O trabalho propõe avaliar a influência da reconfiguração de pista na ocorrência de esperas em voo. Inicialmente, foi verificada a relação entre a ocorrência de espera e a reconfiguração de pista, por meio do teste Qui-quadrado com p<0,001. Além disso, o teste de Wilcoxon indicou que o evento de reconfiguração de pista ocasiona uma diferença na mediana de esperas quando comparados os períodos de 3 horas antes e 3 horas depois da reconfiguração de pista, com p < 0,001. Os resultados indicaram que a reconfiguração de pista exerce influência na ocorrência de espera em voo no aeroporto internacional de Guarulhos.
Aprimorar a eficiência do espaço aéreo é uma prioridade contínua na gestão do tráfego aéreo. Um indicador fundamental de desempenho nesse contexto é o tempo adicional gasto nas áreas terminais. Ao monitorarmos esse indicador de forma adequada, podemos tomar decisões que trarão benefícios tanto econômicos quanto ambientais. Um requisito para calcular esse indicador é identificar os setores de entrada na área terminal. Com base no cenário apresentado, o objetivo do estudo foi implementar um modelo para identificação automática de setores de entrada nas regiões a 40 NM (Milhas Náuticas) e 100 NM de distância de cada um dos quatro aeroporto mais movimentados do Brasil. Para isso, foram testados os métodos de aprendizado não supervisionado K-Means e DBSCAN, para a tarefa de clusterizar movimentos e, portanto, identificar regiões com maior densidade de entrada de voos em área terminal. A avaliação foi conduzida de forma qualitativa, levando em consideração a densidade de voos nas áreas terminais em comparação com os procedimentos de aproximação disponíveis nas cartas aeronáuticas dos aeroportos. Além disso, uma análise de ruído foi realizada, indicando a presença de ruídos entre 0,06% e 6,59% em cada setor, considerados como outliers ou movimentos atípicos. Esses resultados apontaram que os clusters foram bem identificados, mantendo um conjunto significativo de dados. As principais vantagens apresentadas neste artigo foram: a utilização de um modelo para identificação automática dos setores de entrada em área terminal, que, com um único conjunto de parâmetros, demonstrou robustez nos experimentos para os diferentes aeroportos, indicando a escalabilidade computacional da solução proposta; e a possibilidade de calcular o indicador de tempo adicional em área terminal de forma mais assertiva.
The increasing global air traffic requires efficient management systems. One way to achieve that is optimizing arrival sequences, reducing flight duration, and enhancing operational safety. A method to address these requirements is a sequencing technique called Point Merge system (PM). This study analyzes the adherence to the PM at Guarulhos International Airport (GRU), aiming to understand the peak and off-peak usage times, and provide a detailed sector-by-sector analysis. Two methods were applied for trajectories classification according to their adherence to the PM: A method based on the 8th ICAO’s Key Performance Indicator (KPI08-based method), which consists of thresholding based on additional time in terminal area, and clustering by the agglomerative method using trajectories data. The experimental data were collected from: KPI08 dataset, airport movements and radar synthesis, focusing on medium-sized commercial aircraft. The total number of flights analyzed was 2427. The results obtained indicate a better performance of the KPI08-based method, which achieved a more defined representation between the classified trajectory groups. The results highlighted a behavior pattern in relation to peak and off-peak times in the use of PM in all periods analyzed. Evaluating the North and West sectors, it was possible to show that some time slots exhibited similar behaviors when using the KPI08-based method. The time slots 8h-9h and 18h-19h stand out as they show that more than 70% of flights execute the PM. The time slot 20h-22h showed similar behavior in both sectors, with approximately 50% of flights executing the PM. This work contributes by proposing the KPI08-based method, an innovative method to classify flights according to the adherence to the PM, which proved to be more accurate than the agglomerative trajectory clustering method. Additionally, this study may indicate an improvement in the arrival flow management at GRU, due to demand predictability. This predictability enables better route planning, which may result in additional fuel reduction.
The complexity and criticality of airspace control, coupled with the increasing demand for flights, highlight the need for advanced tools to optimize and secure operations. Additionally, many variables influence the air traffic scenario and, given the difficulty of analytically assessing the impact of each variable, a possible strategy is to use Machine Learning techniques to more accurately assign the weight of each factor, by processing historical data. In this context, considering that one of the variables that most influences air traffic is runway reconfiguration, this paper presents the development of a system to predict which runway will be in use at Congonhas Airport (SBSP) using a Gradient Boosting technique, called LightGBM. A classification model based on LightGBM was developed using data from the Meteorological Aerodrome Report (METAR), Terminal Aerodrome Forecast (TAF), and Weather Research and Forecasting (WRF). The model makes predictions about possible runway reconfigurations, deciding which side of the runway will be used in the coming hours. With this data, an accuracy of 98% in predictions was achieved and when testing the model on a period outside the training and test dataset, an accuracy of 88% was obtained. These results highlight the efficiency of the developed model compared to the currently used rule-based methods, which achieved approximately 81%, according to related works. This work contributes by providing a better quantification of the influence of meteorological variables on airspace operations, according to a feature importance analysis, which can support a more efficient and safer airspace activity planning.
This study aims to improve the efficiency of Brazilian air traffic by analyzing the Brazilian Airspace Control System (SISCEAB) performance indicators. The methodology used combined alternative data sources, namely BIMTRA, TATIC FLOW, and VRA, which were employed to examine the impact of different variations in taxiing times. Specifically, Additional Taxi-Out Time (KPI 02) and Additional Taxi-In Time (KPI 13) were analyzed to identify discrepancies among these data sources and determine the most precise combination. The results indicate that airport layout, gate distribution, and runway threshold selection significantly impacted taxiing times. Statistical analysis revealed substantial variations in unimpeded taxi times across different gates and runway thresholds, emphasizing optimizing operational flows. Based on these findings, integrating BIMTRA and VRA is recommended for more accurate KPI measurement. Therefore, this study contributes to implementing operational enhancements, optimizing airport operation flow, and leading to a more efficient management of Brazilian air traffic.
Escritório de Projetos
Gestão da Inovação
Performance ATM