AI Meets Aviation: Smarter Air Traffic Control

As the aviation industry continues to grow, the complexity of managing air traffic increases. It is estimated that Europe will see an increase of 53% more flights by 2040 [1].  Ensuring safety and efficiency in the future skies is a monumental task, and air traffic controllers (ATCOs) are at the heart of this operation.

Enter machine learning (ML), a branch of artificial intelligence (AI) that’s making waves in the aviation sector. But how can ML be leveraged to both help ATCOs reduce their ever-increasing workload, while still making air travel safer and more efficient?

Understanding Air Traffic Control Workload

ATCOs are responsible for managing the flow of aircraft in and out of airports, guiding pilots during takeoff and landing, and ensuring safe distances between aircraft in the air. This job is incredibly demanding, with controllers needing to make quick decisions based on vast amounts of information. The workload can vary significantly depending on traffic volume, weather conditions, and unexpected events such as military training operations.

Machine Learning to the Rescue

Machine learning (ML) algorithms analyse data to find patterns and make predictions. In aviation, these algorithms can help predict traffic patterns, optimise flight paths, and even anticipate potential conflicts. Here are five ways ML is being applied to reduce ATC workload and enhance air traffic management (ATM):

#1 – Predicting Air Traffic Complexity

Researchers have developed various complexity metrics to better understand and manage air traffic. For instance, a study from 2011 [2] proposed a probabilistic complexity metric that helps in predicting deviations from expected flight paths. This metric supports onboard conflict detection and trajectory management, especially useful in high-density airspaces.

#2 – Optimising Flight Paths

ML models can analyse historical flight data to predict the most efficient routes for aircraft. This not only reduces fuel consumption but also minimises the risk of mid-air conflicts. An example is the use of Long Short-Term Memory (LSTM) networks to predict air traffic congestion by looking at aircraft speed vectors [3].

#3 – Clustering Aircraft Trajectories

By clustering aircraft trajectories based on control actions rather than spatial positions, ML can help identify patterns in ATC decisions. This approach was highlighted in a 2020 study [4] that used Hidden Markov Models (HMMs) to transform trajectories into sequences of control actions. Such techniques enable more precise clustering and identification of outlier trajectories, improving traffic management strategies.

#4 – Recognizing Congestion

Deep learning, a subset of ML, has been used to recognize congestion levels in ATC sectors. A 2022 study [5] modelled aircraft as nodes within a network and used complexity indices to classify congestion levels. This method provides insights into controller workload, allowing for better resource allocation and sector management.

#5 Predicting ATC Workload

Predictive models can assess ATC workload by analysing past sector operations. For example, an Evolving Graph Convolutional Network (EvolveGCN) was employed in a 2023 study [6] to predict workload under various traffic scenarios. This model incorporates human-in-the-loop simulations to enhance accuracy, providing confidence bounds for workload labels.

The Role of Explainable AI

One of the challenges in deploying ML in critical fields like aviation is the “black box” nature of many algorithms. Explainable AI (XAI) aims to make these models more transparent, helping human operators understand how decisions are made. For instance, integrating XAI into air traffic management systems can improve trust between ATCs and AI systems, facilitating real-time risk prediction and operational decision-making [7].

The Road Ahead

The integration of ML in aviation is still evolving, but the potential benefits are enormous. By reducing the workload on air traffic controllers and optimising the management of air traffic, ML can make air travel safer and more efficient. As these technologies continue to develop, we can expect even greater advancements in how we manage our skies.

References

1. “Long-term Forecast of Annual Numbers of IFR Flights up to 2040.” Available: https://www.eurocontrol.int/publication/long-term-forecast-annual-numbers-ifr-flights-2040. [Accessed: Aug. 06, 2024]
2. S. L. Brázdilová, P. Cásek, and J. Kubalčík, “Air traffic complexity for a distributed air traffic management system,” Proc. Inst. Mech. Eng. G J. Aerosp. Eng., vol. 225, no. 6, pp. 665–674, Jun. 2011.
3. L. Shi-Garrier, D. Delahaye, and N. C. Bouaynaya, “Predicting air traffic congested areas with long short-Term Memory networks,” in Fourteenth USA/Europe Air Traffic Management Research and Development Seminar (ATM2021), Sep. 2021. Available: https://enac.hal.science/hal-03344406. [Accessed: Nov. 06, 2023]
4. S. Chakrabarti and A. E. Vela, “Clustering Aircraft Trajectories According to Air Traffic Controllers’ Decisions,” presented at the 2020 IEEE/AIAA 39th Digital Avionics Systems Conference (DASC), Available: https://www.researchgate.net/publication/347043534_Clustering_Aircraft_Trajectories_According_to_Air_Traffic_Controllers’_Decisions. [Accessed: Aug. 06, 2024]
5. X. Tan, Y. Sun, W. Zeng, and Z. Quan, “Congestion Recognition of the Air Traffic Control Sector Based on Deep Active Learning,” Aerospace, vol. 9, no. 6, p. 302, Jun. 2022.
6. Y. Pang, J. Hu, C. S. Lieber, N. J. Cooke, and Y. Liu, “Air traffic controller workload level prediction using conformalized dynamical graph learning,” Advanced Engineering Informatics, vol. 57, p. 102113, Aug. 2023.
7. Y. Xie, N. Pongsakornsathien, A. Gardi, and R. Sabatini, “Explanation of Machine-Learning Solutions in Air-Traffic Management,” Aerospace, vol. 8, no. 8, p. 224, Aug. 2021.