Artificial intelligence (AI) based tools are increasingly playing a critical role in air traffic control (ATC) by enhancing the efficiency and safety of air travel. AI techniques like machine learning assist systems in learning from data and improving performance in ATC. This article deliberates on the importance of various AI-based tools in ATC.
Image Credit: Gorodenkoff /Shutterstock
ATC and Its Challenges
Air traffic services (ATS) refer to controlling and managing the air activities of aircraft. ATS includes alerting services, flight information services, and ATC services. The ATC services aim to maintain and accelerate the orderly and safe operation of air traffic and avoid collisions between aircraft and obstacles and between aircraft.
ATC tasks are conventionally performed by human air traffic controllers, who grasp the altitude and the position of the aircraft based on the pilot's position report in flight and the scheduled flight plan to ensure a safe and orderly flight. However, automation has been gradually introduced into ATC as the number of flights has been rising continuously, significantly straining the system.
Currently, ATC has many challenges in managing the air traffic flow for flights, with unpredictable weather conditions being the most significant among all challenges, which lead to flight delays. Additionally, the risk of potential conflict between air flights/accidents has increased substantially owing to the sheer volume of air traffic.
Thus, careful consideration of several factors is necessary while deciding flight routes. AI can effectively address many challenges in ATC as this technology can recommend optimal routes, predict weather patterns, and identify potential conflicts, thereby improving efficiency and safety.
Importance of AI
The development of AI techniques has positively impacted ATC. Using AI-based tools, intelligent ATC systems that enable a richer analysis of existing air traffic problems can be developed. AI tools also facilitate the development of intelligent conflict detection and resolution module systems for flight conflict detection to ensure safer flights.
The aspects of ATC that can be covered by AI-based tools include situational awareness for air traffic controllers, support for wireless communication, computer model creation, human-machine interfaces, traffic forecasting, abnormal traffic flow predictions, and air traffic flow management.
AI techniques have been applied to automate the entire air traffic controller's function instead of focusing on a single aspect of the controller's work. Similarly, methods from the fields of AI and qualitative physics have been combined to understand the effects of aircraft performance on the ATC actions of the controller.
AI methods in ATC facilitated human-machine interaction, which led to the development of expert systems for ATC. Many research groups, including the Rand Corporation research team and Lincoln Laboratory Group, have studied distributed expert systems for planning and control. An expert system has been developed and used in aircraft gates for cost assignments.
Decision support systems with expert systems can be utilized for addressing airline station operation problems. An air traffic management expert system was developed as an accessory tool to assist air traffic controllers with rescheduling. A study introduced entropy-based efficiency calculations and investigated how these calculations, coupled with AI methods, could be utilized for ATC.
Similarly, distributed planning and problem-solving have been proven to an effective and reliable ATC methods. This includes the implementation and design of a distributed planning system/a location-centered collaborative planning system for a distributed ATC system. Air traffic controllers have to provide conflict resolution upon the detection of potential conflicts.
Recently, AI has been proposed to support decision-making in ATC. For instance, a knowledge-based conflict resolution process has been proposed that allows the resolution of predictive conflicts according to the controller practices, including multiple degrees of freedom blending and prioritization of resolution strategies to achieve separation.
In a recent study, an AI system has been built as a digital assistant to support air traffic controllers in resolving potential conflicts. This proposed system contained two primary components, including an AI agent that utilized reinforcement learning (RL) and an intelligent interaction conflict solution that acquired the preferences of air traffic controllers. Results demonstrated the effectiveness of the system in conflict resolution strategies proposing conflict resolution strategies.
Artificial neural networks have been employed for ATC automation, with a back-propagation network being specifically utilized for making intelligent decisions. A self-enforcing network was developed to address the issue of runway direction selection in airports. The measured data for various periods for wind condition forecasting was given to the network, which then provided suggestions for selecting suitable operation directions.
In many studies, multi-agent-based models were used to represent physical resources for ATC, such as runways, airports, and control centers, or the tasks that had to be performed. Despite the extensive utility of AI-based tools in ATC, these tools put forward higher requirements on the ATC system's input data. Additionally, system users have to possess more professional domain knowledge to use these tools.
Moreover, the opaqueness of AI-based tools increases the difficulties of using them. This issue can be addressed by incorporating explainable AI strategies in AI-based systems to bolster transparency and trust among human operators.
AI Techniques in ATC
Different techniques have been used for air traffic prediction, including deep learning, long short-term memory (LSTM) model, recurrent three-dimensional convolutional neural network (R-3DCNN) model, deep multi-agent RL method, recurrent neural network (RNN), extreme learning machines (ELM), and bidirectional long short-term memory (Bi-LSTM).
Machine learning algorithms and neural networks excel in trajectory prediction to avoid air traffic conflicts. RL has proven effective in multiple aspects of ATC. In Brazil, the air holding problem module significantly improved traffic performance, which indicated the effectiveness of RL as a decision aid.
AI algorithms like RNN are utilized to forecast future scenarios/make early predictions from historical flight and weather data to efficiently manage air traffic and avoid flight accidents. Bi-LSTM and ELM, coupled with deep learning networks, have been used to improve the accuracy and capacity of air traffic management.
Deep multi-agent RL method and analytical air-ground data link model can be employed to resolve aircraft conflicts and control air traffic, and predict future ATC communications, respectively. The R-3DCNN model has been proposed to predict air traffic effectively with higher accuracy.
Air traffic flow with different flight levels can be predicted using the LSTM model. In recent years, disruptions have frequently occurred in airline operation control systems. This problem can be addressed using the multi-agent system/multi-agent system for disruption management in airline operation control.
The accuracy of aircraft coordinates can be predicted using deep learning, inception modules, and LSTM. Automatic dependent surveillance-broadcast technology is used with these algorithms for accurate prediction.
Recent Developments
A paper recently published in Aerospace demonstrated the safety of automatic speech recognition (ASR) applications in ATC operational environments. Specifically, the paper described the safety assessment performed in SESAR2020 project PJ.10-W2-96 ASR on ASR technology implemented for ATC centers.
The safety assessment process involved defining design requirements for ASR applications in degraded, abnormal, and normal modes of ATC operations. Overall, eight functional hazards were identified by analyzing four use cases. The assessment was supported by bottom-up and top-down modeling and the analysis of the hazard causes to obtain system design requirements for mitigating the hazards.
The assessment displayed that the eight ASR functional hazards have no substantial impact on overall air traffic management safety. Mitigations were derived from operational requirements for each hazard to ensure acceptable air traffic controller performance without degrading the controller's task execution.
Additionally, the assessment of realizing the specified design requirements was supported by the evidence obtained from two real-time simulations with pre-industrial ASR prototypes in en-route and approach operational environments. These simulations validated the hypotheses that ASR decreases controllers’ workload and improves situational awareness. Thus, the paper demonstrated that the use of ASR does not increase safety risks, and thus, the technology is ready for industrialization.
To summarize, AI-based tools are revolutionizing ATC by paving the way for advanced automation and improved management of air traffic challenges.
References and Further Reading
Helmke, H., Dokic, J., Hartikainen, P., Ohneiser, O., Lasheras, R. G. (2023). Ensuring Safety for Artificial-Intelligence-Based Automatic Speech Recognition in Air Traffic Control Environment. Aerospace, 10(11), 941. DOI: 10.3390/aerospace10110941, https://www.mdpi.com/2226-4310/10/11/941
Xie, Y., Pongsakornsathien, N., Gardi, A., Sabatini, R. (2021). Explanation of Machine-Learning Solutions in Air-Traffic Management. Aerospace, 8(8), 224. DOI: 10.3390/aerospace8080224, https://www.mdpi.com/2226-4310/8/8/224
Tang, J., Liu, G., Pan, Q. (2022). Review on artificial intelligence techniques for improving representative air traffic management capability. Journal of Systems Engineering and Electronics, 33(5), 1123-1134. DOI: 10.23919/JSEE.2022.000109, https://ieeexplore.ieee.org/abstract/document/9940157
Abdillah, R. E., Moenaf, H., Rasyid, L. F., Achmad, S., Sutoyo, R. (2024). Implementation of Artificial Intelligence on Air Traffic Control-A Systematic Literature Review. 2024 18th International Conference on Ubiquitous Information Management and Communication (IMCOM), 1-7. DOI: 10.1109/IMCOM60618.2024.10418350, https://ieeexplore.ieee.org/abstract/document/10418350
Sangeetha, V., Andrews, S. K., Rajavarman, V. N. (2022). Air traffic control using machine learning and artificial neural network. Journal of Positive School Psychology, 6(2), 4737-4746. https://www.journalppw.com/index.php/jpsp/article/view/2929