Browsing by Author "Koopman, Cynthia"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Open Access AI for real-time tolerance to critical flight data errors in large aircraft(AIAA, 2023-06-08) Koopman, Cynthia; Zammit-Mangion, DavidThe environment in the cockpit of large transport aircraft is currently highly complex due to an increasing amount of automation systems. This complexity can cause pilots to become less aware of how all systems work and interact. It becomes a severe issue when sensor or data failures occur, as such failures can contribute to a situation in which it is difficult for a pilot to assess what actually is happening and, possibly, where the fault originated from and how to resolve the problem. Erroneous sensor information is known to cause automation to fail or malfunction and there are several instances where such errors led to fatal accidents. This paper presents a method, based on artificial intelligence, for detecting and identifying incorrect critical flight control data in real-time. The aim of this method is to help the pilot assess the state of the aircraft and reduce the risk of confusion due to automation. A novel combination of Reinforcement Learning and an auxiliary denoising autoencoder is proposed to identify where the failures are occurring and to provide command inputs to the aircraft’s flight control and guidance systems, allowing the aircraft to perform the correct manoeuvre to counter the failure and/or to avoid or recover from flight upsets and loss of control. Tests in nominal as well as stall conditions with a partially blocked Pitot tube were conducted. These tests show that the proposed combination of Machine Learning methods creates a system to accurately detect failures (2.5s average detection time), reconstruct input data (RMSE < 6 ft/s for airspeed), and provide stable directions for the flight controls. Due to the specifically designed architecture and training schedule it is possible for the proposed system to achieve this level of performance using only a single neural network. To conclude, a comparison with the performance of the system trained without the auxiliary denoising autoencoder was made to highlight the significant advantages of the proposed architecture for learning meaningful neural connections and how this relates to creating systems with AI to improve situational awareness for pilots and execute appropriate automatic manoeuvres to successfully counter the effect of sensor failures.Item Open Access Increasing predictability in the response of an AI-assisted stall recovery system in complex stall conditions by expanding the knowledge-base of AI(IOP Publishing, 2024-03-13) Koopman, Cynthia; Zammit-Mangion, DavidThe environment in the cockpit of commercial aircraft is becoming increasingly complex due to the introduction of automation systems. This complexity is especially evident when malfunctions take place, making it difficult for pilots to comprehend the interconnectedness of the systems and potentially leading to loss of control. This paper investigates a novel method for creating an Artificial Intelligence-based stall recovery assistant using Reinforcement Learning by training the agent to generate a stall and subsequently recover from it. This enables training in a large training space with a simple reward function, where the agent has the ability to develop a deep understanding of the environment. Tests show that the agent is able to recover from stall at a variety of altitudes while experiencing unreliable airspeed information originating from a blocked Pitot tube system and with a better response than all baseline agents. The results indicate that restricting AI is not always necessary and, further, that too many restrictions can lead to a system that learns only shallow features, causing it to be unreliable in unforeseen circumstances.