Large aircraft takeoff path optimization out of terrain challenging airports
Bertrand Masson · UNSWorks (UNSW Sydney) · 2022
Large Aircraft operators conducting regular passenger transport must satisfy regulatory requirements such as considering engine failure at takeoff at the worst point of the takeoff roll. Such constraints can severely restrict commercial payload, for example in airports surrounded by high terrain. However, current methods for the analysis of takeoff paths are largely manual and require significant time to yield an allowable payload. These current methods may induce higher than necessary engine thrust levels that increase engine maintenance cost and drive engine designers to design costlier higher thrust engines. In contrast research in robotics, particularly on unmanned aerial vehicles, has created a wealth of automatic path planning techniques that enable high speed online guidance and navigation and provide solutions to multi-constraint multi-cost path planning problems. Building on robotic path planning techniques, in this research we address this complex problem in order to provide the aircraft performance engineer with automated methods of generating high quality escape paths that combine complex aircraft kinematics, terrain models and regulatory constraints in a unified mathematical framework. Specifically, we develop a general approach to the problem, formalize our method, extending it out to engine failure on climb-out as well as in missed approach. In order to test our approach over the complete range of world runways, we start off by creating a runway classification using machine learning classification methods, yielding four runway categories: Open, Coastal, Valley and Channel. Classification results show that close to 30% of all world runways present a challenge to current manual based methods, therefore justifying the creation of our automated path generation. Runways are then selected out of these four categories to thoroughly test the model under varying conditions. Results from our extensive evaluation of its implementation on a sample of real-world airports yields superior payload capability, safer paths based on a unique set of metrics, and high repeatability of the path generator. Furthermore, engine failure simulations in climbout and missed approach appear to challenge common regulatory beliefs, therefore paving the way to regulatory changes and safer flight operations.