Title: AI-based Emergency Landing Support for Pilots

10:10 ~ 10:50 a.m., 21 February 2024 (Wednesday), The Sukosol Hotel, Bangkok, Thailand

Wolfram Schiffmann and Steffen Flämig

University of Hagen

Abstract

A loss of thrust constitutes a crucial issue for every pilot and requires fast as well as target-oriented action. The most urgent decision is to select an emergency landing site. Even if there is enough excess altitude to fly to a registered airport, the decision can be difficult if there are several airports within range. However, it becomes even more difficult if there are no official airports within range. The pilot then has to decide quickly on a suitable emergency landing option. Since the field of vision from the cockpit is very limited, deciding on the best emergency landing opportunity is made even more difficult. One solution is to use geo-data to determine in advance a database of landing options for different types of aircraft. During the flight, a smart pilot assistance system can provide the pilot with permanent support for the aircraft in use. In the event of an engine failure, nearby landing options can be displayed immediately, greatly facilitating and accelerating the decision-making process. The keynote will discuss AI-based techniques to solve the problem for support of the pilots and detail our own method for identifying emergency landing sites. The core idea of our approach is to search for landable areas using elevation data, satellite imagery and road maps. Of course, the characteristics of the specific aircraft are used as a basis for selecting the most suitable landing fields for it. A suitable emergency landing field must be as flat as possible, free of obstacles and have as low slope as possible. Therefore, we use high resolution elevation data to find such areas and label them as landable. Those areas are also supplemented by information from the open street map project. In total, the sketched procedure provides enough information to generate training sets for the equally geo-registered satellite images. This training data will later be used to train convolutional neural networks (CNN), which can then detect emergency landing fields under real-time conditions. To bring an aircraft with engine failure in gliding flight to an emergency landing field, we have developed the Safe2land technology, which is also briefly introduced. There are also a variety of other AI-based methods like reinforcement learning to find optimized routes from the emergency point to the selected landing site. Safe2Land can even autonomously land the aircraft after the pilot selected the destination. Emergency situations can thus be managed smarter using AI-based approaches and help to save the lives of the passengers and the crew.

Bio

Wolfram Schiffmann studied electrical and computer engineering at the Universities of Kaiserslautern and Saarbrücken. From 1983 to 2000 he worked as a scientific assistant at the computer science department of the University of Koblenz. He simultaneously worked as a lecturer at the University of Saarbrücken (1983-1992). Wolfram Schiffmann received a Diploma in Electrical Engineering in 1983, a Doctoral degree in 1986 both at the University of Saarbrücken, and a Habilitation degree in computer engineering at the University of Koblenz in 1998. From 2000 to 2022 he got a full professorship of Computer Engineering at the University of Hagen. His research interests were in future computer architectures for parallel processing, artifical neural networks, cluster- and gridcomputing, scheduling algorithms for task graphs and immersive Imaging. During the last ten years Prof. Schiffmann moved his research focus on emergency landing assistance and energy-efficient flying. As a private pilot and flight instructor he is familiar with the complexity of emergency management. Together with his team, which also includes Steffen Flämig as another privat pilot, Schiffmann developed various solutions that enable even autopiloted emergency landings in case of engine failure and thus contribute to autonomous flying, e.g. for airtaxi applications.

Steffen Flämig was born in Zwickau, Saxony in 1966. After graduating from the Polytechnic High School, he completed two vocational training courses, as a maintenance mechanic and as an electrician. He has been self-employed in the field of renewable energies since 2004. He studied computer science at the University in Hagen and completed his studies in 2020 with a thesis on the topic of “emergency landing field identification based on artificial neural networks". After that he joint the team of Prof. Schiffmann and works currently as a doctoral student at the University in Hagen, where he conducts research on the subject of flight assistance systems.