Research Project: Energy Efficient Landing Approaches (EELA) 2020
A typical approach to a large commercial airport takes place via predefined routes. Sink and horizontal flight phases alternate. While the engine power is reduced during descent, it must be increased again and again for horizontal flight. This has several disadvantages compared to a continuous descent.
As the level of horizontal flight takes place in the lower layers of the air with a higher air density, more fuel is burned during this approach and consequently more harmful fuel gases such as CO2 are emitted. The multiple load changes of the engines lead to higher maintenance costs and cause increased noise pollution for local residents.
EELA provides sweeping solutions and helps to save the environment.
Researchers at the FernUniversität have developed an emergency landing system that helps pilots “gliding” to a landing site. An aviation magazine recognized this innovation.
"Whether due to a technical problem, a bird strike or lack of fuel: aircraft engines can always fail!" says Prof. Dr. Wolfram Schiffmann with great emphasis. The computer science professor at the FernUniversität is a passionated pilot and flight instructor in his spare time. If the aircraft engine fails completely, the emergency landing assistance system developed in its computer architecture department can provide effective support for pilots. It determines a landing site that can be reached in gliding flight and helps to steer the non-powered machine down to the ground. The "Emergency Landing Assistant" (ELA) and the "Emergency Landing Field Identification" (ELFI) have now been extended by an aircraft control component to "Safe2Land" and can now automatically land a machine even if the crew fails.
Learning to Fly:
Building an Autopilot System based on Neural Networks and Reinforcement Learning (Eckstein, Schiffmann)
This work contributes to the final goal of building an autopilot system (Neuro-Pilot) based on neural networks to follow an emergency landing trajectory in loss of thrust situations. A learning environment based on reinforcement learning shall be set up and procedures shall be developed to automatically train the system. Once set up, this environment is used to train the Neuro-Pilot which can eventually be compared to conventional proportional–integral–derivative (PID) based control algorithms. A focus will be laid on identifying the necessary learning time and the sample efficiency of such a learning approach.
Multi-Modal Image Processing Pipeline for a Reliable Emergency Landing Field Identification (Klos, Lenhardt, Klein, Schiffmann):
In the case that the pilot of an aircraft is forced to perform an emergency
landing, quick and reliable decisions about the flight path are necessary. Besides,
it is not guaranteed that an official landing field is located within reach. Therefore,
in such a situation the selection of an appropriate emergency landing field denotes a crucial task to the pilot. The choice of a suitable emergency landing field determines the damage of the aircraft, the civil population, the crew and the passengers on board. For that reason, we have developed an image processing pipeline which is capable to support the pilot in his decision about the most suitable emergency landing field, if no official – in the best case, paved – landing field is reachable. Our approach is based on satellite imagery, rasterized road maps and interpolated digital elevation grids. The chosen processing pipeline consists of eight processing steps. The results have shown that our approach is capable of a reliable selection of appropriate emergency landing fields for a certain configuration of the algorithms in our image processing pipeline.
Emergency Landing Field Recognition Based on Elevation Data Using Parallel Processing (Eckstein, Wittig, Schiffmann):
In case of a forced emergency landing situation it is crucial to identify suitable landing fields rapidly and precisely. To safely land an aircraft, various options are possible. A premium choice would be a published airfield. But there are situations where such a location is unreachable – e. g. in regions with few widely scattered airfields or if the airplane’s altitude is too low to reach a nearby airfield. In these cases, other flat terrains may be suitable to safely land the plane. If a satisfactory landing field is known, an Emergency Landing Assistant can be used to calculate an optimized route from the current position .
In this paper we present a parallel processing approach based on POSIX threads to identify suitable emergency landing fields. As input of the algorithm elevation data of a surface model is used. The landing fields are charaterized by a set of algorithm parameters which are based on different types of slope among the data points within a certain landing field. The recognized landing fields are scored according to their bumpiness measured by the variance of the altitude values and stored in a geo database for later usage in an emergency case.
Wind-aware Emergency Landing Assistant Based on Dubins Curves (Klein, Klos, Lenhardt, Schiffmann):
A total engine failure poses a major threat to passengers as well as the aircraft and requires a fast decision by the pilot. We develop an assistant system to support the pilot in this decision process. An aircraft is able to glide a certain distance without thrust power by converting the potential energy of the altitude into distance. The objective of our work is to calculate an approach route which allows the aircraft to reach a suitable landing field at an appropriate altitude. This is a non-trivial problem because of many free parameters like wind direction and wind velocity. Our solution computes an approach route with two tangents and two co-rotating circular segments. For this purpose, valid approach routes can be calculated for many general cases. The method has a constant complexity and can dispense with iterative approaches. The route is calculated entirely in the wind frame which avoids complex calculations with trochoids. The central idea is to take the wind into account by moving the target towards the wind direction.