|M.Sc Student||Adler Avishai|
|Subject||Real-Time Computation of Flight Paths for Emergency|
|Department||Department of Electrical Engineering||Supervisors||Professor Nahum Shimkin|
|Dr. Aharon Bar-Gill|
|Full Thesis text|
A significant part of air transportation activity is carried out by General Aviation (GA) aircraft, often flown by a Single Pilot - either in the Visual Flight Rules (SPVFR) or in the Instrument Flight Rules (SPIFR). In recent years technological advances in avionics (GPS, GIS and powerful low-weight & miniaturized computers) have reached the GA cockpit - clearing the way for safety enhancement in these flight regimes. Engine cut constitutes a typical emergency situation, requiring location of a safe-to-land strip (incl. its altitude and approach direction) and guidance towards it - in particular under SPIFR conditions.
This research addresses the capability to reach a given strip in a manner, which should enable safe landing. In an un-thrusted aircraft this means planning a three-dimensional (3D) flight-path while carefully monitoring the remaining energy reserves. Because of this limiting constraint, a dynamic model of the aircraft is adopted, instead of a kinematic one. Also, the flight-path should avoid obstacles - geographic and man-made.
We study candidate algorithms, which may allow on-board shaping of optimal approach 3D flight paths, and synthesize the most efficient solution to satisfy this objective. We start by formulating a continuous optimal-control problem, and then transform it to a discrete one. To retain the properties of the continuous problem we employ “motion flight primitives” in the discretization scheme, which are pre-defined flight segments. Finally, the discrete problem is efficiently solved by an optimal graph search (OGS) approach, thus producing a globally-optimal flight -path.
We study this method further, and suggest several extensions for the basic algorithm, which are designed for achieving real-time performance. Also, the produced flight-path has several discontinuities, caused by the discrete nature of the algorithm. These are relieved using a flight-path smoothing technique. The presented methods were tested in a simulation. The test consists of a variety of scenarios, which were designed to demonstrate the effectiveness of methods.