Modeling of the Visual Approach to Landing Using Neural Networks and Fuzzy Supervisory Control

摘要: During the visual approach to landing of a x ed wing aircraft, human pilot bases control and timing subsequent maneuvers mainly on out-the-window view, as there is not su cient time read all instruments. The skill making smooth soft landings acquired through experience. Research has been done identify most important features in scene (cues) for two phases landing: glide slope tracking are maneuver. Using simulator real ight data, neural networks have trained both mimic pilot’s based cues available. By using operator neuron transfer functions, transparent model obtained. Fuzzy supervisory proposed couple thus provide insight decision process with respect initiation. generally considered one demanding [1]. combination high workload, having interpret scene, initiation executing those maneuvers, risks inherent lowaltitude ight, makes this di cult learn new pilots. Real and/or simulated experience indispensable obtain maintain skills, performance feedback thought greatly improve learning e ciency [1, 2]. However, pilots cannot explain what they look at or how make their decisions even training methods consistent. research presented paper focuses nding uses, analysis data. A method construct which takes generates longitudinal actions during landing. This numerical data from by itself however merely used verify correspondence between model. Of main interest structure parameters resulting model, i.e., driving inputs, internal relations thresholds, these give (subconscious) behavior. knowledge gained ireconstruction mindi would be useful evaluation pilots: if we know experienced use available landings, insights can taught