NEURAL NETWORK ANALYSIS OF PILOT MANEUVER DURING LANDING PHASE

摘要

Pilot control at the visual approach has been modeled using neural networks. The quality of the constructed pilot model depends on the given inputs and learning scheme because measured data includes noise and uncertain inputs. In order to cope with these uncertainties, a new learning scheme is proposed in this paper. Using the proposed scheme, pilot controls are analyzed for different flight conditions, such as dayight and no wind/gusty. The contribution ratios and sensitivity analysis results show clear differences between the flight conditions.