Air Traffic Safety Issues Based On Neural Network Technology Research

This paper generally describes problems of traffic safety about the civil aviation and multiple UAVs (Unmanned Air Vehicles) formation flight, sums up basic research directions for the air traffic safety, generally describes some problems about formation flight of UAV, and introduces SWARM plan of the U.S. navy specially. To differ from the concept of multiple UAVs formation flight, this paper presents the freedom formation flight, studies the track prediction and collision avoiding between two vehicles of multiple UAVs formation flight, and presents continuous fuzzy control based on neural network. It gives one-step-prediction of assumed tracks by neural network. In this study, Matlab toolbox of linear neural network is used to predict beeline and curve track, the results indicate it's effectiveness. Before applying the compensatory fuzzy neural network, we establish seven-level fuzzy rule combining theory knowledge, the vehicles' variety ratio of relative distance and angle of relative position are dealt with seven-level fuzzy rule, and trained by compensatory fuzzy neural network. After 178 steps' training, Outputs of the network reaches the goal precision. In the chapter 5 the trained fuzzy rule is confirmed by the data adding random noise, the result shows that compensatory fuzzy neural network can respond the trend of theory variety. In the end the paper presents the further research direction.