A Highly Reliable Aircraft Search And Rescue Positioning Technology Under Extreme Conditions

Automatic dependent surveillance-broadcast(ADS-B)technology is a promising technology in the next generation air traffic control(ATC).Compared with traditional radar,ADS-B systems have the advantages of high positioning accuracy,large data update rate,low system structure and low maintenance cost.However,the existing ADS-B system has two major issues.On the one hand,in certain areas,ADS-B receivers are sparsely deployed,mainly limited by terrain(especially in marine areas).In addition,ADS-B receivers in certain areas lack maintenance,and faulty equipment may experience errors when parsing ADS-B messages.The absence or malfunction of the ADS-B receiver affects the entire system’s positioning of the aircraft.On the other hand,the biggest drawback of the ADS-B system is the openness of its links,which poses a risk of data tampering.In certain extreme situations where an aircraft malfunctions or is hijacked,relying on tampered coordinate data to rescue the aircraft is considered an "invalid rescue".If aircraft coordinates are obtained through complex and time-consuming satellite positioning and other means,the best rescue time will be missed.Regarding the above two issues,fully consider the "reliability" of ADS-B data.Then,combined with deep learning algorithms,predict the coordinates of the aircraft.Specifically,firstly,data visualization is carried out on the ADS-B dataset of "tens of millions" in order to deeply mine the ADS-B dataset,eliminate unreliable data,and identify highly reliable data.Secondly,a scheme is proposed to optimize data by combining particle filter and Kalman filter.It provides stable and reliable data for subsequent neural network training.Thirdly,the constructed Inception-LSTM neural network is used for aircraft positioning.The Inception neural network is responsible for mining spatial features in the data,and the LSTM neural network is responsible for mining temporal features in the data.Finally,a neural network with N-Inception-LSTM parallel structure is constructed.The inspiration for parallel structure comes from the internal structure of Inception neural network.N parallel Inception-LSTM neural networks can expand the width of the network and expand the data perception domain.The simulation results show that the improved Kalman filter scheme significantly eliminates most of the noise interference in the received signal strength of ADS-B.In addition,the larger the parallel number N and the longer the continuous time slice of the N-Inception-LSTM,the higher the prediction accuracy.When N is 2 and the continuous time slice length is 64,the average prediction error of N-Inception-LSTM is within 10 kilometers.Its performance is superior to Inception-LSTM without parallel structures and Inception without composite structures.It is concluded that the proposed method sacrifices a certain degree of prediction accuracy in exchange for absolute reliability of the predicted data.This method is interdependent with traditional ADS-B systems.In most normal situations,high-precision,high refresh rate,and low-cost positioning can be achieved through traditional ADS-B systems.In suspected dangerous situations,secure and reliable positioning can be achieved through the new scheme proposed in this paper.Reliable aircraft coordinates can help relevant departments quickly determine aircraft coordinates,reduce inference time,and obtain more rescue time.