Convolutional Bidirectional Long Short-Term Memory-based Dynamic Noise Adaptive Distributed Collaborative Maneuvering Target Tracking

A distributed cooperative target tracking algorithm based on a convolutional bidirectional long short-term memory (ConvBiLSTM) neural network is proposed to address the nonnegligible and changing observation noise caused by previous passive maneuvering target tracking.The algorithm improves the time difference of arrival/frequency difference of arrival (TDOA/ FDOA) measurement accuracy by utilizing the ConvBiLSTM neural network to correct the sensor observations to adapt to the dynamically changing observation environment, as well as combining with the weighted two-step least squares method to reduce the initial estimation error.The experimental results show that the algorithm can be used to estimate the target position and velocity more accurately in maneuvering target tracking environments with large observation errors, and at the same time, improves the stability of target tracking.