Research On Target Tracking Algorithms For Asynchronous Underwater Sensor Networks

Underwater sensor networks (UWSN) have broad application prospects in marine environment monitoring, undersea resources exploration, disaster prevention, assisted navigation, intrusion detection and target tracking, et.al. UWSN has become one research hotspot concerned by research institutions and scholars. UWSN (?) communicates using underwater acoustic signals, and sensor nodes in UWSN have a wide and sparse distribution. Thus, different sensor nodes can hardly obtain synchronous measurements of the same instant’s state of a target, in the intrusion detection and target tracking application. The conventional target tracking algorithms which are based on synchronous measurements will not be applicable in UWSN.Owing to the asynchronism of different sensors’measurements in UWSN, the target tracking algorithms based on asynchronous measurements in asynchronous UWSN are researched in this study. The details can be described as follows:(1) By researching on the target tracking problem in asynchronous UWSN based on filtering, asynchronous sequential filtering algorithms based on predicted estimates are proposed. Firstly, the concept of the predicted estimate is utilized to combine the measurements of different instants’ target states, to estimate one instant’s state of the target. Then, the asynchronous sequential filtering process based on predicted estimates is derived, by reference to the sequential filtering algorithm. Lastly, the asynchronous sequential Kalman filtering algorithm and the asynchronous sequential Particle filtering algorithm are proposed, based on the proposed asynchronous sequential filtering process.(2) By researching on the target tracking problem in asynchronous UWSN based on smoothing, an asynchronous filtering algorithm based on the fixed-point smoothing is proposed. The proposed algorithm discards the conventional algorithm thought which tries to combine the measurements of different instants’target states to directly estimate one instant’s state of the target. Only one measurement of each state is utilized in the forward filtering process, to compute the partial posterior estimate of a state by the proposed algorithm. Then, the partial posterior estimates are smoothed by the fixed-point smoother in the forward smoothing process, to obtain the completed posterior estimate of the states.(3) By researching on the problem that measurement bias and Non-Gaussian random measurement noise exist in sensor nodes of asynchronous UWSN, a robust asynchronous filtering algorithm with measurement bias estimation is proposed. Firstly, the Gaussian distribution with non-zero mean and dynamic covariance is utilized to model a sensor’s measurement bias and random measurement noise. Secondly, the Normal-inverse-Wishart distribution is employed to represent the parameters of the noise model, and the Variational Bayesian approximation method is utilized to update the parameters of the Normal-inverse-Wishart distribution. Lastly, the asynchronous filtering algorithm based on the fixed-point smoothing is employed to estimate the target’s states and the measurement biases iteratively and robustly.(4) By researching on the problem that precise sensor locations are unknown or varying locally in asynchronous UWSN, an asynchronous filtering algorithm with simultaneous sensor localization and target tracking is proposed. Firstly, the Gaussian distribution with static process transition matrix is utilized to model the dynamics of one sensor’s location. Then, the augmented state vector which comprises of the target’s state vector and one sensor’s location vector is utilized in the forward filtering process. Lastly, the partial posterior estimates of the target’s states are smoothed by the fixed-point smoothing algorithm, to obtain the completed posterior estimate of the target’s states.