Joint Data Registration And Fusion For Multiple Sensors Under Moving Platforms

In recent years, sensor networks are widely employed in target tracking, traffic management, military surveillance, network control, and remote monitoring. However, communication errors such as random time delays and packet loss will degrade the fusion performance because of unreliable network channels. Especially in a battlefield environment, this phenomenon is more serious due to the existence of electromagnetic interference. In addition, the inherent systematic sensor biases will deteriorate the precision of received measurement data. If sensor biases are too large, the system performance will be heavily degraded, even unstable. In order to solve these problems, this paper proposed a joint data registration and fusion algorithm with imperfect communication channels for multiple moving platforms. This algorithm can realize the estimate of target state and sensor biases simultaneously for nonlinear systems with one-step random delay and multiple packet dropouts. The main research contents are as follows:Joint data registration and fusion algorithm with imperfect communication channels for multiple moving platforms of non-maneuvering target tracking system was proposed. Multiple-moving sensors are used to form a network for target tracking. Assume that a fusion center is placed on the ground. Measurements from multiple sensors on mobile platforms are transmitted to the fusion center for further processing. In order to establish the measurement model of the system with one-step random delay and multiple packet dropouts, two random variables of Bernoulli distribution are used to describe the arrival conditions of the received measurement data. State augmentation approach is adopted here to convert the original system with random parameters that of Bernoulli distribution into a stochastic parameterized augmented system. For this augmented system, expectation maximum (EM) based on the extended Kalman smoother with random parameters (EM-pEKS) algorithm is proposed to estimate target state and sensor biases simultaneously. According to the projection theory, extended Kalman filter with random parameters (pEKF) is proposed, this algorithm can realize the linear minimum variance filtering estimation. Extended Kalman smoother with random parameters (pEKS) by using the maximum likelihood estimation (MLE) is derived to reduce the estimation error of filtering algorithm.Joint data registration and fusion algorithm with imperfect communication channels for multiple moving platforms of maneuvering target tracking system was proposed. Firstly, the maneuvering target tracking model with one-step random delay and multiple packet dropouts is formulated. Secondly, State augmentation approach is adopted here to convert the original system with random parameters that of Bernoulli distribution into a stochastic parameterized augmented system. In order to realize joint data registration and fusion for multiple moving platforms for the maneuvering target tracking system, EM based on the interacting multiple model smoother with stochastic parameters (EM-pSIMM) algorithm is derived to estimate target state and sensor biases simultaneously. According to the projection theory, interacting multiple model filtering algorithm with stochastic parameters (pIMM) is proposed to solve the state estimation problem of maneuvering target. In order to reduce the filtering error of pIMM algorithm, interacting multiple model smoother with stochastic parameters (pSIMM) is propose by using MLE method.Posterior Cramer-Rao bound (PCRB) is developed here to evaluate the joint data registration and fusion algorithm with imperfect communication channels for multiple moving platforms. The PCRB for non-maneuvering target tracking system and maneuvering target tracking system is derived respectively. Simulation results show that the proposed algorithm is effective in sensor registration and target tracking for nonlinear system with one-step random delay and multiple packet dropouts.