Research On Target Tracking Algorithms In Wireless Sensor Networks Based On Kalman Filter

With the rapid development of micro-electro-mechanical systems (MEMS), wireless communications and embedded microprocessor technologies, wireless sen-sor networks (WSNs) are more and more widely being applied in the world. Target tracking is one of the most fundamental applications of wireless sensor networks. Due to the limited computational capability, the energy and bandwidth resource constraints of the sensors, the dynamic changes in the network topology and the real-time requirements of the tracking applications, the algorithms for target track-ing in wireless sensor networks are required to solve the above problems. The target tracking algorithms in wireless sensor networks remain to be perfected further.Serval algorithms for target tracking in wireless sensor networks are proposed based on the Kalman filtering theory in this dissertation. The main works and con-tributions are summarized as follows:1. An extended H∞filter-based algorithm is proposed for target tracking in wireless sensor networks for the case of having no knowledge of the statistics of sen-sor measurement noises and the maneuvering properties of the target to be tracked. And also the extended H∞filter and the existence condition of the filter are derived using the Krein space Kalman filtering theory.2. A new approach is proposed based on the combination of the maximum likelihood estimation and the Kalman filter for target tracking in wireless sensor networks with multiplicative and additive noisy measurement model. First the non-linear measurements are converted into a linear observation of the target state, and the converted noise covariances are also evaluated using the maximum likelihood es-timation. Then the converted measurements and the associated covariance are used by the Kalman filter to update the target state estimate. In this approach the Newton iterative method is utilized to solve this maximum likelihood estimation problem, and the initial values are assigned using the Kalman predictor. 3. Since sensor nodes are energy constraints, on the basis of the above result two sensor selection methods are proposed based on the Fisher information matri-ces of the maximum likelihood estimation and the Kalman filter, respectively. The tracking approaches integrated with the two methods can be used to efficiently and accurately track the target.4. Since the leaders are required to spend huge energy on receiving measure-ments from all its members and estimating the target state in the extended Kalman filter based target tracking algorithms, a new tracking approach based on the sequen-tial extended Kalman filter is proposed, in which at each time instant all sensors in the tracking cluster sequentially estimate the target state and transmit their results to the next sensor. Every sensor estimates the target state only using the received data and its measurement. The above procedure repeats until the last sensor of the tracking cluster, and the last sensor obtains the tracking result for that time instant.In conclusion, this dissertation focuses on the target tracking problem in wire-less sensor networks. The obtained results have not only important theoretic values, but also extensive practical values.