The Study Of Distributed Estimation Based On Optimal Power Allocation

The wireless sensor networks is a kind of monitoring network, which is a combination of sensor technology, wireless communication and computer networks, and becomes a research hotspot at home and abroad in recent decades. It has a broad application in civilian and military fields, such as target tracking and localization, remote sensing and military surveillance. Since the sensor node itself has some limitations on power, transmission bandwidth and storage capacity, which makes the data transmission efficiency and the network life decrease. In order to avoid wasting the band width and the energy as well as reducing the efficiency of collection of information, the data fusion technology plays an important role.There are a lot of difficulties on researching the data fusion:the optimal power allocation problem of multi-sensor system, the impact of channel estimation error on distributed fusion. In view of studying the existing problem, main jobs completed are as follow:1. In most distributed fusion algorithms, the measurement noises in different sensors are often assumed to be uncorrelated, but in practical occasions the assumption may not be met and the measurement noises are often cross-correlated between sensors. Therefore, in the case of cross-correlated measurement noises, by taking linear transformation of the raw measurement of each sensor, the optimal distributed fusion algorithms are proposed. We find the distributed fusion can achieve the performance of the centralized one under the sufficient condition that the transformation matrix is of full column rank.2. Based on the power constraints in WSNs, we study the optimal power allocation problem for scalar and vector signal. We consider that the phase synchronization is imperfect at the fusion center. The residual phase error is molded by the Tikhonov probability distribution function. We propose the global optimal power allocation algorithm, and drive the closed-form solutions for power coefficient of the sensor node. Monte Carlo simulations are carried out to verify the performance of the proposed methods. Simulation results show that the proposed scheme outperforms the uniform power allocation.3. To the best of our knowledge, we note that previous studies on distributed estimation are based on the common assumption that the perfect channel state information is available at the FC. However, this assumption is not realistic due to channel estimation error, which degrades the system performance. We focus on the effect of the imperfect channel state information upon the distributed estimation and the power allocation between the training data and transmission data. Extensive simulations are carried out in terms of both the linear minimum mean square error and the optimal weighted least squares fusion rules.