Consensus State Estimation Algorithm Based On Wireless Sensor Networks

Wireless sensor networks (WSNs), as an emerging technology, has drawn theworldwide increasing attention of the military departments, industry and academia.One basic applications is to cooperatively sense the physical information of the interestthrough wireless transmission and information interaction, in order to provide neces-sary support for the real-time monitoring and immediate processing of the event in thearea of interest.The perception accuracy of the wireless sensor will be infuenced by the ambientnoise, the sensed data exists a large discrepancy among the sensor displayed nearby.In order to obtain more accurate information, a large amount of sensors are generallydeployed in the monitoring regional. So WSNs also faced with data redundancy andcorrelation problems between the diferent sensors. Moreover, the sensor nodes havelimited energy and usually deployed in the remote or hard-to-reach areas. It is difcultto recharge or change the power source. These disadvantages have brought about greatchallenges for the state estimation in the WSNs, such as the the issues of inaccurateperception data, inconsistent and redundant data of the nodes. Besides, how to reducethe power consumption of the WSNs is another signifcant issue for the informationinteraction process.Based on the above considerations, we adopt the strategy of consensus estimationto cover the problems of inaccurate sensed data and information discrepancy. Dis-tributed information processing is employed, coupled with the event-triggered and se-quential waked up mechanisms, in order to reduce the energy consumption. Firstly, thedistributed estimation has been studied in order to improve the data accuracy and makea reduction in discrepancy between the sensors. For the sake of energy saving, we con- sider an event-driven transmission scheme based on the Kullback-Leibler divergence(KL divergence) to reduce the data transmission rate; and then apply an sequentialwaked up mechanism to reduce the waiting time of the sensor nodes.Inordertominimizetheestimateerrors, weadoptthedistributedestimationbasedon above schemes respectively to analyze its unbiasedness and optimality and presentthe design guideline of the optimal flters. Simulation studies are presented to validatethe performance of the the distributed optimal consensus flter based on the event-driven transmission scheme, through a maneuvering target in2-D space governed bythe Singer model. Last but not the least, with the aid of industrial wireless network, weimplement the sequential estimation algorithm on the hot strip mill, in order to realizethe continuous monitoring and estimating of the steel bar’s temperature at the entranceof the rough rolled.