Research On Localization Algorithms For Wireless Sensor Networks Based On Adaptive Kalman Filters

In many applications of wireless sensor networks,such as environmental moni-toring and indoor positioning,wireless sensor nodes need to be deployed in a com-plex environment.The wireless sensor network is a self-organizing network,and there is no fixed communication link between nodes which return the collected data to the sink node in a multi-hop manner.For randomly deployed wireless sensor net-works,the exact location of the data-acquisition node is often unknown.Therefore,localization technology is one of the hotspots in the research field of wireless sensor networks.However,due to the complexity of the deployment environment of wireless sensor networks,it is still a challenging task to obtain accurate node locations.This paper focuses on the localization of wireless sensor networks in complex environ-ments and studies three problems of the adaptive Kalman filter based sensor network positioning refinement,robust noise-adaptive Kalman filtering localization for sensor networks,and optimization of noise-adaptive Kalman filtering fingerprint positioning for sensor networks in multipath environments.(1)In consideration of the wireless sensor network being a kind of net-work with limited resources such as energy,computing power,communication bandwidth,cost,etc.,typically,the most easily acquired received signal strength between nodes is used to determine the distance between nodes in wireless sensor networks,then,according to the relative distance between the nodes,the relative position between the nodes in the network is obtained.However,the measurement noise of the received signal strength will make the positioning accuracy of this method become poor.In order to reduce the noise's impact on the positioning accuracy,the Kalman filter is utilized to refine the localization results.Due to the fact that the statistical properties of the noise are often unknown or time-varying in the actual environment,we first proposed a multidimensional scaling localization refinement algorithm for wireless sensor networks based on the existing adaptive extended Kalman filter.Then,in order to further improve the refinement effect,we deduced a new adaptive unscented Kalman filter from the adaptive extended Kalman filter,and put forward a multidimensional scaling localization refinement al-gorithm with higher precision for wireless sensor networks based on the new filter.The previous algorithm has a relatively small computational complexity,while the latter algorithm has a higher positioning accuracy.At the same time,we also proposed a kind of adaptive fingerprint Kalman filter for refining wireless sensor network fingerprint positioning results under the noise envi-ronment.Extensive experimental results show that no matter whether the noise is known or unknown,whether it changes with time or not,our proposed algorithms are able to better improve the refinement effect of traditional Kalman filter to the wireless sensor network localization.(2)Kalman filter is used to estimate the node's position,speed and other state information through the process model and measurement model of wire-less sensor nodes.In general,due to some external and internal interference factors,the process model and the measurement model will contain random variables representing the process noise and the measurement noise.At present,if the statistical properties of these two kinds of noise simultaneously change with time,the estimation result of the existing adaptive Kalman filter will have a large deviation,and even the filter is not working to lose robust-ness.In order to improve the accuracy and robustness of localization algo-rithm based on adaptive Kalman filter for wireless sensor networks,we first propose a robust adaptive extended Kalman filter.Then,on the basis of this robust filter,we further derive a robust adaptive unscented Kalman filter with higher accuracy.In addition,the robustness of these two new filters is strictly proved in theory.In the case that both process noise and measurement noise are time-varying,the results of a large number of simulation experiments show that two proposed Kalman filters can ensure that the nodes' positions in wireless sensor networks are robustly and accurately estimated.(3)At present,the research of fingerprint localization in wireless sensor networks based on adaptive Kalman filter is focused on how to improve the accuracy and robustness of the algorithm,but there are few researches on how to obtain the optimal node position estimation in the environments of noise and multipath.Therefore,firstly,the adaptive Kalman filter and the mul-ti-objective evolutionary algorithm are employed to optimize the fingerprint localization results in noise environments.Because the adaptive Kalman filter can only filter out part measurement noise of the received signal strength,there is still a small noise in the filtered received signal strength,which will affect the effect of fingerprint localization refinement.In order to minimize the influence of residual noise on the localization results,we also use the multi-objective evolutionary algorithm to further optimize the fingerprint lo-calization results in the noise environment.The existing RSSI(received signal strength indication)distance measurement model in multipath environments will cause that the weight of the fingerprint for estimating the node location is not matched with the position weight during the optimization,so in order to get a better fingerprint positioning optimization result,we derive a mul-ti-channel weighted RSSI ranging model according to the relationship be-tween signal strength and distance in multipath environments.Extensive ex-perimental results show that the newly established multi-objective evolutio-nary model and the multi-channel weighted RSSI ranging model can make the adaptive Kalman filtering fingerprint localization obtain better position esti-mation results in the noise and multipath environments.