Study On Similarity Metric Based Localization Algorithm In Wireless Sensor Networks

Wireless sensor network emerging as a bridge connecting the virtual information world and the physical world, people can feel the feedback from the objective world, and then make the surrounding environment "observable and controllable". What the most important in WSN applications, obviously, is to locate where the messages obtained come from.Measurement technologies play an important role in most typical localization algorithms. Little error would decrease the accuracy of the localization rapidly. From the control science point of view, the tolerance of fault is weak, i.e., the robustness of these algorithms is not good. In perfect classifications, the learning algorithms use medical information, which only meet a certain rank order. Therefore, this paper attempts to locate the nodes by the similarity / dissimilarity metric, the main content consists of the following three aspects:① Briefly introducing the research background and the structure of WSN, and then to elaborate the importance of the node localization for WSN applications. Typical localization algorithms are excessively dependent on measurement technologies. Therefore,trying to locate nodes by the similarity/dissimilarity data,which used in machine learning algorithms。② Aim at the node localization problem in wireless sensor networks consisted of inexpensive sensing devices with limited resources, a novel distributed algorithm called LKNN(Localization based on K-Nearest Neighbor) is posed. The localization problem can be solved as a binary classification using the k-nearest neighbor algorithm. LKNN localizes the networks based on connectivity information(i.e., hop counts only) and beacon locations. And then a modified version of mass-spring optimization is posed to further improve the location estimation in LKNN. The simulation study shows that, the accuracy of the KNN classification is remarkably high and this supports our approach of using KNN classification for the node localization problem; mass-spring optimization addresses the border problem effectively; and compared with the DV-Hop, LKNN has a better position performance, especially in a random C-shaped placement. Further experimental results verify the effectiveness and practicality of the proposed algorithm.③ Among indoor localization algorithms based on wireless sensor networks(WSNs), RSSI methods can be interfered with a lot influencing effects and the errors of localization have a very high variance. To solve this problem, we present a novel algorithm called RSSI-CMDS(received signal strength indicator and classic MDS).The RSSIs of the nodes in WSNs are used to construct the proximity matrix. Apply classical MDS to the proximity matrix to achieve the relative map. Use the particle swarm optimization(PSO) to optimize the four arguments mode, and then transform the relative map to an absolute map based on the absolute positions of the reference nodes. Simulations and experimental results all show the proposed RSSI-MDS algorithm, which has a good performance even in the presence of random RSS fluctuations due to multi-path fading and shadowing, can meet the requirements of a low-cost indoor tracking and localization system.