Research On Localization Of Wireless Sensor Networks

Wireless sensor network (WSN) has attracted much attention from industrial and academiccircles since its appearance, because traditional information-access technology of sensors arechanging from single-mode to intelligent and networked mode with the development of electronicscience, information science, computer networking technology, and etc. The value of data sensed bynodes in many WSN applications (e.g. forest-fire prevention, ecological environmental monitoring,etc.) is discounted even meaningless without position information, therefore, localization is one ofthe key technologies of WSN. In practice, most nodes are randomly distributed and traditionallocalization technologies (e.g. GPS) are not suitable for localization of each node due to the powerlimitation of WSN; hence it is necessary to study the technology of WSN localization.Firstly, the dissertation introduces the basic conceptions, characteristics, applications, and keytechnologies of WSN; secondly, the problem of WSN localization, especially the range-basedlocalization algorithms and their improved algorithms are studied; At last, it proposes a range-basedrefinement algorithm and a range-based localization algorithm for WSN. The contributions of thisdissertation are as following:The dissertation presents a refinement algorithm, RAMP, which is based on the thought ofmagnetic pole and distance-hops contradiction. On the basis of analyzing errors of classiclocalization model, the algorithm identifies “error nodes” and “effective nodes” in each iteration;then chooses two nodes with minimum efficient node relative deviations in the neighbors set oferror nodes, and gets the intersection points of two circles whose centers are the positions of abovetwo nodes and radiuses are the range distances from above two nodes to the error node respectively;finally, it chooses the position with the minimum error as the refinement position of error node fromits current position and above two intersection points. Simulation results show that RAMP canreduces average relative localization error of classic WSN localization model with more than7%,and can reduces the average number of “error nodes” with more than40%. Hence it can improvethe localization accuracy of network efficiently.The dissertation presents a distributed localization algorithm of WSN based on weighted twodimensional logarithmic search, W-TDLS. The algorithm determines the weight coefficient of eachnode in objective function according to the type of node; then search the best estimated point (BEP)by weighted two dimensional logarithmic search; finally, it achieves the localization result of thisiteration and the type of node in next iteration by comparing the BEP with localization results ofprevious iterations.At last, the dissertation presents a refinement algorithm for W-TDLS based on the thought of searh and RAMP. The algorithm includes a sub-algorithm for normal nodes and a sub-algorithm forlonely nodes. In an iteration, the two sub-algorithms both determine the weight coefficient of eachnode in objective function by error level, and they search the BEP with different objective functionsand achieve the refinement results with different rules. Simulation results show that, compared withthe some current complex algorithms, DSOCP+NCSG, the average relative localization error ofW-TDLS is18%lower than that of DSOCP, and the average relative localization error ofW-TDLS+R-TDLS is12%lower than that of DSOCP+NCSG.