الفهرس | Only 14 pages are availabe for public view |
Abstract Node localization is one of the essential and supporting requirements to most appli- cations of wireless sensor networks (WSNs). When dealing with a network with a large number of randomly deployed nodes, localizing nodes, when possible, is an expensive and time consuming task. Moreover, in many applications, nodes can move autonomously, thus positions need to be tracked as time evolves. This research proposes a clustered localization approach for WSNs based on Second order Cone Programming (SOCP). The proposed approach divides the large network into smaller sub networks. For each cluster, the cluster head formulates the localiza- tion problem as a SOCP problem, and this requires all the data needed for problem formulation (distance measurements and anchor positions) to be available to the cluster head. The cluster head solves the SOCP problem as a global minimization over the entire cluster to get positions of the cluster sensor nodes. Because of noisy distance measurements and weak relaxation of SOCP, most sensors of each cluster do not get positioned accurately (specially on the border), and some others are not positioned. To enhance localization accuracy, a cluster level refinement step is used. Each clus- ter solves the network localization problem using Gauss-Newton (iterative least-squares) approach. The initial position guess for the Gauss-Newton optimization is the position drawn from the preprocessor SOCP solver which is close to the global solution. Hence Gauss-Newton optimization converges rapidly to the global optimum in few iterations, and get a refined accurate position estimates for the cluster nodes. The proposed approach controls the cluster size, and hence it scales well for large networks of thousands of nodes and provides a considerable reduction in computation time while yielding good localization accuracy. The performance of the proposed approach is evaluated using Castalia (open-source simulator based on OMNET++ simulation environment). We measure the performance metrics: localization error, computation time, and problem size (number of variables and constraints); and study the effect of varying network size, cluster size, anchor percentage, communication radio range, and noise values added to distances measurements on the performance metrics. |