Wireless Sensor Network Node Self-positioning Technology

Automatic localization of every sensor is a key enabling technology of wireless sensor networks. With a network of thousands of nodes, it is unlikely that the position of each node can be pre-determined. Thus, estimating nodes'position is needed after deploying. Localization approaches based on Multidimensional Scaleing could work efficiently on range-based or range-free, so this paper will be focused on wireless sensor networks localization approaches based on Multidimensional Scaling.First of all, the research status of wireless location techniques and positioning systems for wireless sensor networks are summarized based on the study of plenty of related literatures. The criterion of performance evaluation and the taxonomy for wireless sensor networks localization systems and algorithsms are described. The principles and characteristics of recent representative localization approaches are also discussed and presented.Then, this paper introduced sensor networks localization technology based on classical Multidimensional Scaling. On the basis of analyses on classical Multidimensional Scaling algorithm based on centroid reference point, this paper derives the Multidimensional Scaling algorithm which using network nodes as reference point. Simulation result shows that classical Multidimensional Scaling algorithm based on centroid reference point is more precise than other one.Furthermore, this paper proposed modified iteration Multidimensional Scaling algorithm. It utilized node's communication radius and hop count information between nodes. Node's initial position matrix is computed by classical Multidimensional Scaling and cost function's form.is also improved. Simulation result proves that the modified approach is effective and has better performance than conventional one.At last, this paper introduced relaxation solution into node position estimation and proposes a fast algorithm. This algorithm did not expect minimizing whole cost function on each step and minimized one term on each step. Simulation result shows that the fast algorithm efficiently reduced computation complexity and improved computing efficiency.