Localization Algorithm Based On Constrained Optimization In Wireless Sensor Networks

Wireless Sensor Networks (WSN) can be designed and developed because of development in sensor technology, Micro-Electro-Mechanism System (MEMS), modern network and wireless communication technology. It has been considered as one of the most important technologies in 21st century due to its bright future in application. More and more researchers focus on wireless sensor network. As a most important support technology in Wireless Sensor Networks (WSN), node localization can greatly influence the performance of WSN.This paper focused on the localization of nodes in WSN. The algorithms based on constrained optimization are analyzed and enhanced. The concrete content is as follows:1. In wireless sensor networks, connections between nodes can be represented as constraints on the node positions. The localization can be abstracted to constrained optimization model. Sector-Convex is put forward based on convex, in that the relative angle and distance between nodes are used to make constraint condition into sector. It makes the feasible sets smaller to reduce error and improve accuracy in the process.2. Besides above, two hop constraints are also used in the Sector-Convex. Fan-shaped location model is influenced by multi-hop neighbors. Estimated location area is smaller to reduce error in the process. The simulation results are that the improvement in accuracy will be more in Sector-Convex than Convex in the same condition, and enhancement is higher with more connections. Lower consistency of nodes and anchors is used in Sector-Convex. 3. In order to resolve low accuracy problem of constrained optimization with measuring distance, SQ-NLP is developed which is based on constrained optimization. The nonlinear distance equations between nodes are considered as tight constraints of feasible solution, and more efficient algorithm SQP is used to solve the node location. SQ-NLP algorithm can enhance the localization accuracy more efficiently than variable-Convex in the same condition. Also the location could be done well in SQ-NLP with lower consistency anchors or nodes.