Research On Mobility Pattern Prediction And Localization For Underwater Sensor Nodes

With the increasing proportion of marine economy in the national one,more and more attention has been paid to the exploitation and utilization of marine resources.Therefore,with the high flexibility and self-organizing ability,the sensor networks are widely used in the field of ocean.The localization technology of underwater sensor nodes helps to combine the monitoring information and location information,making the former one valuable.As a result,the localization for underwater sensor networks has been one of the research hotspots.In the marine environment,nodes move frequently,making positions changing with time.Besides,it is difficult to charge energy or replace battery for nodes in the seawater.Therefore,it brings great challenges to achieve localization for underwater nodes.The paper mainly focuses on localization algorithm for underwater mobile sensor nodes with energy-efficiency.First,establish and predict the node mobility pattern and estimate velocities with the pattern.Then,design the localization method for anchor nodes and ordinary nodes to reduce communication cost.Moreover,the particle swarm optimization localization algorithm based on dynamic search space is designed to improve localization accuracy.The main contents are as follows:(1)In the offshore environment,Gauss radial basis function is chosen as spatial basis function to construct the node mobility pattern with high temporal and spatial resolution.Besides,in order to improve the real-time performance of the mobility pattern as well as reduce data update frequency,the K-medoids method and the extended Kalman algorithm are used to predict model coefficients,laying the foundation for node localization.(2)Considering the differences between anchor nodes and ordinary nodes,the localization algorithm based on mobility pattern is designed to get nodes' location in time and reduce communication cost.Anchor nodes utilize K-medoids method and the extended Kalman algorithm to predict model coefficients and get location combining it with history position.According to the error threshold and prediction window,anchor nodes update coefficients to improve the pattern accuracy.As for ordinary nodes,the mobility pattern is computed based on reference node information and estimate location with it.The update method of reference node list is designed to adjust the information,ensuring the accuracy.(3)Aiming at the scene with requirement of high location precision,the node location optimization model is established and the particle swarm optimization localization algorithm based on dynamic search space is designed.For mobile nodes,the dynamic two-dimensional search space is designed after extending space with predictable position as the center.In this way,it improves the search efficiency and position accuracy.(4)Simulations are carried out for the localization algorithm based on mobility pattern and the particle swarm optimization localization algorithm based on dynamic search space.Results show that the proposed algorithm outperforms SLMP scheme and MP-PSO scheme in localization coverage,average position error and average communication cost.Besides,the average location error is obviously reduced after using PSO algorithm.