Sensor Mobility Prediction And Localization Algorithm In UWSNs Based On Sequence Learning

Underwater sensor networks(UWSNs)can be applied to the exploration of marine environment and underwater resources,which have broad prospects for development and have aroused a wide concern of many military departments,industry and academia in various countries around the world.In all applications of UWSNs,the information collected by underwater sensors is of practical significance only if the locations of the sensors are known.Thus researching on the localization of the sensors is fundamentally meaningful.In complex underwater environment the sensors move along with the water flow,while the research findings of the moving pattern of underwater sensors are very few,which results in the difficulty of the localization of mobile sensors.Therefore,it is of great theoretical significance and engineering value to design a suitable dynamic sensor localization algorithm for UWSNs.The main contents and innovations of this thesis are as follows:(1)The sensors localization in dynamic underwater environment was studied and a sensor mobility prediction algorithm based on sequence learning was proposed.This algorithm used the historical coordinate sequences of the sensors to predict the sensors' mobility: firstly,reconstructed the sensors' historical coordinate sequences in phase space;then used the Online-Sequential Learning Machine to train the coordinate prediction model;and then output the current coordinates of the sensors.The prediction model was refreshed with the joining of new coordinates.(2)The results of above mobility prediction and the beacon information of the anchor sensors received by the sensors were combined to localizing the sensors so that the localization accuracy was improved.We built two kinds of dynamic UWSNs models and designed extensive simulation experiments,which demonstrated the effectiveness of our method comparing with the representatively dynamic localization method SLMP.In this thesis,we proposed a sensor mobility prediction and localization method in UWSNs based on sequence learning.This method did not rely on the sensors' moving model and the computing efficiency of mobility prediction was high.It made full use of the distance information of anchor sensors so as to synthesize more accurate underwater sensors position.This thesis is helpful to improve the theoretical level and function of the existing localization of UWSNs.