Marine Wireless Sensor Network Location Method Based on Gradient Boosting Decision Tree

Abstract

Due to the influence of underwater environment, the accuracy of node localization in marine wireless sensor networks (MWSNs) is usually not high, and the adaptability of node localization algorithm is weak. A new positioning method based on gradient boosting decision tree (GBDT) is proposed. Firstly, the average positioning error is defined according to the hop count of the unknown node to the anchor node and the average connectivity degree of the network in this method. Next, the position of the unknown nodes are obtained. Meanwhile, the network topology is optimized. The free-space broadcasting model is used for power control, and signal interference and energy consumption are reduced consequently. A decision tree based on gradient is used to establish an MWSN positioning algorithm. Finally, the simulation experiment was out in Matlab. The relationship among node communication distance, node density, link loss product and positioning error was analyzed. The simulation experimental indicate the high efficiency and good adaptability of the proposed algorithm.