Research on Underwater Acoustic Location Algorithm Based on Multilayer Extreme Learning Machine

摘要

In view of underwater acoustic positioning model which is complex, highly nonlinear and solving difficult, as well as the low accuracy of underwater acoustic gradient measurement, an underwater acoustic positioning algorithm based on multilayer extreme learning machine(MELM) networks was proposed to achieve the high accuracy of underwater target positioning. In this paper, the first layer ELM network in the algorithm is proposed to realize the initial location of sound source, solve its approximate location, and eliminate invalid data. The second layer ELM network calibrates the underwater acoustic velocity relying on the initial position depth, and uses the filtered data to achieve higher positioning accuracy for the target. Simulation results show that the underwater acoustic positioning algorithm using multilayer ELM networks has higher positioning accuracy and stronger fault tolerance.