A Dynamic Adaptive Positioning Method based on Differential Signal Feature Map

摘要

This paper proposes a dynamic adaptive positioning method based on differential signal feature map of cellular network which aims at how to accurately calculate the non-line-of-sight(NLOS) error in dynamic environment. The main innovations include the accurate compensation method of NLOS error based on machine learning and the dynamic adaptive updating method of the differential signal feature map. We build the differential signal feature map and the NLOS error distribution map based on the signal feature after establishing the 3D geographical environment by wireless insite fully considering physical and architectural features. We proposed TD-GA-BPNN algorithm to train the model related to the differential signal characteristics and NLOS distribution as the NLOS compensation method. While the dynamic adaptive updating method adopts a four branch tree partition technique which can dynamically update the feature map by the information of the sparse virtual stations. Through experimental verification, this method eventually achieve high-precision accuracy which can be less than 2m in the actual dynamic environment.