A Highly Trustworthy Fusion Localization Method Based on Evidence Theory

To achieve highly trusted wide-area wireless localization, the on-demand access to multi-source information becomes critical. However, the current means of fusing multi-source heterogeneous information for localization is single, and the selection of weight space often relies on experience. To this end, this paper proposes a hierarchical confidence evaluation method for multi-source information, using Euclidean distance as the objective to optimize the confidence probability of spatial fusion of each localization algorithm, using Dempster combination rules to synthetically assign confidence to the basic information from different moments. In this way, a confidence evaluation model of multi-source information is constructed, and the final confidence results are obtained by the hierarchical fusion of two layers of confidence functions. The simulation results show that the proposed method can effectively fuse the localization results of RSSI fingerprint matching, Kalman filter and LLOP closed-form algorithms, and the algorithm in this paper has stronger environmental adaptability and robustness compared with fixed weight fusion.