Design of Bluetooth Positioning and Identification Algorithm for Automotive PEPS System

本文提出了一种基于差分加权K近邻算法的蓝牙PEPS车内外高精度识别系统。通过Dempster-Shafer (DS)融合方法，将双终端算法与传统单终端算法的结果融合，显著提高了车内外识别的准确率。研究构建了蓝牙信标RSSI指纹数据库，并设计了信念Hellinger距离作为新的证据融合方法。实验使用了NRF52832蓝牙开发板，并结合加权K近邻、朴素贝叶斯和分类回归树等方法对RSSI信号进行识别。经过验证测试，DS融合算法的识别准确率超过97%，溢出距离测试准确率接近100%。研究结果表明，改进的DS证据理论能够有效提升多源数据信息融合的性能，为车内外识别系统提供了可靠的解决方案。This paper presents a high-precision identification system for in-vehicle and out-of-vehicle scenarios based on a differential weighted K-nearest neighbors algorithm for Bluetooth PEPS (Passive Entry Passive Start). By employing the Dempster-Shafer (DS) fusion method, the results of the dual-terminal algorithm are integrated with those of the traditional single-terminal algorithm, significantly improving the accuracy of in-vehicle and out-of-vehicle recognition. The study constructs a Bluetooth beacon RSSI fingerprint database and designs the Hellinger distance as a new evidence fusion method. Experiments were conducted using the NRF52832 Bluetooth development board, combined with weighted K-nearest neighbors, Naive Bayes, and classification and regression trees to recognize RSSI signals. Validation tests showed that the recognition accuracy of the DS fusion algorithm exceeded 97%, with an overflow distance test accuracy approaching 100%. The results indicate that the improved DS evidence theory can effectively enhance the performance of multi-source data information fusion, providing a reliable solution for the in-vehicle and out-of-vehicle identification system.