融合机器学习的WSN中RSSI高精度鲁棒定位算法OA

Wireless Sensor Network(WSN),as a core component of the Internet of Things,has broad application prospects in fields such as environmental monitoring,smart homes,and target tracking.However,in complex environments,localization algorithms based on Received Signal Strength Indication(RSSI)often suffer from insufficient positioning accuracy and robustness due to multipath effects,Non Line of Sight(NLOS)propagation,and signal attenuation.To address these challenges,an optimized RSSI-based localization algorithm for WSNs,focusing on high precision and strong robustness,is proposed.The main innovations of this method are:Ultra-Wideband(UWB)anomaly detection based on isolation forest.Through unsupervised learning,anomalous ranging data in NLOS environments are identified and eliminated,significantly reducing the impact of multipath interference on localization;Multilayer Perceptron(MLP)adaptive noise adjustment.Using a MLP to dynamically model the RSSI-distance relationship,the process noise covariance matrix of the Extended Kalman Filter(EKF)is adjusted in real time,enhancing the algorithm's adaptability to dynamic environments;and an improved EKF framework integrated with machine learning.By combining the Multi-Innovation EKF(MIEKF)with Weighted Least Squares(WLS),historical observation data are fused through a sliding window mechanism to reduce linearization error accumulation,and the localization Root Mean Square Error(RMSE)is reduced to 0.18 m.Experimental results show that in complex scenarios with weak signals and 30%NLOS proportion,the proposed method improves localization accuracy by over 25%compared to traditional RSSI-based localization methods.Moreover,through anomaly detection and dynamic noise suppression mechanisms,the localization success rate remains stable above 90%,significantly enhancing system robustness.This proposed method provides reliable technical support for high-precision WSN localization in complex environments.