基于EGWO-LSTM-LSM算法的TOF定位研究

Research on TOF Localization Based on EGWO-LSTM-LSM Algorithm

针对最小二乘法(Least Squares Method,LSM)对于超宽带(Ultra Wide Band,UWB)中飞行时间(Time of Flight,TOF)法在非视距(Non Line of Sight,NLoS)下定位精度低下的问题,提出了基于扩展灰狼算法(Extended Gray Wolf Algorithm,EGWO)优化后的长短期记忆(Long Short-Term Memory,LSTM)网络改进LSM定位算法(EGWO-LSTM-LSM)。采用LSTM及改进的EGWO建立最优测距误差预测模型,根据预测结果构造权重矩阵,在LSM上加权计算,并添加测距误差校正项,以改进LSM实现静态定位,并结合卡尔曼滤波器(Kalman Filter,KF)实现动态定位追踪。仿真结果表明,EGWO-LSTM预测准确率达98.857%,EGWO-LSTM-LSM将二维和三维位置误差分别稳定控制在10~25 mm,进一步提升了TOF定位精度。

To solve the problem of low localization accuracy of Time of Flight(TOF)in Ultra Wide Band(UWB)under Non Line of Sight(NLoS)by Least Squares Method(LSM),an Extended Gray Wolf Algorithm(EGWO)is proposed to optimize the Long Short-Term Memory(LSTM)network that is then used to improve the LSM localization algorithm(EGWO-LSTM-LSM).First,LSTM and EGWO are used to establish the optimal range error prediction model,then the weight matrix is constructed according to the prediction results,weighted calculation is made on the LSM,and the range error correction item is added.Finally,the improved LSM is used to achieve static positioning,and the Kalman Filter(KF)is combined to achieve dynamic positioning and tracking.The simulation results show that the prediction accuracy of EGWO-LSTM is 98.857%,and the two-dimensional and three-dimensional position errors by EGWO-LSTM-LSM are stably controlled within 10~25 mm respectively,which further improve the positioning accuracy of TOF.