基于翘曲高斯过程与JAYA-KELM的指纹定位算法OA

In allusion to the challenges of fingerprint database expansion errors and insufficient online positioning accuracy in WiFi fingerprinting localization,a novel fingerprint localization algorithm that integrates warped gaussian process regression(WGPR)with JAYA-optimized kernel extreme learning machine(KELM)is proposed.In the offline phase,the MaxMean is used to select high-information access points,thereby reducing redundancy and enhancing feature quality.WGPR is then introduced to map the received signal strength(RSS)from a non-Gaussian space into a latent Gaussian space by means of nonlinear warping function.A composite kernel function is incorporated to strengthen the modeling capability for multi-scale signal variations and construct a highly robust fingerprint database.In the online phase,the JAYA algorithm is used to optimize the hyperparameters of the KELM model,enhancing its nonlinear fitting capability and generalization performance,and ultimately achieving high-precision mapping from RSS to position coordinates.The experimental results in indoor environments demonstrate that,in comparison with the other algorithms,the proposed WGPR-JAYA-KELM model can realize an average positioning error of 0.988 1 m under complex conditions,significantly improving both localization accuracy and robustness.