UWB positioning in indoor environments using a hybrid AI-optimized fingerprint fusion algorithm

Abstract Non-line-of-sight (NLOS) propagation critically degrades indoor positioning accuracy. To address this, we propose the IALA-HKELM-AIEKF fingerprint positioning algorithm. To correct relative clock deviations between devices, the double-sided two-way ranging (DS-TWR) algorithm is employed to obtain high-precision ranging values as fingerprint features. Aiming at the complex nonlinear relationship between distance and coordinates, a hybrid kernel extreme learning machine (HKELM) is used to construct a position estimation model to lay the foundation for high-precision positioning. Meanwhile, recognizing the influence of model parameters on accuracy, the improved artificial lemming algorithm (IALA) with powerful global optimization capability is introduced to optimize the HKELM, which significantly enhances the model’s ability to deal with complex nonlinear problems. Further, in order to suppress the random errors introduced by NLOS, the output coordinates of the model are dynamically corrected using the adaptive iterative extended Kalman filter (AIEKF). Experiments show that the fusion algorithm algorithmic positioning accuracy can reach about 5 cm in NLOS environments, with a maximum reduction of 82.22% in the average positioning error compared to the comparison algorithm in this paper, confirming significant improvements in positioning accuracy.