Improved indoor localization algorithm based on CDCOOT-RBFNN and OPTICS-KF

In indoor wireless positioning, the complex indoor environment will lead to an increase in ranging errors, which will directly affect the accuracy of positioning. This paper proposes an indoor positioning model aimed at enhancing positioning system accuracy, comprising two stages: data preprocessing and positioning. The purpose is to address the issue of the significant impact of the ranging sequence’s initial value on the performance of the Kalman filter during the data preprocessing stage. In this paper, we propose a density clustering algorithm using Ordered Points of the Ordering points to identify the clustering structure (OPTICS) to ameliorate the effect of initial values on Kalman filtering. In view of the problem that the RBF neural network is susceptible to getting trapped in local optimum during the positioning stage, this article incorporates three algorithms: differential evolution (DE), Cauchy mutation, and Coot Optimization Algorithm (COOT) and proposes the CDCOOT optimization algorithm and refines the parameters of the RBF neural network. The experimental results show that in the data preprocessing stage, the OPTICS-KF algorithm can effectively reduce the ranging error compared to the KF algorithm. In the positioning phase, compared to the COOT algorithm, the positioning accuracy of using the CDCOOT algorithm to optimize RBFNN increased by 0.117 m, and compared to using the RBFNN algorithm alone, the positioning accuracy increased by 54%.