A Cooperative Positioning Method Considering Nonlinear Optimization of Relative Measurement Model

As information science continues to evolve, cooperative positioning technology has emerged as the dominant robust positioning algorithm. This approach enhances positioning precision by disseminating data through wireless communication. However, the present research on cooperative positioning algorithms indicates that the nonlinear optimization problem of relative measurements between vehicles in dynamic and complex environments still presents significant challenges. In this paper, a deep learning-based cooperative positioning optimization method is proposed, which uses deep learning to deal with the non-linear characteristics of the relative measurement model, and finally fuses the results of single-vehicle positioning and multi-vehicle cooperative positioning to obtain more accurate positioning results. This paper compares the cooperative positioning method based on deep learning networks (CP-DNN) with the single-vehicle positioning algorithm based on Kalman filtering (SVP-KF) and the cooperative positioning algorithm based on extended Kalman filtering (CP-EKF), using the Prescan simulation platform as a basis for comparison. The experimental results demonstrate that, in comparison with SVP-KF, the positioning performance of CP-DNN has been enhanced by approximately 43%. In comparison to CP-EKF, CP-DNN has demonstrated an improvement in its positioning performance of approximately 22%. CP-DNN method is an effective means of addressing the nonlinear characteristics of relative measurement models and significantly improving positioning accuracy.