GPS/UWB Tightly Coupled Vehicle Cooperative Positioning Based on AOO-CNN- BiGRU-Attention Model

Abstract

Accurate relative positioning is essential for the deployment of an intelligent transportation system. However, in complex environments such as urban canyons and tunnels, the global positioning system (GPS) signals are often blocked or interrupted, resulting in decreased or invalid positioning accuracy. To meet the demand for accurate vehicle positioning in complex environments of urban roads, this article proposes a deep learning model for GPS pseudo-range and Doppler shift prediction based on the fusion of the animated oat optimization (AOO), a convolutional neural network (CNN), a bidirectional gated recurrent unit (BiGRU), and an attention mechanism. CNN is applied to capture spatiotemporal features from the input sequence, while BiGRU explores the long-term dependencies in the data. The attention assigns varying weights according to the importance of input data, enabling the model to focus more effectively on critical parts. To improve predictive accuracy, the AOO algorithm is employed for hyperparameter optimization. Then, the predicted GPS pseudo-range and Doppler shift are used for GPS/ultrawide band (UWB) tightly coupled cooperative positioning by utilizing the characteristics of UWB technology that can provide high-precision ranging information. The results of the experiment show that the proposed fusion model improves the relative positioning accuracy by 13%, 29%, 33%, and 50% over CNN-BiGRU-Attention, CNN-BiGRU, BiGRU, and GRU models, respectively, during a GPS signal loss-of-lock environment, which significantly enhances the stability of vehicle positioning in complex environments.