Indoor Road Positioning Based on RWKV-TS

To address the challenges of low accuracy and susceptibility to environmental interference in traditional positioning technologies, particularly in indoor scenarios such as underground parking garages, this study proposes an indoor positioning method based on the RWKV-TS network. This method integrates Inertial Measurement Unit (IMU) data, non-visual semantic landmarks, and Wi-Fi fingerprints to achieve precise positioning. The RWKV-TS network, with its enhanced time-series processing architecture, effectively captures long-term dependencies in the data. It also offers advantages in terms of low time complexity and memory usage, making it well-suited for real-time applications on edge devices. Experiments conducted in the underground parking garages of two shopping malls demonstrate that the model achieves road recognition accuracy rates of 0.931 and 0.868, with Mean Absolute Errors (MAE) of 4.245 m and 4.171 m, respectively. The performance of the model significantly exceeds that of traditional deep learning models, meeting the practical requirements for applications such as smart parking guidance.