WiFi/IMU Fusion Indoor Positioning Algorithm Based on LSTM

In the field of indoor positioning-based location information services, traditional single-source indoor positioning methods exhibit inherent limitations, while filter-based multi-source fusion positioning approaches are inadequate to meet the high-precision positioning requirements of mobile devices. To tackle this issue, this paper presents a fusion indoor positioning algorithm based on the Long Short-Term Memory (LSTM) network, which integrates WiFi fingerprint positioning and IMU (Inertial Measurement Unit) inertial positioning. The proposed LSTM-based fusion algorithm employs the LSTM network and self-attention mechanism as its core components. It performs heterogeneous synchronization and displacement feature extraction on the results from WiFi fingerprint positioning and IMU inertial positioning. By leveraging the learning capabilities of the network model, the algorithm fuses the two single-source positioning techniques to achieve high-precision indoor positioning. Experimental results demonstrate that the average positioning error of the proposed fusion positioning algorithm is 1.22 meters. This represents improvements of 24.7%, 17.6%, and 11.6% compared to the fusion positioning algorithms based on the Kalman Filter (KF), Particle Filter (PF), and Generalized Regression Neural Network (GRNN), respectively.