Optimized Deep Networks for Indoor Position and Localization Based on IEEE P802.11az

With the exponential growth of mobile devices and wireless networks, the IEEE 802.11az standard has emerged as a leading technology, offering distinct advantages in localization and positioning tasks compared to other wireless solutions. This standard boasts superior accuracy, enhanced scalability, robust security features, and reduced energy consumption. This research investigates localization and positioning through experimentation with simulated datasets based on the IEEE 802.11az standard. We propose a deep convolutional neural network (CNN) architecture for positioning and localization tasks. Specifically, we simulate a simplified three-dimensional (3D) map and generate a synthetic dataset of channel impulse response (CIR) fingerprints by leveraging the IEEE P802.11az standard. Our optimized network architecture then predicts the location of a station (STA) based on a subset of the fingerprints, enabling the precise prediction of the exact position within this virtual indoor environment. Simulation results demonstrate that our proposed architecture outperforms the baseline model, achieving up to 93.04% accuracy and 0.92 root mean square error (RMSE), compared with 86.68% accuracy and 1.63 RMSE of the baseline.