SIM-IPS: Stacked Intelligent Metasurface-Based Indoor Positioning System

In this paper, we introduce a stacked intelligent metasurface-based indoor positioning system (SIM-IPS), which is a novel IPS developed using an SIM, a transmissive metasurface with multiple layers of independently controlled meta-atoms. These layers function analogously to the hidden layers in a deep neural network (DNN) and enable advanced beamforming in the electromagnetic (EM) wave domain, thus improving the control of the power distribution at the receiver plane. The SIM-IPS setup includes a single Wi-Fi access point (AP) dedicated to localization, which transmits all of its power through the SIM for beamforming. This approach utilizes two distinct configurations types for improved performance: zone identification, which determines the zone where the user is located, and precise localization, which uses random radiation patterns to pinpoint the exact location of the user within the identified zone. We employ the gradient descent algorithm and the alternating optimization (AO) technique to optimize the phase shifts of meta-atoms across different SIM layers for zone configuration. The resulting radiation patterns are stored in a database. During the online phase of the IPS, the AP cycles through the available configurations, and the user provides received signal strength indicator (RSSI) measurements to estimate their location. To ensure high localization accuracy, we optimize the selection of configurations and fine-tune system parameters to achieve good performance. Simulations demonstrate a substantial increase in power concentration at target locations within specific zones by incorporating additional SIM layers. This refined power distribution enables precise differentiation between the target zone and surrounding areas, significantly increasing localization accuracy. Furthermore, we present a detailed comparison of localization accuracy improvements across varying numbers of zones and SIM layers, emphasizing the critical importance of parameter optimization for enhancing the overall performance of SIM-IPS. Specifically, we achieve optimal accuracy with an 8 zone system and 5 layers, attaining a localization accuracy of 1.72m and 1.66m by considering continuous-valued and discrete-valued phase shifts, respectively.