Visual Assistant: A Vision-Assisted System for Generating Wi-Fi Device Maps

Abstract

Device-free sensing systems have the potential to enhance human-computer interaction applications, but their effectiveness is highly dependent on the precise placement of devices, which typically requires a significant amount of labor input. To alleviate this labor-intensive process, we present the Visual Assistant, an inventive system engineered to autonomously generate environmental maps and accurately identify the locations of Wi-Fi infrastructure for sensing. Our proposed approach employs commercially available devices in smart homes including the Wi-Fi router, the camera, and the robotic vacuum cleaner. The Visual Assistant consists of the following two stages: i) the camera mapping stage synthesizes the coordinate systems with respect to the camera and the Simultaneous Localization and Mapping (SLAM) robots; ii) the Wi-Fi mapping stage proposes the hyperbolic model to inversely pinpoint the positions of undisclosed Wi-Fi infrastructure based on human movement and the camera. The innovation of this paper lies in the proposed hyperbolic model, which theoretically reveals how to infer Wi-Fi location based on known user trajectories. We conduct comprehensive experiments to verify the performance of the Visual Assistant. The results suggest our system can yield centimeter-level and decimeter-level mapping accuracy in camera mapping and Wi-Fi mapping stages, respectively, marking a substantial stride in the development of device-free sensing systems.