WiFineTrack: Enabling Fine-Grained Position Tracking Using Commodity WiFi

With the development of integrated sensing, communication and computation (ISCC), wireless sensing technology is turning a WiFi device into a special sensor. Specifically, the current research focuses on extending the inherent communication attributes of ubiquitous WiFi signals to achieve indoor position tracking. However, previous works based on geometric positioning like triangulation are limited by the number of antennas and bandwidth, and can only achieve decimeter-level positioning accuracy. To this end, we present WiFineTrack, enabling fine-grained position tracking based on Channel State Information (CSI). Instead of representing the continuous absolute positions directly, WiFineTrack first characterizes the relative motion traces by mapping the target’s movements in relation to its prior position. Then the model correlates the angle-based triangulation with the derived relative motion traces to reconstruct the absolute motion traces. Thus, the model shifts the tracking task from traditional absolute geometry to relative length variation. WiFineTrack designs an elaborate denoising framework for tracking refinement, including phase correcting based on the bidirectional transmission mechanism and accumulated error suppression based on the Extended Kalman Filter (EKF). We implement WiFineTrack using off-the-shelf WiFi devices, and empirical results show that WiFineTrack can achieve an accuracy of 10.45 cm under diverse environment conditions.