Robust Wi-Fi Sensing With Multi-Link Integration and CSI Recovery in Congested Network Environments

In modern wireless networks, Wi-Fi technology serves a dual role — supporting both data communication and environmental sensing. This dual functionality has positioned Wi-Fi sensing as a key enabler of Integrated Sensing and Communication (ISAC), with broad applications across the Internet of Things (IoT). However, this integration poses a significant challenge to Wi-Fi sensing, as data transmission for communication interferes with sensing data collection and degrades the overall performance of sensing models. In this work, we propose a multi-link-based CSI sampling and integration framework combined with a CSI data loss recovery preprocessing method to enhance network resource efficiency, mitigate data loss, and improve the robustness of Wi-Fi sensing models. The proposed preprocessing method utilizes a Context Encoder-based inpainting approach to effectively restore lost CSI data, ensuring reliable system performance even under high-loss conditions. Experimental results show that our approach consistently outperforms existing methods, achieving higher data recovery accuracy and maintaining stable performance despite increasing loss rates.