Decimeter-Level WiFi Tracking in Real-Time

Abstract

This paper presents DeTrack, a tracking system that can continuously trace WiFi objects at decimeter-level in real-time. To enable this, we make three main proposals. The first one is a super-resolution localization scheme that combines compressed sensing and expectation-maximization algorithms to iteratively resolve multi-path, which realizes better resolution compared against traditional MUSIC. The second one is a customized particle filter that takes advantage of WiFi signals and the geometric nature of AOA estimates to properly update location states and particle weights. Finally, an SVD-based multipacket fusion is employed to reinforce the signal space and improve tracking efficiency at the same time. A prototype is built using only commercial WiFi NICs. Extensive experiments demonstrate that DeTrack achieves an 80th percentile localization accuracy of 0.9 meters and a median latency of around 90 milliseconds. As a result, DeTrack is looking to benefit a wide range of applications, e.g., indoor navigation, intelligent logistics, and smart cities.