A Software-Defined and Distributed Wi-Fi Channel-State Information Acquisition Testbed

TL;DR

This paper introduces a flexible, software-defined Wi-Fi CSI testbed using SDRs and a custom IEEE 802.11a stack, enabling real-world CSI data collection and advanced algorithm testing for positioning.

Contribution

It presents a novel distributed testbed for Wi-Fi CSI acquisition that allows real-time algorithm exploration on commercial devices.

Findings

Denoising algorithms improve CSI accuracy

CSI-based positioning is feasible with machine learning

Testbed enables real-world Wi-Fi CSI research

Abstract

We propose a software-defined testbed for Wi-Fi channel-state information (CSI) acquisition. This testbed features distributed software-defined radios (SDRs) and a custom IEEE 802.11a software stack that enables the passive collection of CSI data from commercial off-the-shelf (COTS) devices that connect to an existing Wi-Fi network. Unlike commodity Wi-Fi sniffers or channel sounders, our software-defined testbed enables a quick exploration of advanced CSI estimation algorithms in real-world scenarios from naturally-generated Wi-Fi traffic. We explore the effectiveness of two advanced algorithms that denoise CSI estimates, and we demonstrate that CSI-based positioning of COTS Wi-Fi devices with a multilayer perceptron is feasible in an indoor office/lab space in which people are moving.