Multistatic Sensing of Passive Targets Using 6G Cellular Infrastructure

TL;DR

It is shown that human sized target can be sensed with high accuracy and sub-meter positioning errors in a practical indoor deployment scenario and resolution, coverage and position uncertainty for practical indoor deployments are analyzed.

Abstract

Sensing using cellular infrastructure may be one of the defining feature of sixth generation (6G) wireless systems. Wideband 6G communication channels operating at higher frequency bands (upper mmWave bands) are better modeled using clustered geometric channel models. In this paper, we propose methods for detection of passive targets and estimating their position using communication deployment without any assistance from the target. A novel AI architecture called CsiSenseNet is developed for this purpose. We analyze the resolution, coverage and position uncertainty for practical indoor deployments. Using the proposed method, we show that human sized target can be sensed with high accuracy and sub-meter positioning errors in a practical indoor deployment scenario.

Figures (7)

• Figure 1: Sensing of a passive object in an indoor wireless deployment. The passive object creates shadow regions which in turn creates perturbation in the communication link which can be exploited towards sensing.
• Figure 2: Three different deployment scenarios with varied number of links, $L\in \{1,2,3\}$, are considered.
• Figure 3: AI pipeline for target detection and position estimation considering scenario-3 ($L= 3$). Input to the model is a 2D-CSI frame, $\mathbf{H} \in \mathbb{C}^{L N_{\scriptsize \hbox{r}} \times N_b}$. With $L=3,N_r=8,N_b=7$, and separating real and imaginary values into different channels, we have the 2D-CSI frame dimension of $(24\times7\times2)$. Both target detection and position estimation share the same network except at the last two layers.
• Figure 4: Accuracy versus target size for different deployment scenarios of Fig \ref{['fig:deployment']}.
• Figure 5: Coverage of the proposed AI detector. (a) Coverage for deployment scenario-1 ($L=3$) with target size $\sigma=0.5~\hbox{m}$ (b) Coverage for deployment scenario-1 ($L=3$) with target size $\sigma=0.8~\hbox{m}$ (c) Coverage for deployment scenario-3 ($L=1$) with target size $\sigma=0.8~\hbox{m}$.