Joint Communication and Sensing in OTFS-based UAV Networks

TL;DR

This work tackles precise 3D localization and tracking of a UAV swarm using OTFS-based communications by exploiting relative echo delays (RED) and Doppler shifts extracted from delay-Doppler domain channel estimates. It introduces Turbo Iterative Positioning (TIP), a two-stage algorithm: belief propagation (BP) to infer path-to-UAV associations and a gradient-descent-based refinement that jointly estimates positions and velocities, with an optional cold-start and tracking mode. The approach leverages multipath reflections within a connected network to surpass the intrinsic $c/B$ resolution limits of the communication bandwidth, and it is rigorously benchmarked against Cramér-Rao lower bounds and genie-aided versions, including tests with real-world UAV traces. Results show high localization and velocity accuracy, robustness to noise, and performance approaching theoretical bounds, highlighting the practical potential for integrated sensing in 6G-like UAV networks.

Abstract

We consider the problem of accurately localizing $N$ unmanned aerial vehicles (UAV) in 3D space where the UAVs are part of a swarm and communicate with each other through orthogonal time-frequency space (OTFS) modulated signals. The OTFS communication system operates in the delay-Doppler domain and can simultaneously provide range and velocity information about the scatterers in the channels at no additional cost. Each receiving UAV estimates the multipath wireless channel on each link formed by the line-of-sight (LoS) transmission and by the single reflections from the remaining $N-2$ UAVs. The estimated channel delay profiles are communicated to an edge server to estimate the location and velocity of the UAVs from the relative echo delay (RED) measurements between the LoS and the non-LoS paths. To accurately obtain such estimations, we propose a solution called Turbo Iterative Positioning (TIP), initialized by a belief-propagation approach. Enabling a full cold start (no prior knowledge of initial positions), the belief propagation first provides a map associating each echo to a reflecting UAV. The localization of the $N$ UAVs is then derived by iteratively alternating a gradient descent optimization and a refinement of the association maps between UAVs and echos. Given that the OTFS receivers also acquire the Doppler shifts of each path, the UAV's velocities can be sensed jointly with communication. Our numerical results, obtained also using real-world traces, show how the multipath links are beneficial to achieving very accurate position and velocity for all UAVs, even with a limited delay-Doppler resolution. The robustness of our scheme is proven by its performance approaching the Cramer-Rao bound.

Figures (8)

• Figure 1: Communicating UAVs assisted by an edge server. The UAV $u$ ($u{\in}\{h,i,j,k\}$) has position $\vec{P}_u$ and velocity $\vec{V}_u$, measured with respect to a common Cartesian coordinate system. The notation $|{\bf p}_u-{\bf p}_v|$ denotes the distance between UAV $u$ and UAV $v$, where ${\bf p}_u$ and ${\bf p}_v$ are the components of the geometric vectors $\vec{P}_u$ and $\vec{P}_v$, respectively.
• Figure 2: Scheme of the cold start algorithmic framework. An initial estimate of the maps, $\widehat{\mathlarger{\mathlarger{\boldsymbol{\mu}}}}^{(0)}$, is provided by the BP algorithm and fed to TIP. After $L$ iterations, TIP provides an estimate of the UAVs' positions, $\widehat{{\bf p}}$, and velocities, $\widehat{{\bf v}}$.
• Figure 3: Scheme of the tracking algorithmic framework. At every time step $\Delta t$, TIP is provided with a new set of channel observations $\boldsymbol{{\cal D}}$ and $\boldsymbol{{\cal V}}$. TIP also takes as input an initial estimate of the UAV's positions, obtained at the previous time step and outputs the current estimates $\widehat{{\bf p}}(t)$ and $\widehat{{\bf v}}(t)$.
• Figure 4: RMSE as a function of the number of gradient descent iterations ($I_\alpha$). Left: $L{=}0$ and $B{=}30$ MHz corresponding to $c\Delta\tau{=}10$ m. Right: $I_{\mu}{=}2$ and $B{=}3$ MHz corresponding to $c\Delta\tau{=}100$ m.
• Figure 5: RMSE on position estimates (left) and velocity estimates (right), as a function of the signal bandwidth, for varying $L$, $T_f{=}20$ ms, and $I_{\mu}{=}1$. Average over $R{=}100$ random network scenarios.

Theorems & Definitions (1)

• Remark 1