Dynamic Joint Communications and Sensing Precoding Design: A Lyapunov Approach

Abolfazl Zakeri, Member, IEEE, Nhan Thanh Nguyen, Member, IEEE, Ahmed Alkhateeb, Senior Member, IEEE, Markku Juntti, Fellow, IEEE · 2025

Read Dynamic Joint Communications and Sensing Precoding Design: A Lyapunov Approach by Abolfazl Zakeri, Member, IEEE, Nhan Thanh Nguyen, Member, IEEE, Ahmed Alkhateeb, Senior Member, IEEE, Markku Juntti, Fellow, IEEE (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-14054.txt.

Introduction

Abolfazl Zakeri, Nhan Thanh Nguyen, Ahmed Alkhateeb, and Markku Juntti design wireless beams that must do two jobs at once: carry messages and sense the surroundings. In such joint systems, the beam patterns for the two jobs need steady adjustment as conditions change. The team frames this as a long-run control problem.

Their goal is to keep the sensing signal strong on average while holding the communication quality above a floor and staying within a power budget. Using Lyapunov optimization, they turn the long-run problem into a series of single-slot problems. They then solve those with an approximation method, and also give a closed-form solution based on zero-forcing. In computer tests, both methods beat a conventional baseline design, the authors report.

Catalog record

CREATORAbolfazl Zakeri, Member, IEEE, Nhan Thanh Nguyen, Member, IEEE, Ahmed Alkhateeb, Senior Member, IEEE, Markku Juntti, Fellow, IEEE
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-14054

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Dynamic Joint Communications and Sensing Precoding Design: A Lyapunov Approach Abolfazl Zakeri, Member, IEEE, Nhan Thanh Nguyen, Member, IEEE, Ahmed Alkhateeb, Senior Member, IEEE, Markku Juntti, Fellow, IEEE Abstract This letter proposes a dynamic joint communications and sensing (JCAS) framework to adaptively design dedicated sensing and communications precoders. We first formulate a stochastic control problem to maximize the long-term average signal-to-noise ratio for sensing, subject to a minimum average communications signal-to-interference-plus-noise ratio requirement and a power budget. Using Lyapunov optimization, specifically the drift-plus-penalty method, we cast the problem into a sequence of per-slot non-convex problems. To solve these problems, we develop a successive convex approximation method. Additionally, we derive a closed-form solution to the per-slot problems based on the notion of zero-forcing. Numerical evaluations demonstrate the efficacy of the proposed methods and highlight their superiority compared to a baseline method based on conventional design. I Introduction The coexistence of communications and sensing has been increasingly attracting attention from academia and industry due to its profound potential for resource efficiency, low cost, and hardware convergence [ 1 ] . Existing studies in joint communications and sensing (JCAS) can be broadly categorized into radar-centric, communication-centric, and joint design approaches [ 2 ] . Several interesting relevant problems were addressed, such as waveform design [ 3 ] and transmit beamforming design [ 4 ] , under various performance metrics. Notably, these include the signal-to-clutter-plus-noise ratio [ 5 ] and the Cramér–Rao lower bound [ 6 ] for sensing, and quality of service (QoS) constraints, such as a minimum signal-to-interference-plus-noise ratio (SINR) [ 7 ] , among others for the communications. Most existing studies in JCAS however focus on optimizing the system for a given time, e.g., [ 8 , 3 , 5 , 7 ] , regardless of potential variations in network load, resource availability, and memory-dependent time-varying parameters such as wireless channel characteristics [ 9 ] and impulse responses of sensed targets [ 10 ] . At the same time, satisfying instantaneous (per-slot) constraints for radar and communications might be challenging, if not impossible. Yet, the system can still operate if average radar and communications requirements are met; this relates to the non-guaranteed bit rate within the QoS requirements in 3GPP Release 19 [ 11 ] . Besides, there are inherent differences between radar and communications operations, depending on the time variations. For instance, the demand for communications is generally lower at night compared to daytime, whereas many sensing applications require continuous operation throughout the day and night. Motivated by the above facts, and to account for a dynamic balance in the resource utilization between radar and communications, this study addresses a JCAS problem aiming to dynamically design dedicated precoders for radar and communications to satisfy an average SINR requirement for communications and maximize average radar SNR. We consider a JCAS system with a base station (BS), a radar target, and a single-antenna user, where communications occur in the downlink. At each time slot, the BS simultaneously transmits data to the user and receives the echo signal back from the target. The goal is to dynamically design dedicated precoders for both radar and communications. We formulate a stochastic control problem to maximize the long-term average radar SNR, subject to an average SINR constraint for communications and the BS’s power budget. The main challenge here is to design a dynamic precoding algorithm under uncertainty and time variations while ensuring strict per-slot and average constraints. Using Lyapunov optimization, we cast the problem into a sequence of per-slot non-convex optimization problems. To solve these per-lot problems, we apply the successive convex approximation (SCA) method. Additionally, we provide a closed-form solution based on zero-forcing (ZF), resulting in a more efficient dynamic algorithm. Simulation results demonstrate the effectiveness of the proposed algorithms and their performance compared to a baseline algorithm. The results show that the ZF-based method achieves nearly the same performance as the SCA-based method while offering a considerably lower computational complexity. A handful of papers studied the problem of dynamic JCAS design, e.g., [ 12 , 13 , 14 , 15 , 10 , 16 ] . Work [ 10 ] developed a dynamic power allocation strategy for a vehicular setup to maximize the average mutual information for the radar, subject to minimum average rate and power budget constraints. The work in [ 14 ] focused on maximizing average throughput under a hybrid repeat-request protocol, with constraints on minimum average probability of detection and average power utilization. Reference [ 15 ] proposed a learning-based algorithm for dynamic resource management, while [ 12 ] introduced a memory-based learning method for online channel sensing using backscattered signals. The study in [ 13 ] combined Lyapunov optimization with deep reinforcement learning to provide dynamic power and subcarrier allocation, subject to average data rate and sensing requirements. The work by [ 16 ] proposed a deep reinforcement learning method to dynamically allocate dwell time for target tracking and data transmission. Unlike previous works, we devise a dynamic, dedicated radar and communications precoding algorithm aimed at maximizing the average sensing SNR, subject to constraints on average communications SINR and total power budget of the BS. The most closely related work to this paper is [ 10 ] , which also employs Lyapunov optimization to address a time-averaged problem. However, in contrast to [ 10 ] , we introduce the SCA method and a ZF-based approach with a closed-form solution to the per-slot problems. Additionally, our problem with the precoding design fundamentally differs from the dynamic power allocation problem studied in [ 10 ] . II System Model and Problem Formulation We consider a JCAS system with a BS, a target, and a single-antenna (communications) user. The communications takes place on the downlink. The BS is equipped with N transmit antennas and M receive antennas. For clarity of the presentation, we use subscript “ r ” for notations associated with radar/sensing, and “ c ” for those with communications. At time slot t , the BS simultaneously transmits probing signals to the target and data signals to the communications user using dedicated precoding vectors \mathbb{w}_{\textsf{r}}(t)\in\mathbb{C}^{N\times 1} and {\mathbb{w}_{\textsf{c}}(t)\in\mathbb{C}^{N\times 1}} , respectively. The BS then receives the radar echo signal using M antennas. Let \mathbb{h}(t)\in\mathbb{C}^{N\times 1} denote the channel from the BS to the user in slot t . Note that, due to fluctuations in the wireless channels, \mathbb{h}(t) is essentially a stochastic process. The received signal by the user is then given by y_{\textsf{c}}(t)=\mathbb{h}^{\textsf{H}}(t)\mathbb{w}_{\textsf{c}}(t){s}_{\textsf{c}}(t)+\mathbb{h}^{\textsf{H}}(t)\mathbb{w}_{\textsf{r}}(t){s}_{\textsf{r}}(t)+n_{\textsf{c}}(t), (1) where {s}_{\textsf{c}}(t) is the transmit (data) signal to the communications user, \mathbb{E}\{|{{s}_{\textsf{c}}(t)}|^{2}\}=1 , {s}_{\textsf{r}}(t) is the transmit radar signal with unit energy, and n_{\textsf{c}}(t)\in\mathbb{C} is additive white Gaussian noise (AWGN) following the distribution \mathcal{CN}(0,\sigma^{2}_{\textsf{c}}) , with \sigma^{2}_{\textsf{c}} denoting the noise variance at the communications user. The radar’s dedicated signal becomes interference for the user. As such, the SINR at the user at slot t is given by \gamma_{\textsf{c}}(t)=\frac{|\mathbb{h}^{\textsf{H}}(t)\mathbb{w}_{\textsf{c}}(t)|^{2}}{|\mathbb{h}^{\textsf{H}}(t)\mathbb{w}_{\textsf{r}}(t)|^{2}+\sigma_{\textsf{c}}^{2}}. (2) While transmitting communications signals to the users, the BS also receives echo signals from the target. Let {\mathbb{G}(t)\in\mathbb{C}^{M\times N}} denote the two-way channel between the BS and the target in slot t . We model \mathbb{G}(t) as [ 8 ] \mathbb{G}(t)=\mathbb{b}(t)\mathbb{a}^{\textsf{H}}(t), (3) where \mathbb{a}(t)\in\mathbb{C}^{N\times 1} and \mathbb{b}(t)\in\mathbb{C}^{M\times 1} are the steering vectors of the transmit and receive

Read the complete text