Unsupervised Learning Approach for Beamforming in Cell-Free Integrated Sensing and Communication
Mohamed Elrashidy, Mudassir Masood, Ali Arshad Nasir · 2024
Read Unsupervised Learning Approach for Beamforming in Cell-Free Integrated Sensing and Communication by Mohamed Elrashidy, Mudassir Masood, Ali Arshad Nasir (2024) 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-2412-18162.txt.
Introduction
Mohamed Elrashidy, Mudassir Masood, and Ali Arshad Nasir published this paper in 2024. It tackles beamforming for cell-free networks, where many small access points serve users jointly instead of one big tower. Their setting merges communication with sensing: the same signals carry data and probe the surroundings.
Classical beamforming math is heavy and loads the links back to the central processor. The team instead trains neural networks without labels, using a teacher-student scheme. Two teacher networks each master one goal, sensing quality or communication quality, and a student network learns to match the best of both, avoiding a forced trade between them. The design runs in a scattered way across access points, easing the central load. Results came close to the best classical solution while running at least a thousand times faster. The authors pitch the method for future dense, dual-purpose networks.
Catalog record
| CREATOR | Mohamed Elrashidy, Mudassir Masood, Ali Arshad Nasir |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-18162
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Unsupervised Learning Approach for Beamforming in Cell-Free Integrated Sensing and Communication Mohamed Elrashidy, Mudassir Masood, Ali Arshad Nasir Abstract Cell-free massive multiple input multiple output (MIMO) systems can provide reliable connectivity and increase user throughput and spectral efficiency of integrated sensing and communication (ISAC) systems. This can only be achieved through intelligent beamforming design. While many works have proposed optimization methods to design beamformers for cell-free systems, the underlying algorithms are computationally complex and potentially increase fronthaul link loads. To address this concern, we propose an unsupervised learning algorithm to jointly design the communication and sensing beamformers for cell-free ISAC system. Specifically, we adopt a teacher-student training model to guarantee a balanced maximization of sensing signal to noise ratio (SSNR) and signal to interference plus noise ratio (SINR), which represent the sensing and communication metrics, respectively. The proposed scheme is decentralized, which can reduce the load on the central processing unit (CPU) and the required fronthaul links. To avoid the tradeoff problem between sensing and communication counterparts of the cell-free system, we first train two identical models (teacher models) each biased towards one of the two tasks. A third identical model (a student model) is trained based on the maximum sensing and communication performance information obtained by the teacher models. While the results show that our proposed unsupervised DL approach yields a performance close to the state-of-the-art solution, the proposed approach is more computationally efficient than the state of the art by at least three orders of magnitude. Journal: Physical Communication 1 Introduction Integrated sensing and communication (ISAC) is a promising 6G technology that is expected to support many applications such as radar [ 1 ] , localization [ 2 ] , vehicular networks [ 3 ] , [ 4 ] , [ 5 ] , and UAV assisted networks [ 6 ] . However, the performance of ISAC systems is affected by a tradeoff between communication quality and sensing accuracy. Dong et al. [ 7 ] proposed a general scheme for resource allocation in ISAC systems and studied the mentioned tradeoff between target detection and communication quality of service (QoS). While some works focused on structured beamforming and optimized power and bandwidth allocation coefficients in different ISAC MIMO systems [ 7 , 8 , 9 ] , it can be argued that optimizing unstructured beamformers is more general since the allocated power coefficients are readily included in the optimized beam vectors. Xu et al. [ 10 ] formulated a beamforming problem for ISAC systems to maximize the sum secrecy rate while maintaining a certain user QoS. After handling the non-convexity of the problem, the authors proposed an iterative algorithm based on block coordinate descent to obtain a suboptimal solution. He et al. [ 11 ] proposed two problems to design the transmit and receive beamformers and signal covariance matrix through minimizing the transmit power or maximizing the sum rate in ISAC systems. The two problems are subject to sensing and communication quality thresholds. Hua et al. [ 12 ] proposed an ISAC downlink design by minimizing the beam pattern matching error, where pre-designed beam patterns are considered. The minimization is subjected to a certain signal-to-interference plus noise ratio (SINR) threshold, and the optimization parameters are the beamformers and radar signal covariance matrix. Many other works addressed the beamforming problem in reconfigurable intelligent surface (RIS) assisted ISAC systems [ 13 , 14 , 15 , 16 ] . The use of deep learning (DL) to design the downlink of general ISAC systems has recently been explored [ 17 , 18 ] . Particularly, Liu et al. [ 19 ] trained a couple of deep neural networks (DNNs) to estimate sensing and communication channels for an ISAC system. Liu et al. [ 3 ] used a recurrent neural network (RNN) that is composed of convolutional long-short term memory (CLSTM) subunits. Such network is expected to capture the spatial and temporal properties of the environment. The network was trained to predict the angles of the vehicles within the range of an ISAC multi-antenna BS. The predicted angles were used for designing a predictive transmit beamformer. The training was done offline in a supervised fashion. A similar approach is adopted by Liu et al. [ 20 ] , where an RNN consisting of CLSTM subunits was used for direct beamforming in the same system considered by [ 3 ] . In [ 20 ] , the authors first proposed a sum-rate maximization problem constrained by Cramer-Rao lower bound as a sensing error metric. Solving this problem provided means to both generate an unsupervised dataset and a baseline for comparison between the conventional CVX solver and the unsupervised learning scheme. Mu et al. [ 21 ] proposed an FCNN that predicts the beamformer and the angular parameters of the detected vehicles given the received signal at the ISAC-assisted roadside units. Liu et al. [ 22 ] proposed a DL network architecture to manage interference in ISAC systems through power allocation design. To that end, the authors utilized unsupervised learning for training. They also used transfer learning to predict a proper beamforming scheme for interference management. The aforementioned works considered single cell systems where beamformers are designed at a single base station (BS) to serve a number of users within a designated area. Cell-free systems, on the other hand, are promising in terms of meeting the demands of next generation technologies, such as energy and spectral efficiency, low latency and reliability [ 23 ] . It was found that cell-free schemes outperform conventional small-cell networks tremendously in terms of 95%-likely per-user throughput, energy efficiency and spectral efficiency [ 24 , 25 , 26 , 27 ] . Nevertheless, only few recent works considered beamforming design problems in cell-free ISAC systems. Demirhan and Alkhateeb [ 28 ] proposed a joint sensing and communication beamforming optimization problem, where the objective is to jointly maximize sensing and communication metrics, namely, sensing signal-to-noise ratio (SSNR) and SINR, respectively, subjected to the power budget constraint at each access point (AP). Their results show that the joint optimization problem outperforms other methods where sensing or communication beams are designed a priori. Liu et al. [ 29 ] considered the problem of AP selection (i.e., transmitting or receiving modes) to improve the degrees of freedom in cell-free ISAC networks. The authors formulated a sum-maximization problem to jointly optimize the AP selection mode, the beamformer vectors and the receive filter vectors. Other works that studied the beamforming problem in cell-free ISAC networks include [ 30 , 31 , 32 , 33 ] . Since studies on ISAC-assisted cell-free MIMO systems are limited and recent, the application of DL techniques for solving power allocation or beamforming design has not been considered thus far. The only exception is the recent work in [ 9 ] , where the authors considered a deep reinforcement learning (DRL) solution to optimize power coefficients for the data and the pilot in cell-free ISAC systems. However, this solution requires discretizing the optimization variables, since the action space is usually discrete. Also, the work did not consider solving a more general and challenging beamforming design problem. It is worth mentioning that many works utilized DL solutions to allocate resources in cell-free