A Literature Survey on AI-Aided Beamforming and Beam Management for 5G and 6G Systems
Abstract
:1. Introduction
2. Systematic Review
2.1. The Need for This Review
2.2. Research Questions
- 1:
- What are the beamforming and beam management challenges to face, and which are amenable to AI solutions?
- 2:
- What ML techniques are adequate and often applied for beam-related problems?
- 3:
- What are the benefits and downsides of applying ML algorithms to beamforming and beam management problems?
- 4:
- How were the datasets composed and used for ML training and simulation?
- 5:
- Which are the future directions of research for AI-based beamforming and beam management?
2.3. Search String Definition
2.4. Criteria for Inclusion and Exclusion
2.5. Identify Primary Studies
2.6. Review Results and Contributions
2.7. Scope of This Review
3. Beamforming Architectures
3.1. Analog Beamforming
3.2. Digital Beamforming
3.3. Hybrid Beamforming
4. Related Works
5. Beam Selection in MIMO Systems
6. Mobility and Handover
7. Codebook Design
Challenges | Algorithm | Highlight (Pros) | Limitations (Cons) | Key Contribution | Ref. |
---|---|---|---|---|---|
Hardware and deployment awareness | NN |
● Robust to hardware impairments. ● Decouples learning process from communication. ● The codebook is refined while communication goes on. |
● The offline learning process can be time-consuming and requires accurate channel state information. ● For a reduced number of codewords, the performance is not satisfying. |
● Online machine learning framework. ● Adapts the codebook and avoids the need for explicit channel state information. | [149] |
Limited feedback | K-means |
● Adapts well to the underlying channel distribution. ● Reduces the feedback overhead. |
● K-means clustering does not work well with high-dimensional vectors. ● The method utilizes uninterpretable Grassmannian clustering to find optimal codebooks in a black-box manner. | Reduces the codebook design problem to an unattended clustering problem in a Grassmann collector. | [150] |
● DNN ● K-means |
● Networks could learn the key propagation characteristics of the CSI. ● Clustering algorithms acquire the centroids of the corresponding characteristic. |
● The offline learning process can be time-consuming and requires accurate CSI. ● Network performance for smaller antennas is lower than for larger antennas. ● The number of spatial lobes affects the accuracy of the alignment direction. |
● Reduces the dimension of the full space and the feedback overhead. ● Proposes a deep-clustering-based codebook design for massive MIMO systems. ● Improves the performance of limited feedback massive MIMO systems. | [152] | |
SOM |
● Simple implementation. ● Better than the DFT codebook. |
● Initial codebook depends on prior massive channel data. ● Ignores the impact of noise over the channel samples. | The proposed method can update the codebook adaptively according to the instantaneous channel state information. | [161] | |
Grassmannian line packing |
● Codebook adaptive to any Rician factors. ● The proposed codebook substantially outperforms conventional methods. | When the Rician factor is small, the impact of the NLoS components is greater. As a result, the average quantization distortion increases. | Deduces the angle distribution between the channel vectors and the LoS component and a precisely approximated angle distribution in a tractable form. | [162] | |
Environment awareness | NN |
● Proposes an ML model that adapts codebook beam patterns based on the surrounding environment and user distribution. ● The developed NN architecture takes into account hardware constraints found in analog-only mmWave/THz transceivers. |
● With only 3 bits, the learned codebook reaches more than 85% of the upper limit with 64 beams and 90% of the upper limit with 128 beams. ● A 16-beam DFT codebook performs comparably to a 64-beam DFT codebook. |
● Proposes an NN-based framework for learning environment-aware beamforming codebooks that takes hardware impairments into account. ● The solution is important when the resolution of the analog phase shifters is limited. | [153] |
Beam selection | CEO | Guarantees a result within 98% of that obtained by ES with substantially lower complexity. |
● Limitations due to the use of finite resolution phase shifters. ● Results rely on the simulation of a single static scenario with idealized assumptions. ● Performance could differ substantially in realistic dynamic conditions with unpredictable changes and channel effects. |
● Proposes a low-complexity, near-optimal algorithm for identifying an optimal analog precoder and analog combiner pair in mmWave massive MIMO systems. ● Demonstrates the effectiveness of the proposed algorithm in achieving near-optimal performance. ● Addresses the hardware complexity challenge of mmWave massive MIMO systems. | [154] |
Large codebook sizes | RL |
● Does not require channel knowledge. ● Evaluates hardware impairments. |
● Proposes a complex deep reinforcement learning architecture. ● The proposed Wolpertinger-variant architecture is selected based on heuristics rather than rigorous analysis. Superior alternatives may exist, and performance could likely be improved with principled design. |
● Design of deep reinforcement learning. ● Relies only on receiving power measurements and does not require any channel knowledge. ● Framework capable of learning a codebook for users in the surrounding environment. | [155] |
Maximize the achievable rate | K-means | The proposed codebook design can recognize and adapt to arbitrary propagation environments. |
● Large amounts of CSI are stored as the input data. ● Key design choices, such a the number of clusters (codebook size) and clustering algorithm, are selected based on heuristics, rather than rigorous analysis. | Proposes using characteristics extracted from the clustering centroids used as the key channel information. | [156] |
Optimal precoding policy for complex MIMO | DRL | The proposed precoding framework can outperform the conventional approximation algorithm in the complex MIMO environment. |
● Does not compare with any other solution in the literature. ● Questions remain around how the agents might adapt their precoding policies to non-stationary conditions, including changes in channel statistics, and whether the approach could do so effectively while maintaining near-optimal performance. | DQN- and DDPG-based agents can learn the near-optimal policy for the precoder selection problem. | [157] |
Balanced MRT-ZF combined optimization | DL |
● Outperforms MRT and ZF in terms of data rate. ● Computational complexity below the optimal solution. |
● A small number of users is used in the simulation results. ● Only considers Rayleigh channel model. | This paper uses DL to build beamforming vectors based on the sub-optimal solutions provided by the MRT and ZF methods. | [164] |
Interference mitigation (SI and CCI) | MLP | The trained model presents lower computational complexity than the Optimization-Driven Design (ODD) approach. |
● Training dataset depends on complex optimization problem solution. ● The solution quality is coupled to the dataset size. ● The proposed MLP-based solution has scalability issues. |
● The proposed solution presents a sub-optimal solution that is comparable to the traditional ODD approach. | [165] |
SINR balancing and power minimization | BNN | Achieves high beamforming accuracy when combining supervised and unsupervised learning. |
● The beamforming prediction must be trained previously. ● The approach ignores constraints such as computational resources, hardware limitations, or complexity requirements for realistic implementation. Practical feasibility is not demonstrated. |
● The framework is designed based on the CNN structure. ● Proposes a hybrid two-stage BNN with both supervised and unsupervised learning. | [166] |
Paper | ML Technique | Antenna Type and Frequency | Complexity Analysis | |
---|---|---|---|---|
[149] | NN | 64 Uniform Linear Array (ULA) | 28 GHz | ◯ |
[150] | K-means | 2 × 2, 3 × 3, 4 × 4, 5 × 5, 6 × 6, 8 × 8, 12 × 12 | 2.5 GHz | ◯ |
[152] | K-means | 32, 64, 128 UPA 32, 64 ULA | Not mentioned | ◯ |
[153] | NN | 64 ULA | 28 GHz | ◯ |
[154] | CEO | 16 × 64, 32 × 128 | 28 GHz | ⚫ |
[155] | RL | 32 ULA | 60 GHZ 28 GHz | ◯ |
[156] | K-means | 4 ULA, 16 ULA | Not mentioned | ◯ |
[157] | DRL | 4 × 2 MIMO | Not mentioned | ◯ |
[161] | SOM | 4, 16, 64 ULA | Not mentioned | ◯ |
[162] | Grassmannian LP | 32 ULA | Not mentioned | ◯ |
[164] | DL | 2, 3 ULA | Not mentioned | ⚫ |
[165] | MLP | 4, 2 ULA | Not mentioned | ◯ |
[166] | BNN | 2–12 ULA | Not mentioned | ⚫ |
8. Precoding and Combining in MIMO with Hybrid or Digital Architectures
Challenges | Algorithm | Highlight (Pros.) | Limitations (Cons.) | Key Contributions | Ref. |
---|---|---|---|---|---|
Channel estimation | DL | Solves two problems with a similar approach. |
● Large NN offline training overhead. ● Based on artificial channel measurements. | A comparison between a DL compressed sensing channel estimation for MIMO and deep learning quantized phase hybrid precoding. | [168] |
DL |
● Lower computational complexity. ● Imperfect CSI premise. ● Better than others state-of-the-art greedy and sum-rate optimization precoders. |
● Needs a large training dataset to provide robustness. ● The long-term stability and robustness of CNN-MIMO are uncertain since neural networks can be prone to “catastrophic forgetting” as environments change. ● The exhaustive search method used to generate training labels for CNN-MIMO may not provide optimal hybrid precoding solutions, limiting the performance CNN-MIMO can ultimately achieve. |
● A CNN that accepts an imperfect channel matrix and outputs analog precoder and combiners. ● A exhaustive search algorithm for the analog precoder to feed the CNN training. ● A solution that is capable of training with large amounts of data. | [169] | |
DL | Good results with lower computational complexity if compared to SVD- and GMD-based methods. |
● The simulated communications environment is poorly described. ● The method used to generate training labels for the DNN is not specified, so it is unclear if it can provide optimal hybrid precoding performance for the DNN to achieve. ● The limitations of the approach regarding robustness against parameter estimation errors, channel estimation inaccuracies, or other practical impairments are not analyzed. |
● A novel framework that incorporates DL into hybrid precoding. ● A DNN with lower computational complexity requirements in the training phase. ● The DNN provides accurate hybrid precoding while supporting channel feedback. | [170] | |
DL | The proposed solution can be generalized to unseen environments. |
● The training time is not discussed to assess the feasibility of the proposed solution. ● As DL solutions are tailored for a specific setup, the proposed approach may not generalize well to other network architectures and duplexing modes. ● It does not provide a rigorous mathematical analysis of the approach’s performance gains, complexity, and optimality. |
● Joint DNN architectures for high generalization. ● DNNs achieve outstanding performance in scenarios where downloading training datasets is very limited. | [171] | |
DRL | The hybrid beamforming method spectral efficiency that surpasses the fully digital precoding |
● As it is a new ML scheme, it lacks a complexity assessment to fairly compare it with the other algorithms. ● The impact of different RL algorithms, such as actor–critic and policy gradient, on the proposed approach’s performance is unclear. | The authors propose a new way of combining DL and RL for beamforming, leveraging high spectral efficiency and overall beamforming effectiveness. | [172] | |
ine Dynamic subarrays | AHC | Proposed hybrid precoding, which can efficiently avoid mutually correlated metrics. |
● The authors do not mention the simulation tools used. ● The clustering algorithm misses information about the training phase. |
● Optimal hybrid precoder on PCA. ● Agglomerative hierarchical clustering to group dynamic subarrays. ● Energy efficiency for passive and active antennas. ● A DNN algorithm to predict the dimension output in MIMO. ● A customized DNN algorithm to cope with the requirements. | [174] |
ine Two-stage precoding | DL | Proposes an ML-based approach to finding optimal dimensions with good accuracy and closer to the brute-force solution. |
● The authors do not describe the dataset, nor its size and format. ● The training phase requires too many iterations. | [175] | |
Hybrid, analog, and digital precoding | DL |
● Generalizable for many systems with many parameters. ● Numerical results suggest the performance of the proposed approach is closer to optimal. |
● Missing some ML algorithm details. ● Simultaneous design of analog and digital precoders using a neural network is difficult. ● The approach is not generalizable to systems with different numbers of users. |
● Proposes a joint channel sensing and downlink precoding solution that avoids explicit channel estimation. ● Introduces an end-to-end design that directly builds precoders from the received pilots without the intermediate channel estimation step. | [176] |
ine BF-based on IRS | DL |
● The combination of BF-based on IRS with BS enhances the system sum-rate. ● Uses an NN to achieve the optimal configuration. ● Good generalization rate achieved by ML algorithm. |
● The convergence time is not discussed. ● Different DNN architectures could be evaluated. |
● A combined BF based on BSs and IRS. ● An optimization method for implicit channel estimation. ● A DNN performance assessment for the BF. | [178] |
Location-based | DL | A method capable of handling LoS and NLoS propagation. |
● The solution does not predict the user location in BF. ● The solution does not predict the channel vector directly. |
● A supervised learning method to map user location to an appropriate precoder. ● Reduces the need for pilot symbols. | [179] |
Complexity reduction | DL | The proposed method has low computation complexity when compared with CNNs. |
● The computational complexity relies on the learning technology design (CNN or ELM). ● ELM still requires a large amount of training data to work effectively, which is not specified in the paper. ● The work considers a single-cell MU-MIMO system. Performance in a multi-cell environment with inter-cell interference would likely be different and is not addressed. |
● Novel, robust, and low-complexity hybrid BF algorithms. ● An optimization method based on fractional programming to provide labels for the training set. | [180] |
DL | Using PSO combined with DNN, the authors reduce the computational costs in managing antenna arrays. |
● Does not present accuracy, which hinders the performance assessment. ● It lacks an analysis of the robustness of the proposed method against various parameter estimation errors, channel estimation inaccuracies, or other practical impairments. ● It may not generalize well to different system configurations, antenna array geometries, or channel conditions, as the performance heavily relies on the training dataset. |
● A novel DL with phase-only digital BF for MIMO. ● A metaheuristic method based on DL is used to reduce the computational complexity. | [181] | |
DRL |
● Online adjustment of parameters to minimize the BER. ● Uses a bi-fold approach for finding the optimal precoding policy and the codebook and non-codebook-based precoding. |
● Does not compare with any other solution in the literature. ● Does not discuss the convergence time of the proposed algorithm. | Introduces a hybrid ML approach for precoding policy for complex MIMO systems. | [157] | |
DL |
● Lower baseband precoding and combining training overhead. ● Detailed experimental evaluation description supports reproducibility. |
● Leveraging prior knowledge with DL has an underlying training cost to collect information about the end-to-end channel and network training. ● Assumes a specific channel model, and it is not clear how the proposed approach would perform with different channel models. |
● A reduction of training overhead compared to classical (non-ML) solutions. ● A novel DL-based approach to optimize channel measurement vectors. | [182] | |
CSI feedback | DL |
● Can be implemented in a real-time system due to low computational complexity. ● Works in FDD mode. ● Short training time, as it is unsupervised. |
● Might not be as precise as CSI-trained DL models. ● Assumes the availability of RSSI measurements, which may be difficult to obtain in practice or come with their own costs, complexity, and accuracy issues ● Comparisons are only made to conventional full-CSI solutions, lacking consideration of other learning, optimization, or approximate approaches proposed for this problem. |
● Evaluation of unsupervised learning to design the hybrid BF. ● Use of ray-tracing model in the deployment environment. ● A loss function proposal that is based on the sum-rate for classification and regression. ● Evaluation of non-DL and DL hybrid BF for the realistic channel model. | [163] |
9. Security of AI Models
10. Limitations of AI-Based Beamforming and Beam Management
- Limited applicability: AI-based algorithms may work well in specific scenarios, but may not be suitable for other scenarios. For example, algorithms designed for pedestrian mobility may not work well for high-speed mobility scenarios such as trains or urban vehicles. Additionally, algorithms optimized for indoor environments may not perform well in outdoor settings with different propagation characteristics.
- Reliance on training data: AI-based algorithms require large amounts of training data to learn the optimal beamforming and beam management strategies. If the training data are not representative of the actual operating environment, the performance of the algorithm may suffer. Furthermore, collecting and labeling these data can be time-consuming and expensive, especially in dynamic and rapidly changing environments.
- Limited generalizability: The performance of AI-based algorithms can be heavily influenced by the training data used to develop them. Therefore, the algorithms may not generalize well to new scenarios or environments where the training data do not adequately represent the target environment. Transfer learning techniques can help mitigate this issue, but they may not completely overcome the problem.
- Complexity: AI-based beamforming and beam management algorithms can be complex and require significant computational resources. This can increase the cost and power consumption of the system, making it less attractive for energy-constrained or cost-sensitive applications.
- Limited interpretability: AI-based algorithms often rely on complex deep learning models, which can be difficult to interpret. This can make it challenging to understand why certain decisions are being made or to identify errors or biases in the algorithm’s output. Such limited interpretability can also hinder the adoption of AI-based beamforming and beam management in industries where explainability and regulatory compliance are necessary.
- Limited robustness: AI-based algorithms may be vulnerable to adversarial attacks or other forms of interference that can disrupt their performance. This can limit their reliability in real-world applications, where security and robustness are critical factors. Developing AI-based algorithms that are resistant to such attacks and can maintain stable performance under a variety of conditions is an ongoing research challenge.
- Limited scalability: As the number of antennas and users in a massive MIMO system increases, the complexity of AI-based beamforming and beam management algorithms can become prohibitively high. This can limit their scalability and make them less practical for large-scale deployment. Efficient algorithms that can handle the growing demands of massive MIMO systems while maintaining high performance are needed to address this challenge.
- Latency: The real-time processing requirements of beamforming and beam management may impose strict latency constraints on AI-based algorithms. Designing AI-based methods that can perform fast computations and adapt to dynamic environments within these latency bounds is crucial for their practical implementation.
11. Open Problems and Future Research Directions
11.1. Centralized and Decentralized Learning
11.2. Reproducible Research
11.3. Semi-Supervised, Active, and Reinforcement Learning
11.4. Prototypes and Real-World Demonstrations
11.5. Privacy and Security
11.6. Computer Vision
11.7. Beamforming at Low-SNR Regimes and Joint Optimization
11.8. Channel Estimation
- Developing robust and efficient channel estimation algorithms that can handle the sparsity of the channel and limited coherence time.
- Investigating new channel estimation techniques that can take advantage of the hardware constraints and limitations of mmWave and THz systems, such as low-resolution Analog-to-Digital Converters (ADCs) and limited feedback bandwidth.
- Addressing the challenges of beam misalignment and developing adaptive channel estimation algorithms that can adjust to changes in the user location or mobility.
- Investigating the use of machine learning techniques for channel estimation in mmWave and THz systems, such as deep learning and reinforcement learning, which can potentially improve the accuracy and efficiency of channel estimation.
- Multipath interference: In mmWave and THz systems, the multipath components can arrive at the receiver with different delays and phases, leading to interference and reduced signal quality. Channel estimation algorithms need to be designed to handle the interference and accurately estimate the channel coefficients.
- Environmental effects: The mmWave and THz signals are highly sensitive to environmental factors such as atmospheric absorption, scattering, and reflection. These effects can cause significant variations in the channel characteristics, making it challenging to estimate the channel accurately.
- Scalability: The use of a large number of antenna elements in mmWave and THz systems can lead to scalability issues in channel estimation. Efficient channel estimation algorithms that can handle a large number of antennas are needed to enable the practical deployment of such systems.
- Hybrid beamforming: In practical mmWave and THz systems, hybrid beamforming techniques are often used, which combine digital and analog beamforming. Channel estimation algorithms need to be designed to handle the complexity of such hybrid beamforming architectures.
11.9. Definition of the Optimal ML Algorithm for a Given Beamforming Application
12. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Cisco VNI. Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2017–2022; White Paper, Cisco Visual Networking Index; Cisco: San Jose, CA, USA, 2019. [Google Scholar]
- Mietzner, J.; Schober, R.; Lampe, L.; Gerstacker, W.H.; Hoeher, P.A. Multiple-antenna techniques for wireless communications—A comprehensive literature survey. IEEE Commun. Surv. Tutor. 2009, 11, 87–105. [Google Scholar] [CrossRef]
- de Figueiredo, F.A.; Dias, C.F.; de Lima, E.R.; Fraidenraich, G. Capacity bounds for dense massive MIMO in a line-of-sight propagation environment. Sensors 2020, 20, 520. [Google Scholar] [CrossRef]
- Yang, H.; Herben, M.H.; Akkermans, I.J.; Smulders, P.F. Impact analysis of directional antennas and multiantenna beamformers on radio transmission. IEEE Trans. Veh. Technol. 2008, 57, 1695–1707. [Google Scholar] [CrossRef]
- Sohrabi, F.; Yu, W. Hybrid Digital and Analog Beamforming Design for Large-Scale Antenna Arrays. IEEE J. Sel. Top. Signal Process. 2016, 10, 501–513. [Google Scholar] [CrossRef]
- Steyskal, H. Digital Beamforming. In Proceedings of the 1988 18th European Microwave Conference, Stockholm, Sweden, 12–15 September 1988; pp. 49–57. [Google Scholar] [CrossRef]
- Molisch, A.F.; Ratnam, V.V.; Han, S.; Li, Z.; Nguyen, S.L.H.; Li, L.; Haneda, K. Hybrid beamforming for massive MIMO: A survey. IEEE Commun. Mag. 2017, 55, 134–141. [Google Scholar] [CrossRef]
- Vilas Boas, E.C.; de Figueiredo, F.A.; Mendes, L.L.; de Souza, R.A. Artificial intelligence for channel estimation in multicarrier systems for B5G/6G communications: A survey. EURASIP J. Wirel. Commun. Netw. 2022, 2022, 1–63. [Google Scholar] [CrossRef]
- Mo, J.; Ng, B.L.; Chang, S.; Huang, P.; Kulkarni, M.N.; AlAmmouri, A.; Zhang, J.C.; Lee, J.; Choi, W.J. Beam codebook design for 5G mmWave terminals. IEEE Access 2019, 7, 98387–98404. [Google Scholar] [CrossRef]
- Kutty, S.; Sen, D. Beamforming for millimeter wave communications: An inclusive survey. IEEE Commun. Surv. Tutor. 2015, 18, 949–973. [Google Scholar] [CrossRef]
- Kim, J.; Molisch, A.F. Fast millimeter-wave beam training with receive beamforming. J. Commun. Netw. 2014, 16, 512–522. [Google Scholar] [CrossRef]
- Dala Pegorara Souto, V.; Dester, P.S.; Soares Pereira Facina, M.; Gomes Silva, D.; de Figueiredo, F.A.P.; Rodrigues de Lima Tejerina, G.; Silveira Santos Filho, J.C.; Silveira Ferreira, J.; Mendes, L.L.; Souza, R.D.; et al. Emerging MIMO Technologies for 6G Networks. Sensors 2023, 23, 1921. [Google Scholar] [CrossRef]
- Zong, B.; Fan, C.; Wang, X.; Duan, X.; Wang, B.; Wang, J. 6G technologies: Key drivers, core requirements, system architectures, and enabling technologies. IEEE Veh. Technol. Mag. 2019, 14, 18–27. [Google Scholar] [CrossRef]
- Elayan, H.; Amin, O.; Shubair, R.M.; Alouini, M.S. Terahertz communication: The opportunities of wireless technology beyond 5G. In Proceedings of the 2018 International Conference on Advanced Communication Technologies and Networking (CommNet) IEEE, Marrakech, Morocco, 2–4 April 2018; pp. 1–5. [Google Scholar]
- Roh, W.; Seol, J.Y.; Park, J.; Lee, B.; Lee, J.; Kim, Y.; Cho, J.; Cheun, K.; Aryanfar, F. Millimeter-wave beamforming as an enabling technology for 5G cellular communications: Theoretical feasibility and prototype results. IEEE Commun. Mag. 2014, 52, 106–113. [Google Scholar] [CrossRef]
- Zhao, H.; Mayzus, R.; Sun, S.; Samimi, M.; Schulz, J.K.; Azar, Y.; Wang, K.; Wong, G.N.; Gutierrez, F.; Rappaport, T.S. 28 GHz millimeter wave cellular communication measurements for reflection and penetration loss in and around buildings in New York city. In Proceedings of the 2013 IEEE International Conference on Communications (ICC), Budapest, Hungary, 9–13 June 2013; pp. 5163–5167. [Google Scholar]
- Abouelseoud, M.; Charlton, G. The effect of human blockage on the performance of millimeter-wave access link for outdoor coverage. In Proceedings of the 2013 IEEE 77th Vehicular Technology Conference (VTC Spring), Dresden, Germany, 2–5 June 2013; pp. 1–5. [Google Scholar]
- MacCartney, G.R.; Deng, S.; Sun, S.; Rappaport, T.S. Millimeter-wave human blockage at 73 GHz with a simple double knife-edge diffraction model and extension for directional antennas. In Proceedings of the 2016 IEEE 84th Vehicular Technology Conference (VTC-Fall), Montreal, QC, Canada, 18–21 September 2016; pp. 1–6. [Google Scholar]
- Brownlee, J. Deep Learning for Computer Vision: Image Classification, Object Detection, and Face Recognition in Python; Machine Learning Mastery: Vermont, Australia, 2019. [Google Scholar]
- Jiang, C.; Zhang, H.; Ren, Y.; Han, Z.; Chen, K.C.; Hanzo, L. Machine Learning Paradigms for Next-Generation Wireless Networks. IEEE Wirel. Commun. 2017, 24, 98–105. [Google Scholar] [CrossRef]
- Tang, F.; Kawamoto, Y.; Kato, N.; Liu, J. Future Intelligent and Secure Vehicular Network Toward 6G: Machine-Learning Approaches. Proc. IEEE 2020, 108, 292–307. [Google Scholar] [CrossRef]
- Giordani, M.; Polese, M.; Roy, A.; Castor, D.; Zorzi, M. A tutorial on beam management for 3GPP NR at mmWave frequencies. IEEE Commun. Surv. Tutor. 2018, 21, 173–196. [Google Scholar] [CrossRef]
- Lim, Y.G.; Ji, H.; Park, J.H.; Kim, Y. Artificial intelligence-based beam management for high speed applications in mmWave spectrum. In Proceedings of the 2020 IEEE Globecom Workshops (GC Wkshps. IEEE), Taipei, Taiwan, 7–11 December 2020; pp. 1–6. [Google Scholar]
- Pereira de Figueiredo, F.A.; Mendes Lemes, D.; Ferreira Dias, C.; Fraidenraich, G. Massive MIMO channel estimation considering pilot contamination and spatially correlated channels. Electron. Lett. 2020, 56, 410–413. [Google Scholar] [CrossRef]
- Khan, M.Q.; Gaber, A.; Schulz, P.; Fettweis, G. Machine Learning for Millimeter Wave and Terahertz Beam Management: A Survey and Open Challenges. IEEE Access 2023, 11, 11880–11902. [Google Scholar] [CrossRef]
- Araújo, D.C.; Maksymyuk, T.; de Almeida, A.L.; Maciel, T.; Mota, J.C.; Jo, M. Massive MIMO: Survey and future research topics. IET Commun. 2016, 10, 1938–1946. [Google Scholar] [CrossRef]
- Pham, Q.V.; Nguyen, N.T.; Huynh-The, T.; Bao Le, L.; Lee, K.; Hwang, W.J. Intelligent Radio Signal Processing: A Survey. IEEE Access 2021, 9, 83818–83850. [Google Scholar] [CrossRef]
- Naeem, M.; De Pietro, G.; Coronato, A. Application of Reinforcement Learning and Deep Learning in Multiple-Input and Multiple-Output (MIMO) Systems. Sensors 2022, 22, 309. [Google Scholar] [CrossRef]
- She, C.; Sun, C.; Gu, Z.; Li, Y.; Yang, C.; Poor, H.V.; Vucetic, B. A Tutorial on Ultrareliable and Low-Latency Communications in 6G: Integrating Domain Knowledge Into Deep Learning. Proc. IEEE 2021, 109, 204–246. [Google Scholar] [CrossRef]
- Jiang, W.; Han, B.; Habibi, M.A.; Schotten, H.D. The road towards 6G: A comprehensive survey. IEEE Open J. Commun. Soc. 2021, 2, 334–366. [Google Scholar] [CrossRef]
- Bishop, C.M.; Nasrabadi, N.M. Pattern Recognition and Machine Learning; Springer: Berlin/Heidelberg, Germany, 2006; Volume 4. [Google Scholar]
- Hastie, T.; Tibshirani, R.; Friedman, J.H.; Friedman, J.H. The Elements of Statistical Learning: Data Mining, Inference, and Prediction; Springer: Berlin/Heidelberg, Germany, 2009; Volume 2. [Google Scholar]
- LeCun, Y.; Bengio, Y.; Hinton, G. Deep learning. Nature 2015, 521, 436–444. [Google Scholar] [CrossRef]
- Alpaydin, E. Introduction to Machine Learning; MIT Press: Cambridge, MA, USA, 2020. [Google Scholar]
- Murphy, K.P. Machine Learning: A Probabilistic Perspective; MIT Press: Cambridge, MA, USA, 2012. [Google Scholar]
- Chen, S.; Zhao, J. The requirements, challenges, and technologies for 5G of terrestrial mobile telecommunication. IEEE Commun. Mag. 2014, 52, 36–43. [Google Scholar] [CrossRef]
- Varshney, U. 4G wireless networks. IT Prof. 2012, 14, 34–39. [Google Scholar] [CrossRef]
- Liu, L.; Cheung, S.; Yuk, T. Compact MIMO antenna for portable devices in UWB applications. IEEE Trans. Antennas Propag. 2013, 61, 4257–4264. [Google Scholar] [CrossRef]
- Dang, S.; Amin, O.; Shihada, B.; Alouini, M.S. What should 6G be? Nat. Electron. 2020, 3, 20–29. [Google Scholar] [CrossRef]
- Lopez, A.V.; Chervyakov, A.; Chance, G.; Verma, S.; Tang, Y. Opportunities and Challenges of mmWave NR. IEEE Wirel. Commun. 2019, 26, 4–6. [Google Scholar] [CrossRef]
- Ning, B.; Tian, Z.; Chen, Z.; Han, C.; Yuan, J.; Li, S. Prospective beamforming technologies for ultra-massive MIMO in terahertz communications: A tutorial. arXiv 2021, arXiv:2107.03032. [Google Scholar]
- Masterson, C. Massive MIMO and beamforming: The signal processing behind the 5G buzzwords. Analog. Dialogue 2017, 51, 10–14. [Google Scholar]
- Keysight Technologies. Engineering the 5G World; Keysight Technologies: Santa Rosa, CA, USA, 2020. [Google Scholar]
- Bolcskei, H. MIMO-OFDM wireless systems: Basics, perspectives, and challenges. IEEE Wirel. Commun. 2006, 13, 31–37. [Google Scholar]
- Vannithamby, R.; Talwar, S. Massive MIMO Communications; Wiley Telecom: Hoboken, NJ, USA, 2017. [Google Scholar]
- Agiwal, M.; Roy, A.; Saxena, N. Next generation 5G wireless networks: A comprehensive survey. IEEE Commun. Surv. Tutor. 2016, 18, 1617–1655. [Google Scholar] [CrossRef]
- Mohamed, K.S.; Alias, M.Y.; Roslee, M.; Raji, Y.M. Towards green communication in 5G systems: Survey on beamforming concept. IET Commun. 2021, 15, 142–154. [Google Scholar] [CrossRef]
- Ali, E.; Ismail, M.; Nordin, R.; Abdulah, N.F. Beamforming techniques for massive MIMO systems in 5G: Overview, classification, and trends for future research. Front. Inf. Technol. Electron. Eng. 2017, 18, 753–772. [Google Scholar] [CrossRef]
- Zhang, S.; Guo, C.; Wang, T.; Zhang, W. ON–OFF analog beamforming for massive MIMO. IEEE Trans. Veh. Technol. 2018, 67, 4113–4123. [Google Scholar] [CrossRef]
- Rozé, A.; Crussière, M.; Hélard, M.; Langlais, C. Comparison between a hybrid digital and analog beamforming system and a fully digital Massive MIMO system with adaptive beamsteering receivers in millimeter-Wave transmissions. In Proceedings of the 2016 International Symposium on Wireless Communication Systems (ISWCS), Poznan, Poland, 20–23 September 2016; pp. 86–91. [Google Scholar] [CrossRef]
- Ahmed, I.; Khammari, H.; Shahid, A.; Musa, A.; Kim, K.S.; De Poorter, E.; Moerman, I. A Survey on Hybrid Beamforming Techniques in 5G: Architecture and System Model Perspectives. IEEE Commun. Surv. Tutor. 2018, 20, 3060–3097. [Google Scholar] [CrossRef]
- Barton, P. Digital beam forming for radar. In Proceedings F (Communications, Radar and Signal Processing); IET: Piscataway, NJ, USA, 1980; Volume 127, pp. 266–277. [Google Scholar]
- Talisa, S.H.; O’Haver, K.W.; Comberiate, T.M.; Sharp, M.D.; Somerlock, O.F. Benefits of digital phased array radars. Proc. IEEE 2016, 104, 530–543. [Google Scholar] [CrossRef]
- Bailleul, P.K. A new era in elemental digital beamforming for spaceborne communications phased arrays. Proc. IEEE 2016, 104, 623–632. [Google Scholar] [CrossRef]
- Rao, L.; Pant, M.; Malviya, L.; Parmar, A.; Charhate, S.V. 5G beamforming techniques for the coverage of intended directions in modern wireless communication: In-depth review. Int. J. Microw. Wirel. Technol. 2021, 13, 1039–1062. [Google Scholar] [CrossRef]
- Murray, B.P.; Zaghloul, A.I. A survey of cognitive beamforming techniques. In Proceedings of the 2014 United States National Committee of URSI National Radio Science Meeting (USNC-URSI NRSM), Boulder, CO, USA, 8–11 January 2014; p. 1. [Google Scholar] [CrossRef]
- Rumyancev, I.A.; Korotkov, A.S. Survey on Beamforming Techniques and Integrated Circuits for 5G Systems. In Proceedings of the 2019 IEEE International Conference on Electrical Engineering and Photonics (EExPolytech), Saint Petersburg, Russia, 17–18 October 2019; pp. 76–80. [Google Scholar] [CrossRef]
- El Ayach, O.; Heath, R.W.; Rajagopal, S.; Pi, Z. Multimode precoding in millimeter wave MIMO transmitters with multiple antenna sub-arrays. In Proceedings of the 2013 IEEE Global Communications Conference (GLOBECOM), Atlanta, GA, USA, 9–13 December 2013; pp. 3476–3480. [Google Scholar]
- Alsaba, Y.; Rahim, S.K.A.; Leow, C.Y. Beamforming in Wireless Energy Harvesting Communications Systems: A Survey. IEEE Commun. Surv. Tutor. 2018, 20, 1329–1360. [Google Scholar] [CrossRef]
- Sharma, A.; Jha, R.K. A Comprehensive Survey on Security Issues in 5G Wireless Communication Network using Beamforming Approach. Wirel. Pers. Commun. 2021, 119, 3447–3501. [Google Scholar] [CrossRef]
- Zardi, F.; Nayeri, P.; Rocca, P.; Haupt, R. Artificial Intelligence for Adaptive and Reconfigurable Antenna Arrays: A Review. IEEE Antennas Propag. Mag. 2021, 63, 28–38. [Google Scholar] [CrossRef]
- Rajarajeswarie, B.; Sandanalakshmi, R. A Short Survey: Applications of Artificial Intelligence in Massive MIMO. In Proceedings of the 2020 International Conference on Communication and Signal Processing (ICCSP), Chennai, India, 28–30 July 2020; pp. 1096–1100. [Google Scholar] [CrossRef]
- ElHalawany, B.M.; Hashima, S.; Hatano, K.; Wu, K.; Mohamed, E.M. Leveraging Machine Learning for Millimeter Wave Beamforming in Beyond 5G Networks. IEEE Syst. J. 2021, 16, 1739–1750. [Google Scholar] [CrossRef]
- Wu, C.; Lai, C.F. A survey on improving the wireless communication with adaptive antenna selection by intelligent method. Comput. Commun. 2022, 181, 374–403. [Google Scholar] [CrossRef]
- Kassir, H.A.; Zaharis, Z.D.; Lazaridis, P.I.; Kantartzis, N.V.; Yioultsis, T.V.; Xenos, T.D. A Review of the State of the Art and Future Challenges of Deep Learning-Based Beamforming. IEEE Access 2022, 10, 80869–80882. [Google Scholar] [CrossRef]
- Nor, A.M.; Halunga, S.; Fratu, O. Survey on positioning information assisted mmWave beamforming training. Ad Hoc Netw. 2022, 135, 102947. [Google Scholar] [CrossRef]
- Andrews, J.G.; Buzzi, S.; Choi, W.; Hanly, S.V.; Lozano, A.; Soong, A.C.; Zhang, J.C. What will 5G be? IEEE J. Sel. Areas Commun. 2014, 32, 1065–1082. [Google Scholar] [CrossRef]
- Morocho-Cayamcela, M.E.; Lee, H.; Lim, W. Machine Learning for 5G/B5G Mobile and Wireless Communications: Potential, Limitations, and Future Directions. IEEE Access 2019, 7, 137184–137206. [Google Scholar] [CrossRef]
- Wang, Y.; Klautau, A.; Ribero, M.; Soong, A.C.; Heath, R.W. MmWave vehicular beam selection with situational awareness using machine learning. IEEE Access 2019, 7, 87479–87493. [Google Scholar] [CrossRef]
- Rezaie, S.; Manchón, C.N.; de Carvalho, E. Location- and Orientation-Aided Millimeter Wave Beam Selection Using Deep Learning. In Proceedings of the ICC 2020—2020 IEEE International Conference on Communications (ICC), Virtual, 7–11 June 2020; pp. 1–6. [Google Scholar] [CrossRef]
- Antón-Haro, C.; Mestre, X. Learning and data-driven beam selection for mmWave communications: An angle of arrival-based approach. IEEE Access 2019, 7, 20404–20415. [Google Scholar] [CrossRef]
- Li, D.; Wang, S.; Zhao, H.; Wang, X. Context-and-Social-Aware Online Beam Selection for mmWave Vehicular Communications. IEEE Internet Things J. 2021, 8, 8603–8615. [Google Scholar] [CrossRef]
- Yang, Y.; Gao, Z.; Ma, Y.; Cao, B.; He, D. Machine learning enabling analog beam selection for concurrent transmissions in millimeter-wave v2v communications. IEEE Trans. Veh. Technol. 2020, 69, 9185–9189. [Google Scholar] [CrossRef]
- Wang, S.; Li, D.; Zhao, H.; Wang, X. Cross-Layer Data Driven Beam Selection for mmWave Vehicular Communications. In Proceedings of the 2020 International Conference on Wireless Communications and Signal Processing (WCSP) IEEE, Nanjing, China, 21–23 October 2020; pp. 455–460. [Google Scholar]
- Xu, W.; Gao, F.; Tao, X.; Zhang, J.; Alkhateeb, A. Computer Vision Aided mmWave Beam Alignment in V2X Communications. IEEE Trans. Wirel. Commun. 2022, 22, 2699–2714. [Google Scholar] [CrossRef]
- Heng, Y.; Andrews, J.G. Machine Learning-Assisted Beam Alignment for mmWave Systems. In Proceedings of the 2019 IEEE Global Communications Conference (GLOBECOM), Waikoloa, HI, USA, 9–13 December 2019; pp. 1–6. [Google Scholar]
- Wang, Y.; Narasimha, M.; Heath, R.W. MmWave beam prediction with situational awareness: A machine learning approach. In Proceedings of the 2018 IEEE 19th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Kalamata, Greece, 25–28 June 2018; pp. 1–5. [Google Scholar]
- Sim, G.H.; Klos, S.; Asadi, A.; Klein, A.; Hollick, M. An online context-aware machine learning algorithm for 5G mmWave vehicular communications. IEEE ACM Trans. Netw. 2018, 26, 2487–2500. [Google Scholar] [CrossRef]
- Va, V.; Shimizu, T.; Bansal, G.; Heath, R.W. Position-aided millimeter wave V2I beam alignment: A learning-to-rank approach. In Proceedings of the 2017 IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), Montreal, QC, Canada, 8–13 October 2017; pp. 1–5. [Google Scholar]
- Va, V.; Shimizu, T.; Bansal, G.; Heath, R.W. Online Learning for Position-Aided Millimeter Wave Beam Training. IEEE Access 2019, 7, 30507–30526. [Google Scholar] [CrossRef]
- Xu, W.; Gao, F.; Jin, S.; Alkhateeb, A. 3D Scene-Based Beam Selection for mmWave Communications. IEEE Wirel. Commun. Lett. 2020, 9, 1850–1854. [Google Scholar] [CrossRef]
- Salehi, B.; Belgiovine, M.; Sanchez, S.G.; Dy, J.; Ioannidis, S.; Chowdhury, K. Machine learning on camera images for fast mmwave beamforming. In Proceedings of the 2020 IEEE 17th International Conference on Mobile Ad Hoc and Sensor Systems (MASS), Delhi, India, 10–13 December 2020; pp. 338–346. [Google Scholar]
- Lin, C.H.; Kao, W.C.; Zhan, S.Q.; Lee, T.S. BsNet: A Deep Learning-Based Beam Selection Method for mmWave Communications. In Proceedings of the 2019 IEEE 90th Vehicular Technology Conference (VTC2019-Fall), Honolulu, HI, USA, 22–25 September 2019; pp. 1–6. [Google Scholar] [CrossRef]
- Echigo, H.; Cao, Y.; Bouazizi, M.; Ohtsuki, T. A Deep Learning-Based Low Overhead Beam Selection in mmWave Communications. IEEE Trans. Veh. Technol. 2021, 70, 682–691. [Google Scholar] [CrossRef]
- Tauqir, H.P.; Habib, A. Deep learning based beam allocation in switched-beam multiuser massive MIMO systems. In Proceedings of the 2019 Second International Conference on Latest Trends in Electrical Engineering and Computing Technologies (INTELLECT), IEEE, Karachi, Pakistan, 13–14 November 2019; pp. 1–5. [Google Scholar]
- Jagyasi, D.; Coupechoux, M. DNN Based Beam Selection in mmW Heterogeneous Networks. arXiv 2021, arXiv:2102.02672. [Google Scholar]
- Sim, M.S.; Lim, Y.; Park, S.H.; Dai, L.; Chae, C. Deep Learning-Based mmWave Beam Selection for 5G NR/6G With Sub-6 GHz Channel Information: Algorithms and Prototype Validation. IEEE Access 2020, 8, 51634–51646. [Google Scholar] [CrossRef]
- Alrabeiah, M.; Alkhateeb, A. Deep learning for mmWave beam and blockage prediction using Sub-6 GHz channels. IEEE Trans. Commun. 2020, 68, 5504–5518. [Google Scholar] [CrossRef]
- Alrabeiah, M.; Hredzak, A.; Alkhateeb, A. Millimeter wave base stations with cameras: Vision-aided beam and blockage prediction. In Proceedings of the 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), Antwerp, Belgium, 25–28 May 2020; pp. 1–5. [Google Scholar]
- Borges, J.P.T.; De Oliveira, A.P.; Bastos, F.H.B.E.; Suzuki, D.T.N.D.N.; Junior, E.S.D.O.; Bezerra, L.M.; Nahum, C.V.; dos Santos Batista, P.; Júnior, A.B.D.R.K. Reinforcement Learning for Scheduling and Mimo beam Selection using Caviar Simulations. In Proceedings of the 2021 Conference ITU Kaleidoscope: Connecting Physical and Virtual Worlds (ITU K) IEEE, Virtual, 6–10 December 2021; pp. 1–7. [Google Scholar]
- Sapavath, N.N.; Safavat, S.; Rawat, D.B. On the machine learning-based smart beamforming for wireless virtualization with large-scale MIMO system. Trans. Emerg. Telecommun. Technol. 2019, 30, e3713. [Google Scholar] [CrossRef]
- Shafik, W.; Ghasemzadeh, M.; Matinkhah, S.M. A Fast Machine Learning for 5G Beam Selection for Unmanned Aerial Vehicle Applications. J. Inf. Syst. Telecommun. (JIST) 2020, 4, 262. [Google Scholar]
- Dias, M.; Klautau, A.; González-Prelcic, N.; Heath, R.W. Position and LiDAR-aided mmWave beam selection using deep learning. In Proceedings of the 2019 IEEE 20th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Cannes, France, 2–5 July 2019; pp. 1–5. [Google Scholar]
- Klautau, A.; González-Prelcic, N.; Heath, R.W. LiDAR Data for Deep Learning-Based mmWave Beam-Selection. IEEE Wirel. Commun. Lett. 2019, 8, 909–912. [Google Scholar] [CrossRef]
- Ruseckas, J.; Molis, G.; Bogucka, H. MIMO beam selection in 5G using neural networks. Int. J. Electron. Telecommun. 2021, 67, 693–698. [Google Scholar]
- Gu, J.; Salehi, B.; Roy, D.; Chowdhury, K.R. Multimodality in mmWave MIMO Beam Selection Using Deep Learning: Datasets and Challenges. IEEE Commun. Mag. 2022, 60, 36–41. [Google Scholar] [CrossRef]
- Lizarraga, E.M.; Maggio, G.N.; Dowhuszko, A.A. Hybrid beamforming algorithm using reinforcement learning for millimeter wave wireless systems. In Proceedings of the 2019 XVIII Workshop on Information Processing and Control (RPIC) IEEE, Bahia Blanca, Argentina, 18–20 September 2019; pp. 253–258. [Google Scholar]
- Yan, H.; Domae, B.W.; Cabric, D. MmRAPID: Machine learning assisted noncoherent compressive millimeter-wave beam alignment. arXiv 2020, arXiv:2007.12060. [Google Scholar]
- Long, Y.; Chen, Z.; Fang, J.; Tellambura, C. Data-Driven-Based Analog Beam Selection for Hybrid Beamforming Under mm-Wave Channels. IEEE J. Sel. Top. Signal Process. 2018, 12, 340–352. [Google Scholar] [CrossRef]
- Ma, X.; Chen, Z.; Li, Z.; Chen, W.; Liu, K. Low complexity beam selection scheme for terahertz systems: A machine learning approach. In Proceedings of the 2019 IEEE International Conference on Communications Workshops (ICC Workshops), Shanghai, China, 20–24 May 2019; pp. 1–6. [Google Scholar]
- Li, Z.; Ma, X.; Chen, W.; Kuang, N.; Zhang, B. Neural Network Enhanced Analog Beam Selection Scheme for Terahertz Systems. In Proceedings of the 2019 IEEE/CIC International Conference on Communications Workshops in China (ICCC Workshops), Changchun, China, 11–13 August 2019; pp. 136–141. [Google Scholar] [CrossRef]
- Li, X.; Gao, B.; Wang, Y.; Luo, Q.; Shao, S.; Yang, X.; Yan, W.; Wu, H.; Han, B. Compressed Beam Selection for Single/multi-cell Beam Management. In Proceedings of the 2022 IEEE 95th Vehicular Technology Conference: (VTC2022-Spring), Helsinki, Finland, 19–22 June 2022; pp. 1–5. [Google Scholar] [CrossRef]
- Kim, M.; Lee, W.; Cho, D.H. Deep Scanning—Beam Selection Based on Deep Reinforcement Learning in Massive MIMO Wireless Communication System. Electronics 2020, 9, 1844. [Google Scholar] [CrossRef]
- Zhang, J.; Zheng, G.; Krikidis, I.; Zhang, R. Fast specific absorption rate aware beamforming for downlink SWIPT via deep learning. IEEE Trans. Veh. Technol. 2020, 69, 16178–16182. [Google Scholar] [CrossRef]
- Ma, W.; Qi, C.; Li, G.Y. Machine learning for beam alignment in millimeter wave massive MIMO. IEEE Wirel. Commun. Lett. 2020, 9, 875–878. [Google Scholar] [CrossRef]
- Zhao, Y.; Liu, Y.; Boudreau, G.; Sediq, A.B.; Abou-zeid, H.; Wang, X. A two-step neural network based beamforming in MIMO without reference signal. In Proceedings of the 2019 IEEE Global Communications Conference (GLOBECOM), Waikoloa, HI, USA, 9–13 December 2019; pp. 1–6. [Google Scholar]
- Choi, J.; Yerzhanova, M.; Park, J.; Kim, Y.H. Deep learning driven beam selection for orthogonal beamforming with limited feedback. ICT Express 2021, 8, 473–478. [Google Scholar] [CrossRef]
- Krunz, M.; Aykin, I.; Sarkar, S.; Akgun, B. Online Reinforcement Learning for Beam Tracking and Rate Adaptation in Millimeter-wave Systems. IEEE Trans. Mob. Comput. 2023, 1–16. [Google Scholar] [CrossRef]
- Li, Q.; Kannan, A.; Josh, H.; Yoo, T.; Sisk, P.; Boroujeni, M.T.; Pezeshk, H.; Luo, T. Long Short-Term Memory Based Millimeter Wave Beam Change Prediction via Real-World Data. In Proceedings of the 2022 IEEE Globecom Workshops (GC Wkshps), Rio de Janeiro, Brazil, 4–8 December 2022; pp. 1442–1447. [Google Scholar]
- Ngo, H.; Fang, H.; Wang, H. Deep Learning-based Adaptive Beamforming for mmWave Wireless Body Area Network. In Proceedings of the GLOBECOM 2020—2020 IEEE Global Communications Conference, Taipei, Taiwan, 7–11 December 2020; pp. 1–6. [Google Scholar]
- Alkhateeb, A.; Alex, S.; Varkey, P.; Li, Y.; Qu, Q.; Tujkovic, D. Deep learning coordinated beamforming for highly-mobile millimeter wave systems. IEEE Access 2018, 6, 37328–37348. [Google Scholar] [CrossRef]
- Deng, J.; Tian, K.; Zheng, Q.; Bai, J.; Cui, K.; Liu, Y.; Liu, G. Cloud-assisted distributed edge brains for multi-cell joint beamforming optimization for 6G. China Commun. 2022, 19, 36–49. [Google Scholar] [CrossRef]
- Yang, Y.; Deng, X.; He, D.; You, Y.; Song, R. Machine learning inspired codeword selection for dual connectivity in 5G user-centric ultra-dense networks. IEEE Trans. Veh. Technol. 2019, 68, 8284–8288. [Google Scholar] [CrossRef]
- Silva, D.H.; Ribeiro, D.A.; Ramírez, M.A.; Rosa, R.L.; Chaudhary, S.; Rodríguez, D.Z. Selection of Beamforming in 5G MIMO scenarios using Machine Learning approach. In Proceedings of the 2022 19th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON) IEEE, Prachuap Khiri Khan, Thailand, 24–27 May 2022; pp. 1–4. [Google Scholar]
- Ahn, H.; Orikumhi, I.; Kang, J.; Park, H.; Jwa, H.; Na, J.; Kim, S. Machine Learning-based Vision-aided Beam Selection for mmWave Multi-User MISO System. IEEE Wirel. Commun. Lett. 2022, 11, 1263–1267. [Google Scholar] [CrossRef]
- Hua, Z.; Lu, Y.; Pan, G.; Gao, K.; da Costa, D.B.; Chen, S. Computer Vision Aided mmWave UAV Communication Systems. IEEE Internet Things J. 2023, 1. [Google Scholar] [CrossRef]
- Nguyen, T.T.; Nguyen, K.K. A Deep Learning Framework for Beam Selection and Power Control in Massive MIMO—Millimeter-Wave Communications. IEEE Trans. Mob. Comput. 2022, 1. [Google Scholar] [CrossRef]
- Rezaie, S.; de Carvalho, E.; Manchón, C.N. A Deep Learning Approach to Location- and Orientation-Aided 3D Beam Selection for mmWave Communications. IEEE Trans. Wirel. Commun. 2022, 21, 11110–11124. [Google Scholar] [CrossRef]
- Zhang, J.; Xia, W.; You, M.; Zheng, G.; Lambotharan, S.; Wong, K.K. Deep learning enabled optimization of downlink beamforming under per-antenna power constraints: Algorithms and experimental demonstration. IEEE Trans. Wirel. Commun. 2020, 19, 3738–3752. [Google Scholar] [CrossRef]
- Sang-Lim, J.; Nam-Il, K.; Kyung-Seok, K. Machine-Learning-Based User Group and Beam Selection for Coordinated Millimeter-wave Systems. Int. J. Adv. Smart Converg. 2020, 9, 156–166. [Google Scholar]
- Ramezanpour, P.; Rezaei, M.J.; Mosavi, M.R. Deep-learning-based beamforming for rejecting interferences. IET Signal Process. 2020, 14, 467–473. [Google Scholar] [CrossRef]
- Liu, Y.; Zhong, R.; Jaber, M. A Reinforcement Learning Approach for Energy Efficient Beamforming in NOMA Systems. In Proceedings of the GLOBECOM 2022—2022 IEEE Global Communications Conference, Rio de Janeiro, Brazil, 4–8 December 2022; pp. 3827–3832. [Google Scholar] [CrossRef]
- Booth, M.B.; Suresh, V.; Michelusi, N.; Love, D.J. Multi-Armed Bandit Beam Alignment and Tracking for Mobile Millimeter Wave Communications. IEEE Commun. Lett. 2019, 23, 1244–1248. [Google Scholar] [CrossRef]
- Yang, Y.; He, Y.; He, D.; G, Z.; and Luo, Y. Machine Learning based Analog Beam Selection for 5G mmWave Small Cell Networks. In Proceedings of the 2019 IEEE Globecom Workshops (GC Wkshps), Waikoloa, HI, USA, 9–13 December 2019; pp. 1–5. [Google Scholar]
- Jia, C.; Gao, H.; Chen, N.; He, Y. Machine learning empowered beam management for intelligent reflecting surface assisted MmWave networks. China Commun. 2020, 17, 100–114. [Google Scholar] [CrossRef]
- de Figueiredo, F.A.P.; Facina, M.S.P.; Ferreira, R.C.; Ai, Y.; Ruby, R.; Pham, Q.V.; Fraidenraich, G. Large Intelligent Surfaces With Discrete Set of Phase-Shifts Communicating Through Double-Rayleigh Fading Channels. IEEE Access 2021, 9, 20768–20787. [Google Scholar] [CrossRef]
- Klautau, A.; Batista, P.; González-Prelcic, N.; Wang, Y.; Heath, R.W. 5G MIMO data for machine learning: Application to beam selection using deep learning. In Proceedings of the 2018 Information Theory and Applications Workshop (ITA) IEEE, San Diego, CA, USA, 11–16 February 2018; pp. 1–9. [Google Scholar]
- Alrabeiah, M.; Booth, J.; Hredzak, A.; Alkhateeb, A. Viwi vision-aided mmwave beam tracking: Dataset, task, and baseline solutions. arXiv 2020, arXiv:2002.02445. [Google Scholar]
- Alkhateeb, A.; Beltagy, I.; Alex, S. Machine learning for reliable mmwave systems: Blockage prediction and proactive handoff. In Proceedings of the 2018 IEEE Global Conference on Signal and Information Processing (GlobalSIP), Anaheim, CA, USA, 26–29 November 2018; pp. 1055–1059. [Google Scholar]
- Van Quang, B.; Prasad, R.V.; Niemegeers, I. A survey on handoffs-Lessons for 60 GHz based wireless systems. IEEE Commun. Surv. Tutor. 2010, 14, 64–86. [Google Scholar] [CrossRef]
- Sun, L.; Hou, J.; Shu, T. Optimal handover policy for mmwave cellular networks: A multi-armed bandit approach. In Proceedings of the 2019 IEEE Global Communications Conference (GLOBECOM), Waikoloa, HI, USA, 9–13 December 2019; pp. 1–6. [Google Scholar]
- Sutton, R.S.; Barto, A.G. Reinforcement Learning: An Introduction; MIT Press: Cambridge, MA, USA, 2018. [Google Scholar]
- Okamoto, H.; Nishio, T.; Morikura, M.; Yamamoto, K.; Murayama, D.; Nakahira, K. Machine-learning-based throughput estimation using images for mmWave communications. In Proceedings of the 2017 IEEE 85th Vehicular Technology Conference (VTC Spring), Sydney, Australia, 4–7 June 2017; pp. 1–6. [Google Scholar]
- Koda, Y.; Nakashima, K.; Yamamoto, K.; Nishio, T.; Morikura, M. Handover management for mmwave networks with proactive performance prediction using camera images and deep reinforcement learning. IEEE Trans. Cogn. Commun. Netw. 2019, 6, 802–816. [Google Scholar] [CrossRef]
- Koda, Y.; Nakashima, K.; Yamamoto, K.; Nishio, T.; Morikura, M. End-to-end learning of proactive handover policy for camera-assisted mmWave networks using deep reinforcement learning. arXiv 2019, arXiv:1904.04585. [Google Scholar]
- Koda, Y.; Yamamoto, K.; Nishio, T.; Morikura, M. Reinforcement learning based predictive handover for pedestrian-aware mmWave networks. In Proceedings of the IEEE INFOCOM 2018—IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), Virtual, 2–5 May 2018; pp. 692–697. [Google Scholar]
- Koda, Y.; Nakashima, K.; Yamamoto, K.; Nishio, T.; Morikura, M. Cooperative sensing in deep RL-based image-to-decision proactive handover for mmWave networks. In Proceedings of the 2020 IEEE 17th Annual Consumer Communications & Networking Conference (CCNC), Las Vegas, NV, USA, 10–13 January 2020; pp. 1–6. [Google Scholar]
- Mismar, F.B.; Evans, B.L. Partially blind handovers for mmWave new radio aided by sub-6 GHz LTE signaling. In Proceedings of the 2018 IEEE International Conference on Communications Workshops (ICC Workshops), Kansas City, MO, USA, 20–24 May 2018; pp. 1–5. [Google Scholar]
- Yan, L.; Ding, H.; Zhang, L.; Liu, J.; Fang, X.; Fang, Y.; Xiao, M.; Huang, X. Machine learning-based handovers for Sub-6 GHz and mmWave integrated vehicular networks. IEEE Trans. Wirel. Commun. 2019, 18, 4873–4885. [Google Scholar] [CrossRef]
- Guo, D.; Tang, L.; Zhang, X.; Liang, Y.C. Joint Optimization of Handover Control and Power Allocation Based on Multi-Agent Deep Reinforcement Learning. IEEE Trans. Veh. Technol. 2020, 69, 13124–13138. [Google Scholar] [CrossRef]
- Mollel, M.S.; Kaijage, S.; Kisangiri, M.; Imran, M.A.; Abbasi, Q.H. Multi-User Position Based on Trajectories-Aware Handover Strategy for Base Station Selection with Multi-Agent Learning. In Proceedings of the 2020 IEEE International Conference on Communications Workshops (ICC Workshops), Dublin, Ireland, 7–11 June 2020; pp. 1–6. [Google Scholar] [CrossRef]
- Klus, R.; Klus, L.; Solomitckii, D.; Valkama, M.; Talvitie, J. Deep Learning Based Localization and HO Optimization in 5G NR Networks. In Proceedings of the 2020 International Conference on Localization and GNSS (ICL-GNSS), Tampere, Finland, 2–4 June 2020; pp. 1–6. [Google Scholar] [CrossRef]
- Nayakwadi, N.; Fatima, R. Machine Learning based Handover Execution Algorithm for Heterogeneous Wireless Networks. In Proceedings of the 2020 Fifth International Conference on Research in Computational Intelligence and Communication Networks (ICRCICN), Virtual, 26–27 November 2020; pp. 54–58. [Google Scholar] [CrossRef]
- Khosravi, S.; Ghadikolaei, H.S.; Petrova, M. Learning-based load balancing handover in mobile millimeter wave networks. In Proceedings of the GLOBECOM 2020—2020 IEEE Global Communications Conference, Taipei, Taiwan, 7–11 December 2020; pp. 1–7. [Google Scholar]
- Yajnanarayana, V.; Rydén, H.; Hévizi, L. 5G handover using reinforcement learning. In Proceedings of the 2020 IEEE 3rd 5G World Forum (5GWF), Bangalore, India, 10–12 September 2020; pp. 349–354. [Google Scholar]
- Khosravi, S.; Shokri-Ghadikolaei, H.; Petrova, M. Learning-based handover in mobile millimeter-wave networks. IEEE Trans. Cogn. Commun. Netw. 2020, 7, 663–674. [Google Scholar] [CrossRef]
- Sun, Y.; Feng, G.; Qin, S.; Liang, Y.C.; Yum, T.S.P. The SMART handoff policy for millimeter wave heterogeneous cellular networks. IEEE Trans. Mob. Comput. 2017, 17, 1456–1468. [Google Scholar] [CrossRef]
- Sana, M.; De Domenico, A.; Strinati, E.C.; Clemente, A. Multi-agent deep reinforcement learning for distributed handover management in dense mmWave networks. In Proceedings of the ICASSP 2020—2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 4–8 May 2020; pp. 8976–8980. [Google Scholar]
- Alrabeiah, M.; Zhang, Y.; Alkhateeb, A. Neural Networks Based Beam Codebooks: Learning mmWave Massive MIMO Beams that Adapt to Deployment and Hardware. arXiv 2020, arXiv:2006.14501. [Google Scholar] [CrossRef]
- Bhogi, K.; Saha, C.; Dhillon, H.S. Learning on a Grassmann Manifold: CSI Quantization for Massive MIMO Systems. arXiv 2020, arXiv:2005.08413. [Google Scholar]
- Linde, Y.; Buzo, A.; Gray, R. An algorithm for vector quantizer design. IEEE Trans. Commun. 1980, 28, 84–95. [Google Scholar] [CrossRef]
- Jiang, J.; Wang, X.; Wang, W.J.; Zhen, L.; Wang, J. Deep Clustering-Based Codebook Design for Massive MIMO Systems. IEEE Access 2019, 7, 172654–172664. [Google Scholar] [CrossRef]
- Zhang, Y.; Alrabeiah, M.; Alkhateeb, A. Learning beam codebooks with neural networks: Towards environment-aware mmWave MIMO. In Proceedings of the 2020 IEEE 21st International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Atlanta, GA, USA, 26–29 May 2020; pp. 1–5. [Google Scholar]
- Chen, J.C. Efficient codebook-based beamforming algorithm for millimeter-wave massive MIMO systems. IEEE Trans. Veh. Technol. 2017, 66, 7809–7817. [Google Scholar] [CrossRef]
- Zhang, Y.; Alrabeiah, M.; Alkhateeb, A. Reinforcement Learning of Beam Codebooks in Millimeter Wave and Terahertz MIMO Systems. arXiv 2021, arXiv:2102.11392. [Google Scholar] [CrossRef]
- Jiang, J.; Wang, X.; Sidhu, G.A.S.; Zhen, L.; Gao, R. Clustering-based codebook design for MIMO communication system. In Proceedings of the ICC 2019—2019 IEEE International Conference on Communications (ICC), Shanghai, China, 21–23 May 2019; pp. 1–6. [Google Scholar]
- Lee, H.; Girnyk, M.; Jeong, J. Deep reinforcement learning approach to MIMO precoding problem: Optimality and Robustness. arXiv 2020, arXiv:2006.16646. [Google Scholar]
- Takabe, S.; Imanishi, M.; Wadayama, T.; Hayakawa, R.; Hayashi, K. Trainable projected gradient detector for massive overloaded MIMO channels: Data-driven tuning approach. IEEE Access 2019, 7, 93326–93338. [Google Scholar] [CrossRef]
- He, H.; Wen, C.K.; Jin, S.; Li, G.Y. A model-driven deep learning network for MIMO detection. In Proceedings of the 2018 IEEE Global Conference on Signal and Information Processing (GlobalSIP), Anaheim, CA, USA, 26–29 November 2018; pp. 584–588. [Google Scholar]
- Balatsoukas-Stimming, A.; Castañeda, O.; Jacobsson, S.; Durisi, G.; Studer, C. Neural-network optimized 1-bit precoding for massive MU-MIMO. In Proceedings of the 2019 IEEE 20th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Cannes, France, 2–5 July 2019; pp. 1–5. [Google Scholar]
- Jiang, J.; Lu, D.; Zhen, L. Codebook Design Based on Self-Organizing Map Clustering for Limited Feedback MIMO Systems. In Proceedings of the 2019 IEEE/CIC International Conference on Communications in China (ICCC), Changchun, China, 11–13 August 2019; pp. 146–151. [Google Scholar]
- Kang, J.; Choi, W. Novel codebook design for channel state information quantization in MIMO rician fading channels with limited feedback. IEEE Trans. Signal Process. 2021, 69, 2858–2872. [Google Scholar] [CrossRef]
- Hojatian, H.; Nadal, J.; Frigon, J.F.; Leduc-Primeau, F. Unsupervised Deep Learning for Massive MIMO Hybrid Beamforming. arXiv 2020, arXiv:2007.00038. [Google Scholar] [CrossRef]
- Kwon, H.J.; Lee, J.H.; Choi, W. Machine learning-based beamforming in K-user MISO interference channels. IEEE Access 2021, 9, 28066–28075. [Google Scholar] [CrossRef]
- Biswas, S.; Singh, U.; Nag, K. Multi-Layer Perceptron-based Beamformer Design for Next-Generation Full-Duplex Cellular Systems. In Proceedings of the 2021 IEEE/ACIS 22nd International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing (SNPD), Taichung, Taiwan, 24–26 November 2021; pp. 49–55. [Google Scholar]
- Xia, W.; Zheng, G.; Zhu, Y.; Zhang, J.; Wang, J.; Petropulu, A.P. A deep learning framework for optimization of MISO downlink beamforming. IEEE Trans. Commun. 2019, 68, 1866–1880. [Google Scholar] [CrossRef]
- Liu, G.; Deng, H.; Yang, K.; Zhu, Z.; Liu, J.; Dong, H. A New Design of Codebook for Hybrid Precoding in Millimeter-Wave Massive MIMO Systems. Symmetry 2021, 13, 743. [Google Scholar] [CrossRef]
- Ma, W.; Qi, C.; Zhang, Z.; Cheng, J. Sparse channel estimation and hybrid precoding using deep learning for millimeter wave massive MIMO. IEEE Trans. Commun. 2020, 68, 2838–2849. [Google Scholar] [CrossRef]
- Elbir, A.M.; Papazafeiropoulos, A.K. Hybrid precoding for multiuser millimeter wave massive MIMO systems: A deep learning approach. IEEE Trans. Veh. Technol. 2019, 69, 552–563. [Google Scholar] [CrossRef]
- Huang, H.; Song, Y.; Yang, J.; Gui, G.; Adachi, F. Deep-learning-based millimeter-wave massive MIMO for hybrid precoding. IEEE Trans. Veh. Technol. 2019, 68, 3027–3032. [Google Scholar] [CrossRef]
- Sohrabi, F.; Attiah, K.M.; Yu, W. Deep learning for distributed channel feedback and multiuser precoding in FDD massive MIMO. IEEE Trans. Wirel. Commun. 2021, 20, 4044–4057. [Google Scholar] [CrossRef]
- Eappen, G.; Cosmas, J.; T, S.; A, R.; Nilavalan, R.; Thomas, J. Deep learning integrated reinforcement learning for adaptive beamforming in B5G networks. IET Commun. 2022, 16, 2454–2466. [Google Scholar] [CrossRef]
- Aljumaily, M.S.; Li, H. Machine Learning Aided Hybrid Beamforming in Massive-MIMO Millimeter Wave Systems. In Proceedings of the 2019 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), Newark, NJ, USA, 11–14 November 2019; pp. 1–6. [Google Scholar]
- Sun, Y.; Gao, Z.; Wang, H.; Wu, D. Machine learning based hybrid precoding for mmWave MIMO-OFDM with dynamic subarray. arXiv 2018, arXiv:1809.03378. [Google Scholar]
- Kang, J.; Lee, J.H.; Choi, W. Machine Learning-Based Dimension Optimization for Two-Stage Precoder in Massive MIMO Systems with Limited Feedback. Appl. Sci. 2019, 9, 2894. [Google Scholar] [CrossRef]
- Attiah, K.M.; Sohrabi, F.; Yu, W. Deep Learning Approach to Channel Sensing and Hybrid Precoding for TDD Massive MIMO Systems. arXiv 2020, arXiv:2011.10709. [Google Scholar]
- Li, X.; Huang, Y.; Heng, W.; Wu, J. Machine Learning-Inspired Hybrid Precoding for mmWave MU-MIMO Systems with Domestic Switch Network. Sensors 2021, 21, 3019. [Google Scholar] [CrossRef]
- Jiang, T.; Cheng, H.V.; Yu, W. Learning to Beamform for Intelligent Reflecting Surface with Implicit Channel Estimate. arXiv 2020, arXiv:2009.14404. [Google Scholar]
- Le Magoarou, L.; Yassine, T.; Paquelet, S.; Crussière, M. Deep learning for location based beamforming with NLoS channels. In Proceedings of the ICASSP 2022—2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Singapore, 22–27 May 2022; pp. 8812–8816. [Google Scholar]
- Huang, S.; Ye, Y.; Xiao, M. Hybrid beamforming for millimeter wave multi-user MIMO systems using learning machine. IEEE Wirel. Commun. Lett. 2020, 9, 1914–1918. [Google Scholar] [CrossRef]
- Almagboul, M.A.; Shu, F.; Abdelgader, A.M. Deep-Learning-Based Phase-Only Robust Massive MU-MIMO Hybrid Beamforming. IEEE Commun. Lett. 2021, 25, 2280–2284. [Google Scholar] [CrossRef]
- Li, X.; Alkhateeb, A. Deep learning for direct hybrid precoding in millimeter wave massive MIMO systems. In Proceedings of the IEEE 2019 53rd Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, USA, 3–6 November 2019; pp. 800–805. [Google Scholar]
- Zheng, Z.; Wang, L.; Zhu, F.; Liu, L. Potential Technologies and Applications Based on Deep Learning in the 6G Networks. Comput. Electr. Eng. 2021, 95, 107373. [Google Scholar] [CrossRef]
- Goodfellow, I.; McDaniel, P.; Papernot, N. Making Machine Learning Robust against Adversarial Inputs. Commun. ACM 2018, 61, 56–66. [Google Scholar] [CrossRef]
- Geiping, J.; Fowl, L.; Huang, W.R.; Czaja, W.; Taylor, G.; Moeller, M.; Goldstein, T. Witches’ Brew: Industrial Scale Data Poisoning via Gradient Matching. arXiv 2020, arXiv:2009.02276. [Google Scholar] [CrossRef]
- El-Mhamdi, E.M.; Farhadkhani, S.; Guerraoui, R.; Guirguis, A.; Hoang, L.N.; Rouault, S. Collaborative Learning in the Jungle (Decentralized, Byzantine, Heterogeneous, Asynchronous and Nonconvex Learning). In Advances in Neural Information Processing Systems; Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P., Vaughan, J.W., Eds.; Curran Associates, Inc.: New York, NY, USA, 2021; Volume 34, pp. 25044–25057. [Google Scholar]
- Tramèr, F.; Zhang, F.; Juels, A.; Reiter, M.K.; Ristenpart, T. Stealing Machine Learning Models via Prediction APIs. In Proceedings of the 25th USENIX Security Symposium (USENIX Security 16), Austin, TX, USA, 10–12 August 2016; pp. 601–618. [Google Scholar]
- Catak, E.; Catak, F.O.; Moldsvor, A. Adversarial machine learning security problems for 6G: MmWave beam prediction use-case. In Proceedings of the 2021 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom), Sofia, Bulgaria, 6–9 June 2021; pp. 1–6. [Google Scholar]
- Kuzlu, M.; Catak, F.O.; Cali, U.; Catak, E.; Guler, O. The Adversarial Security Mitigations of mmWave Beamforming Prediction Models using Defensive Distillation and Adversarial Retraining. arXiv 2022, arXiv:2202.08185. [Google Scholar] [CrossRef]
- Meng, H.; Shafik, W.; Matinkhah, S.M.; Ahmad, Z. A 5 g beam selection machine learning algorithm for unmanned aerial vehicle applications. Wirel. Commun. Mob. Comput. 2020, 2020, 1428968. [Google Scholar] [CrossRef]
- I, C.L.; Huang, J.; Duan, R.; Cui, C.; Jiang, J.; Li, L. Recent Progress on C-RAN Centralization and Cloudification. IEEE Access 2014, 2, 1030–1039. [Google Scholar] [CrossRef]
- Park, S.H.; Lee, K.J.; Song, C.; Lee, I. Joint Design of Fronthaul and Access Links for C-RAN With Wireless Fronthauling. IEEE Signal Process. Lett. 2016, 23, 1657–1661. [Google Scholar] [CrossRef]
- Yu, D.; Lee, H.; Park, S.H.; Hong, S.E. Deep Learning Methods for Joint Optimization of Beamforming and Fronthaul Quantization in Cloud Radio Access Networks. IEEE Wirel. Commun. Lett. 2021, 10, 2180–2184. [Google Scholar] [CrossRef]
- Zhong, C.H.; Guo, K.; Zhao, M. Online Sparse Beamforming in C-RAN: A Deep Reinforcement Learning Approach. In Proceedings of the 2021 IEEE Wireless Communications and Networking Conference (WCNC), Nanjing, China, 29 March–1 April 2021; pp. 1–6. [Google Scholar] [CrossRef]
- Costa, L.R.; Silva, Y.C.; Lima, F.R.M.; Klein, A. Beam allocation based on spatial compatibility for hybrid beamforming C-RAN networks. In Proceedings of the WSA 2019—23rd International ITG Workshop on Smart Antennas VDE, Vienna, Austria, 24–26 April 2019; pp. 1–6. [Google Scholar]
- Pan, C.; Ren, H.; Elkashlan, M.; Nallanathan, A.; Hanzo, L. Robust beamforming design for ultra-dense user-centric C-RAN in the face of realistic pilot contamination and limited feedback. IEEE Trans. Wirel. Commun. 2018, 18, 780–795. [Google Scholar] [CrossRef]
- Bjornson, E. Reproducible research: Best practices and potential misuse [perspectives]. IEEE Signal Process. Mag. 2019, 36, 106–123. [Google Scholar] [CrossRef]
- Gundersen, O.E.; Gil, Y.; Aha, D.W. On reproducible AI: Towards reproducible research, open science, and digital scholarship in AI publications. AI Mag. 2018, 39, 56–68. [Google Scholar] [CrossRef]
- Huang, H.; Xia, W.; Xiong, J.; Yang, J.; Zheng, G.; Zhu, X. Unsupervised Learning-Based Fast Beamforming Design for Downlink MIMO. IEEE Access 2019, 7, 7599–7605. [Google Scholar] [CrossRef]
- Chiu, S.E.; Ronquillo, N.; Javidi, T. Active Learning and CSI Acquisition for mmWave Initial Alignment. IEEE J. Sel. Areas Commun. 2019, 37, 2474–2489. [Google Scholar] [CrossRef]
- Ding, J.; Tramel, E.; Sahu, A.K.; Wu, S.; Avestimehr, S.; Zhang, T. Federated Learning Challenges and Opportunities: An Outlook. arXiv 2022, arXiv:2202.00807. [Google Scholar] [CrossRef]
- Elbir, A.M.; Coleri, S. Federated Learning for Hybrid Beamforming in mm-Wave Massive MIMO. arXiv 2020, arXiv:2005.09969. [Google Scholar] [CrossRef]
- Ren, K.; Zheng, T.; Qin, Z.; Liu, X. Adversarial Attacks and Defenses in Deep Learning. Engineering 2020, 6, 346–360. [Google Scholar] [CrossRef]
- Sakurada, M.; Yairi, T. Anomaly Detection Using Autoencoders with Nonlinear Dimensionality Reduction. In Proceedings of the MLSDA 2014 2nd Workshop on Machine Learning for Sensory Data Analysis, MLSDA’14, New York, NY, USA, 2 December 2014; pp. 4–11. [Google Scholar] [CrossRef]
- Thing, V.L.L. IEEE 802.11 Network Anomaly Detection and Attack Classification: A Deep Learning Approach. In Proceedings of the 2017 IEEE Wireless Communications and Networking Conference (WCNC), San Francisco, CA, USA, 19–22 March 2017; pp. 1–6. [Google Scholar] [CrossRef]
- Simpson, R. Computer vision: An overview. IEEE Expert 1991, 6, 11–15. [Google Scholar] [CrossRef]
- Alrabeiah, M.; Hredzak, A.; Alkhateeb, A. Millimeter Wave Base Stations with Cameras: Vision Aided Beam and Blockage Prediction. arXiv 2019, arXiv:1911.06255. [Google Scholar]
- Roy, D.; Salehi, B.; Banou, S.; Mohanti, S.; Reus-Muns, G.; Belgiovine, M.; Ganesh, P.; Bocanegra, C.; Dick, C.; Chowdhury, K. Going Beyond RF: How AI-enabled Multimodal Beamforming will Shape the NextG Standard. arXiv 2022, arXiv:2203.16706. [Google Scholar] [CrossRef]
- Charan, G.; Alrabeiah, M.; Alkhateeb, A. Vision-Aided 6G Wireless Communications: Blockage Prediction and Proactive Handoff. arXiv 2021, arXiv:2102.09527. [Google Scholar] [CrossRef]
- Xie, H.; Feng, D.; Yu, H. Fast and robust adaptive beamforming method based on l1-norm constraint for large array. Electron. Lett. 2015, 51, 98–99. [Google Scholar] [CrossRef]
- Hussain, M.; Scalabrin, M.; Rossi, M.; Michelusi, N. Adaptive Millimeter-Wave Communications Exploiting Mobility and Blockage Dynamics. In Proceedings of the ICC 2020—2020 IEEE International Conference on Communications (ICC), London, UK, 7–11 June 2020; pp. 1–6. [Google Scholar] [CrossRef]
- Lin, C.; Li, G.Y.L. Terahertz communications: An array-of-subarrays solution. IEEE Commun. Mag. 2016, 54, 124–131. [Google Scholar] [CrossRef]
- Noh, S.; Lee, J.; Lee, G.; Seo, K.; Sung, Y.; Yu, H. Channel estimation techniques for RIS-assisted communication: Millimeter-wave and sub-THz systems. IEEE Veh. Technol. Mag. 2022, 17, 64–73. [Google Scholar] [CrossRef]
- Alkhateeb, A.; El Ayach, O.; Leus, G.; Heath, R.W. Channel estimation and hybrid precoding for millimeter wave cellular systems. IEEE J. Sel. Top. Signal Process. 2014, 8, 831–846. [Google Scholar] [CrossRef]
- Tanveer, J.; Haider, A.; Ali, R.; Kim, A. Machine learning for physical layer in 5G and beyond wireless networks: A survey. Electronics 2022, 11, 121. [Google Scholar] [CrossRef]
- Chen, M.; Challita, U.; Saad, W.; Yin, C.; Debbah, M. Machine learning for wireless networks with artificial intelligence: A tutorial on neural networks. arXiv 2017, arXiv:1710.02913. [Google Scholar]
- Dai, L.; Jiao, R.; Adachi, F.; Poor, H.V.; Hanzo, L. Deep learning for wireless communications: An emerging interdisciplinary paradigm. IEEE Wirel. Commun. 2020, 27, 133–139. [Google Scholar] [CrossRef]
- Salh, A.; Audah, L.; Shah, N.S.M.; Alhammadi, A.; Abdullah, Q.; Kim, Y.H.; Al-Gailani, S.A.; Hamzah, S.A.; Esmail, B.A.F.; Almohammedi, A.A. A survey on deep learning for ultra-reliable and low-latency communications challenges on 6G wireless systems. IEEE Access 2021, 9, 55098–55131. [Google Scholar] [CrossRef]
- Fourati, H.; Maaloul, R.; Chaari, L. A survey of 5G network systems: Challenges and machine learning approaches. Int. J. Mach. Learn. Cybern. 2021, 12, 385–431. [Google Scholar] [CrossRef]
- Jagannath, J.; Polosky, N.; Jagannath, A.; Restuccia, F.; Melodia, T. Machine learning for wireless communications in the Internet of Things: A comprehensive survey. Ad Hoc Netw. 2019, 93, 101913. [Google Scholar] [CrossRef]
- Mao, Q.; Hu, F.; Hao, Q. Deep learning for intelligent wireless networks: A comprehensive survey. IEEE Commun. Surv. Tutor. 2018, 20, 2595–2621. [Google Scholar] [CrossRef]
- Xiong, Z.; Zhang, Y.; Niyato, D.; Deng, R.; Wang, P.; Wang, L.C. Deep reinforcement learning for mobile 5G and beyond: Fundamentals, applications, and challenges. IEEE Veh. Technol. Mag. 2019, 14, 44–52. [Google Scholar] [CrossRef]
- Huang, H.; Guo, S.; Gui, G.; Yang, Z.; Zhang, J.; Sari, H.; Adachi, F. Deep Learning for Physical-Layer 5G Wireless Techniques: Opportunities, Challenges and Solutions. IEEE Wirel. Commun. 2020, 27, 214–222. [Google Scholar] [CrossRef]
Database | Date of Search | Search Strings | Number of Selected Papers |
---|---|---|---|
Google Scholar | March 2021 | “machine learning”, “beam selection” | 36 |
April 2021 | “machine learning”, “codebook”, “mimo” | 21 | |
April 2022 | “beamforming”, “machine learning” | 7 | |
“beamforming”, “artificial intelligence” | 5 | ||
“beam selection”, “machine learning” | 6 | ||
“machine learning”, “beam selection”, “mmwave” | 16 | ||
“machine learning”, “handover”, “mmwave” | 29 | ||
December 2022 | “Beamforming”, “Beam-selection”, “machine learning”, “artificial intelligence” | 25 | |
IEEE Explore | April 2021 | “Beam selection”, “machine learning”, “artificial intelligence” | 36 |
Paper | ML-Enabled | Frequency | Antenna | Radio Transmission | Mobility Support | Array Type |
---|---|---|---|---|---|---|
Araujo et al. [26] | ◯ | ⚫ | ⚫ | ⚫ | ⚫ | ◯ |
Zardi et al. [61] | ⚫ | ⚫ | ◯ | ◯ | ⚫ | ⚫ |
Pham et al. [27] | ◯ | ⚫ | ◯ | ⚫ | ◯ | ◯ |
Murray et al. [56] | ◯ | ◯ | ⚫ | ⚫ | ◯ | ⚫ |
Naeem et al. [28] | ⚫ | ◯ | ◯ | ◯ | ◯ | ◯ |
ElHalawany et al. [63] | ⚫ | ◯ | ◯ | ◯ | ◯ | ◯ |
Wu et al. [64] | ⚫ | ◯ | ⚫ | ◯ | ◯ | ⚫ |
Kassir et al. [65] | ⚫ | ◯ | ⚫ | ◯ | ◯ | ⚫ |
Nor et al. [66] | ⚫ | ◯ | ◯ | ◯ | ⚫ | ◯ |
Ours | ⚫ | ⚫ | ⚫ | ⚫ | ⚫ | ⚫ |
Challenges | Algorithm | Highlight (Pros) | Limitations (Cons) | Key Contribution | Ref. |
---|---|---|---|---|---|
Situational awareness |
● RBF-SVM ● GB ● RF ● FNN | Leverages situational awareness, such as vehicle coordinates, type, and speed. |
● Requires the neighboring vehicles to be connected to the network for best accuracy. ● Localization at the BS can be outdated or incomplete due to mobility, limited location reporting frequencies, and being connected to the BS. | This paper evaluates different coordinate systems and several levels of available side information. | [69] |
● Regression ● SVM ● RF-R ● GB |
● Leverages situational awareness. ● Requires low overhead. | The lack of information on trucks’ positions has a large impact on the method’s performance. | Proposes predicting the received power with different beam power quantizations using regression models through situational awareness. | [77] | |
● RF-C ● MLP ● SVM ● Adaboost | The classification models have smaller feedback and better overall performance. |
● Regression models require feedback. ● Reliance on ray-tracing may limit the generalizability of the results to other scenarios. ● Accurate UE localization may not always be feasible in practice. ● It does not consider other factors that may affect the overall performance of mmWave systems, such as interference and mobility. | Proposes optimal access point and beam pair predictions for establishing communication by exploiting the UE’s localization and machine learning tools. | [76] | |
CSML |
● Shows that context and social information of vehicles and passengers are relevant for beam allocation. ● The results show improvements in the received data. |
● Only permanent blockage is considered. ● It does not evaluate or compare the proposed algorithm with existing methods or benchmarks. | This paper brings a double-layer online learning algorithm based on user context and social preference information. | [74] | |
RL | Using only GNSS data, the ML algorithm has a good beam prediction accuracy. The accuracy is improved further when LiDAR data are considered. |
● The reliance on GNSS and LiDAR for context information may be limited in some scenarios, affecting its performance. ● Beam prediction with LiDAR data is computationally demanding. | Investigates the use of GNSS and GNSS + LiDAR data for beam selection with NN using Raymobtime datasets. | [95] | |
FML |
● Low-complexity and scalable online learning algorithm. ● Does not require either accurate or previous information. ● Proposes a real scenario protocol for supporting the mechanism. |
● The algorithm relies on GPS coordinates, which can be inaccurate in domestic devices. ● A standard-compliant protocol may not be available or compatible with all types of mmWave vehicular systems. | An online learning algorithm based on CMAB is proposed, enabling the mmWave BS to learn from the context autonomously, and it provides a scalable solution to increase the deployment density of mmWave BS. | [78] | |
MAB |
● Defines an exploration and exploitation algorithm for each algorithm layer. ● Criticizes the model limitations. ● Efficiently broadcasts content to UEs with the same interest, maintaining the SNR. |
● Does not specify the control functions for exploration timing. ● The content is only related to movies, which might be of limited scope for real scenarios. |
● Uses a two-layer RL online algorithm to learn surrounding blockages from context information instead of using the CSI. ● The algorithm aggregates UEs with interest in the same content through beam broadcasting. | [72] | |
Position-aided | DNN |
● Reduced outage and beam misalignment probability. |
● Having access to user data might be difficult. ● Needs large training datasets. ● Depends on precise and consistent receiver location and orientation information, which might not be consistently or always accessible. | The results vary with the number of obstacles for training and test datasets, highlighting the robustness of train–test mismatch. | [70] |
MAB |
● The algorithm assumes errors in the positioning coordinates. ● Two mechanisms are proposed for beam pair selection, greedy and risk-aware. ● The authors also propose a beam pair refinement based on hierarchical optimization. |
● The paper lacks a discussion on practical implementations and the algorithm’s computational complexity. ● Simulated evaluation may not reflect real-world scenarios with varying conditions and user behaviors. | Proposes an online method for beam selection to speed up the process. | [80] | |
LtR |
● The authors define the scoring and ranking functions to determine the best beam pairs. ● A communication concise protocol is described as an example for implementing the technique in a real scenario. |
● The offline training requires careful data updates and periodic re-training. ● The baseline algorithm is not well-explained. | Authors use context information and past beam measurements to determine potential beam pointing directions. | [79] | |
● CNN ● DNN | The algorithm presents high accuracy for low-resolution images. |
● No comparison with other ML techniques is provided. ● The images used as the input are unusual. ● The positioning info, when available, could be used instead for a simpler system. | Proposes a CNN algorithm for beam selection and switching using low-resolution images as the input. | [85] | |
Angle-of-arrival-aided |
● KNN ● DNN ● Singular vector class | Evaluates the impact of using imperfect and realistic information for the AoA and received power estimation by using Capon and MUSIC estimation methods. |
● The BS performance degrades for a low number of UEs compared to the available antennas. ● The performance of the proposed approach is heavily dependent on the quality of the CSI estimates, specifically the AoAs and powers. | Proposes the use of AoA and received power as the input of a DNN to select the best beamformer on a codebook rather than the complete channel matrix, which is a realistic approach. | [71] |
Vehicular networks | SVM | Higher sum-rate and lower complexity than channel estimation-based method. | The training depends on the link density, which is hard to estimate and may vary substantially in real scenarios. | Proposes a tailored SVM/SMO algorithm for beam training. | [73] |
3D-scene-based | DNN | The 3D scene reconstruction achieves better accuracy than LiDAR, which is more expensive. |
● The UE coordinate estimation can be erroneous. ● Computational complexity is not evaluated. | In this paper, a 3D scene reconstruction is used to identify the best beams. | [81] |
Beam domain image reconstruction |
● CNN ● DNN | Reduced beam selection overhead without degrading the beamforming performance. |
● The training is based on historical data. ● Simulated evaluation may not reflect real-world scenarios with varying conditions and user behaviors. | This paper treats the beam selection as an image reconstruction problem without requiring channel knowledge. | [83] |
Low overhead | LSTM | The proposed scheme finds the narrow best beam based only on wide beam measurements, reducing the beam training overhead. |
● Only a DFT codebook is tested as both a high and low-resolution codebook. ● Lacks a thorough investigation of the proposed deep-learning-based model’s generalization, which may limit its applicability in real-world scenarios. | This paper proposes a DL-based low-overhead analog beam selection scheme. | [84] |
Sub-6GHz channel information | DNN |
● Detailed DNN description. ● Good accuracy with a partial dataset. | Lacks comparisons with other algorithms using the same scenario (i.e., DeepMIMO O1). | This paper relies on sub-6GHz channel information to deduce the resources in the mmWave band. | [86] |
DNN |
● Compares the results with prior work. ● Robust to NLoS conditions. |
● Marginal gain increasing the number of neurons. ● Implementation cost and energy efficiency not taken into account. | A dual-band scheme to predict beam and blockage from the sub-6GHz band to aid in the mmWave band. | [88] | |
DNN | Presents a prototype validation of an indoor scenario, which shows that the ray-tracing and the beam selection method are very close to the real scenario. |
● The sub-6GHz channel was modeled like a SISO channel. ● Although there is a prototype validation, the results are not compared with any other ML-based beam selection approaches. | The PDP of the sub-6 GHz channel, which is highly available and does not demand beamforming, was used as the input of a DNN for beam selection estimation in indoor and outdoor scenarios. | [87] | |
Blockage prediction | CNN | The use of RGB images reduces beam selection and blockage prediction overhead. |
● High training complexity. ● Simplistic scenario. ● It does not work in dynamic environments. | Joins images and sub-6GHz channel information to identify mmWave blocked users. | [89] |
Intercarrier Interference (ICI) Mitigation | DNN | Low computation time, yet high-spectral-efficiency algorithm. | The paper lacks profound analysis for more users and if the grouping is effective. | This paper proposes an optimal user group beam selection scheme aiming at the spectral efficiency maximization. | [120] |
Small cell networks | SVM | Reduced complexity with quick and high Average Sum-Rate (ASR) in the BS. |
● Though the paper assumes analog beamforming, the side-lobe interference is ignored. ● Assumes the Poisson point process for small-cell BS distribution, which may not accurately represent all scenarios and limit applicability. | This paper aims to maximize ASR for concurrent transmission on an analog beamforming mmWave network by analyzing the BS spatial distribution. | [124] |
LiDAR data | DNN | High accuracy for top-M beam selection classification. |
● The one LiDAR per vehicle premise is not feasible due to LiDAR cost. ● Assumes that the LiDAR data are perfect, but in reality, LiDAR data may be noisy or incomplete, which can affect the performance of the proposed approach. | Proposes using LiDAR information to select beams in vehicular applications using deep learning, comparing centralized and distributed LiDAR. | [93] |
CNN | The use of LiDAR data reduces beam selection overhead for LoS situations. |
● Overhead increases on NLoS occasions to maintain a tolerable throughput ratio. ● Assumes that the LiDAR data are perfect, but in reality, they may be noisy or incomplete, which can affect the performance of the proposed approach. | The use of LiDAR data with CNN reduces the beam selection overhead for Vehicle-to-Infrastructure (V2I) communications. | [94] | |
DL |
● Uses multiple sensors, such as cameras, LiDAR, and GPS ● Accuracy improved and delay decreased when compared to IEEE 802.11ad | The measurement setup is complex and hard to reproduce. | Establishes guidelines for beam selection dataset generation and releases its dataset and paper results | [96] | |
IRS-assisted beam selection | DL |
● The beam management mechanism utilizes user positioning information and environmental information to build reliable beam selection. ● Enabling mobility is approached through updates on the historical database. |
● The algorithm depends on BS-UE and UE computational capabilities to provide full mobility awareness. ● Lacks real-world evaluation, which is crucial for validating its effectiveness. | This work presents an IRS-assisted mmWave network to improve coverage, handover, and beam-switching. | [125] |
Channel data generation |
● SVM ● Adaboost ● DNN ● DQL ● Decision tree | Beam selection performance is simulated for several classification methods. |
● Focuses only on data generation and classification methods for beam selection. ● Relies on a combination of vehicle traffic and ray-tracing simulators to generate channel realizations, which might affect the performance of the proposed methodology. | Describes a methodology for generating mmWave channel data, including realistic traffic simulation. | [127] |
Low-complexity | SVM | The computational complexity of the proposed data-driven approach is significantly lower than the sub-optimization method. |
● The number of analog beams considered is too small. ● Algorithm requires a diverse and large mmWave channel training dataset, which may not be obtainable or representative, limiting its applicability. | The authors propose a novel method, called biased-SVM, that determines the optimal parameter of the Gaussian kernel function to achieve optimal beam selection with low computational complexity. | [99] |
RF-C | The model complexity decreases as the number of users increases and is lower than the other compared methods, which is an advantage for delay-sensitive applications. | The simulation tool is not mentioned, which inhibits the results’ reproductivity. ● The algorithm only tackles hybrid beamforming-based beam selection for THz systems, ignoring other THz communication challenges such as interference, mobility, and handover. |
● Authors compared the computational complexity with a large number of users. ● The results show a better trade-off between computational complexity and system performance compared to exhaustive and SVM approaches. | [100] | |
DL | The authors propose a sampling method, reducing the number of seeped beams, and the DL predicts all beams, increasing the search space for the beam selection |
● The beam combination method cannot be generalized, so in practice, each scenario may require a different combination. ● Does not consider the robustness of the proposed approach to channel variations, such as changes in the environment, mobility, or interference. | A method for sampling a fraction of the beam pairs is proposed, combined with a DL for predicting the RSRP of all beams from the samples. | [102] | |
RBF | Reduces complexity by several orders of magnitude, with near-optimal performance, compared with conventional methods. |
● Needs large training datasets. ● Performance depends on the dataset size. ● Does not provide a detailed analysis of how the size of the training data affects the performance of the proposed approach. |
● In this work, the authors propose using an RBF-NN model to perform the beam selection procedure. ● Reduction in the complexity, beam selection overhead, and latency. | [101] | |
Q-learning | The performance is close to the optimal solution, but takes fewer iterations. |
● Depends on knowledge of the channel matrix. ● Assuming codebook-based analog beamforming may limit algorithm flexibility. | The paper minimizes the training time for beam selection using Q-learning to find the best-quantized analog precoders. | [97] | |
DNN | This approach is appropriate for practical massive MIMO systems due to the complexity of the algorithm, which is not proportional to the number of beamforming vectors, using only one pilot signal. |
● Good accuracy is only achieved for a large number of training epochs. ● The capacity comparisons do not include other beam selection mechanisms. | This work proposes a novel algorithm (named Deep Scanning) based on deep Q-learning. | [103] | |
CNN |
● The model-driven solution proposed in this work reduces the computational complexity and execution time of the data-driven technique. ● Authors include optimal solutions, providing upper bounds for the simulation. |
● Assumes perfect complex channel matrices as the input, which can be hard to obtain in a real scenario. ● Relies on simulations to demonstrate its effectiveness, and its performance in real-world scenarios is unknown. | Proposes a novel model-driven technique based on CNN, which calculates only the essential and passes it through a low-complex beamforming recovery algorithm. | [104] | |
Body area network | GAN | Authors generated a dataset for WBAN based on a human pose dataset used for computer vision. |
● Does not address how the beam prediction would be made without an external camera, and only one set of sensors’ location is provided. ● Only considers a limited number of body types and poses, and it is not clear whether the approach can be generalized to different body types and poses. | Proposes employing a non-intrusive beamforming method in the WBAN with the use of the GAN method for mmWave beam predictions using human pose images. | [110] |
Highly mobile systems | DNN | Authors develop low-complexity mmWave coordination strategies for coverage coordination and latency reduction using omni-directional + directional beams in the offline training phase and only omni-directional transmission in the testing phase. |
● Single user and simplistic channel scenario. ● Although the effective achievable rate is greater than the baseline, the proposed method is more sensible to NLoS scenarios as the rate variation in such scenarios is larger. | To reduce the overhead, the BSs use a DNN to determine the best beams using quasi omni-directional patterns during the online test phase. | [111] |
Out-of-band information | CNN |
● The dataset generated is of academic and industry interest. ● The proposed technique reduced the beam sweeping time by 93% in different scenarios. |
● The proposed method was not compared with other algorithms. ● Only one transmitter and receiver positioning was tested, as well as only one camera location. | The authors create an experimental setup with mmWave hardware, obstacles, and cameras, which originated a dataset of images and beam pairs. Furthermore, the dataset is used for image-based beam prediction. | [82] |
Large-scale MIMO | Q-learning | Outperformed the state-of-the-art in terms of capacity. |
● Assumes Rayleigh fading channel only. ● Performance of the proposed approach in scenarios with varying network conditions or different types of MVNOs is unclear. | Beam scheduling method for enhancing the RF spectrum utilization by subleasing RF slices. | [91] |
Limited feedback | DNN | The method achieves high sum-rates in low SNR regimes and Rician fading. |
● Uses a MISO system only. ● The operating band is not described in the paper. ● Only considers Rician fading, and it is not clear whether the proposed approach can be generalized to other fading models, such as Rayleigh fading and Nakagami-m fading. |
● The beam allocation problem is treated with two different strategies, classification and regression. ● The time prediction of the proposed approach is 6-times shorter than the optimal solution prediction time. | [107] |
Interference rejection | CNN |
● No prior knowledge of the DoA is required. ● The computational complexity is reduced for both space and space–time processing. |
● Needs large training datasets and offline training. ● Lack of implementation/validation, limited comparison with other approaches, and limited NN architecture explanation. | The CNN is employed for space and space–time processing, evaluated in two scenarios with different interference and DoA configurations. | [121] |
Power restrictions | CNN | The intensive computational training phase is performed offline. |
● Considers perfect CSI only. ● It only tackles downlink beamforming in multiuser MISO systems with per-antenna power constraints and ignores other wireless communication challenges such as interference, mobility, and handover. | The goal of this paper is to maximize the downlink SINR based on power restrictions per antenna at the base station and improve the performance complexity trade-off. | [119] |
Cloud-assisted | Conv-LSTM | The proposed solution improves positioning prediction accuracy while reducing storage costs by using the cloud and edge collaboratively. |
● The load added to the backhaul and the cloud service is not taken into account. ● Does not consider the impact of different traffic patterns on the performance of the proposed approach, such as varying user densities, mobility patterns, and traffic types. | This paper proposes a collaborative cloud–edge architecture. The BS uses Conv-LSTM to predict the user distribution and, through this, decide on a better set of beams. | [112] |
Scheduling | RL |
● B-BeamOracle RL agent presents the best performance ● The proposed environment emulates a variety of scenarios. |
● Poor agent modeling. ● B-RL performs close to B-Dummy, which simply uses random actions. ● Lacks exploration of other RL algorithms for scheduling and MIMO beam selection. ● Lacks exploration of alternative reward functions for the RL agent. | It uses the CAVIAR methodology for communication systems combined with the AI models and the virtual world components for terrestrial and aerial beam selection. | [90] |
Dataset generation | GRNN | Provides a baseline solution that predicts future beams based on the sequence of previous ones. |
● The baseline solution does not take the generated images into account. ● Assumes a specific camera and sensor setup, and unclear how it performs with varying configurations such as the resolution, field of view, and sensor noise. | This work uses computer vision with AI algorithms to predict blockage through image-classification-aided beam selection. | [128] |
Beam alignment | KSBL-LTS |
● Beam selection policies are employed using both theoretical and real-world channel models. ● The proposed algorithm obtains a faster learning rate when compared with omnidirectional training with slowly time-varying channel support. |
● Only MISO systems are used. ● The algorithm’s complexity is not evaluated. ● Assumes a specific antenna array configuration, and it is not clear whether the proposed approach can be generalized to different antenna array configurations. | Development of an algorithm for mmWave beam alignment and selection policy to validate which policy results in the most-efficient beamformer: linear Thompsom sampling, omnidirectional, random, or greedy. | [123] |
No reference signal | NN | Does not depend on prior knowledge. |
● The proposed technique only works in LoS conditions. ● The algorithm relies on the availability of a diverse dataset of simulated environments based on the mmWave channel model, which may not be representative or easy to obtain. |
● Authors propose an AMPBML algorithm for beam alignment and beam training reduction. ● Partial beams are used to predict the beam distribution vector. | [105] |
DL | More efficient and accurate than MUSIC, but with comparable performance. |
● Requires parameter tuning. ● As it is based on AoA estimation, it is limited to LoS. ● Relies on the assumption of angle reciprocity, which may not always hold true, especially if the base station and mobile devices have different antenna configurations or positions |
● A two-step NN model is proposed to estimate the uplink signal’s AoA with high accuracy. ● Compared with MUSIC, the results show the same or similar estimation performance in terms of the data rate in moderate to high SNR regimes and outperforms it in low SNR ones. | [106] | |
Dual-connectivity | SVM |
● Low computational complexity. ● Memory-efficient approach. |
● Training time significantly increases with the dataset size. ● The algorithm assumes a homogeneous Poisson point process for base station distribution, limiting applicability to diverse scenarios. |
● An SVM algorithm is used with the Sequential Minimal Optimization (SMO) algorithm in each iteration. ● The proposed method based on channel parameters and transmitted power is compared to the optimal codeword, and the results show a reduction in the beam selection complexity with a high ASR. | [113] |
Non-ideal channel conditions | NN | Reduced overhead compared to the exhaustive search and model-based approaches. |
● Marginal post-alignment beamforming gain loss of 1 dB. ● Neglects NLoS channels. ● Latency, jitter, or throughput improvements are not directly measured, which could mask other relevant limitations. |
● This work proposes a compressive-sensing-based method for reducing the number of channel measurements. ● An NN model addresses the CS dictionary mismatch issue caused by radio hardware impairments. ● The results show a 90.2% reduction in the overhead compared to an exhaustive search approach. | [98] |
Beam tracking and rate adaptation | MAB |
● The proposed online RL method achieves significant throughput gains compared to other methods. ● Uses ACK/NACK messages that are part of the HARQ procedure instead of explicit control messages, thus reducing signaling overhead. ● Both real and simulated indoor and outdoor data are used. ● The method selects both the best beam and Modulation Coding Scheme (MCS) without making assumptions about the channel model or mobility pattern. |
● The proposed RL method performance degrades at high speeds. ● Only a single UE is considered. ● Implementation issues around channel estimation, beam selection parameters, HARQ feedback integration, etc., are not discussed in depth, limiting the assessment of feasibility and scalability. ● The approach assumes ACK-/NACK-based HARQ feedback is always available and sufficient for beam tracking decisions. ● Packet losses or limited feedback scenarios are not addressed, potentially reducing accuracy. |
● Proposal of a novel restless MAB framework for beam tracking for mmWave cellular systems using ACK/NACK messages instead of explicit control signaling. ● The method implements an online RL technique called adaptive Thompson sampling, which selects the best beam and MCS pair. | [108] |
Data augmentation | SMOTE |
● Offers a solution to the lack of datasets containing complete 5G scenarios. ● Evaluates the performance of several classification models, providing insights into which models are best-suited for beam selection. |
● Lack of comparison of the SMOTE-based method with other algorithms found in the literature. ● Generating synthetic data to expand limited datasets may introduce biases or inaccuracies, worsening model performance. ● Generalization ability is unclear due to limited datasets and simulation-based evaluation. Real-world performance, especially for complex 5G scenarios, may differ significantly and pose additional challenges. |
● Proposal of a method to augment datasets with synthetic data. ● Demonstrates that including synthetic data can improve the performance of machine learning algorithms in selecting beamforming in 5G MIMO scenarios. | [114] |
Angle estimation and user selection | DL |
● Reduction of the beam selection overhead and consequent reduction of the computational complexity involved in this task. ● Good performance in terms of achievable sum-rate and multi-user angle estimation with a single camera. |
● The angle estimation accuracy is limited by the single camera’s field of view and resolution and the quality of the image-processing algorithm used for angle estimation. ● No experimental validation is provided. Numerical simulation may not capture all the real-world factors that can impact system performance. | A computer-vision-based method to estimate the beam angle, consequently selecting the beam and user. | [115] |
CV-based UAV localization | CNN |
● Compared to traditional schemes, their proposal significantly saves implementation costs and overhead (e.g., pilot transmission and consequent bandwidth waste). ● The proposed joint optimization scheme can help improve the efficiency of the system. |
● Assumes the UAVs can obtain accurate visual information from the cameras, which may not always be possible in real-world scenarios. ● Requires prior knowledge of the locations of the grounded receivers, which may not always be available or may be subject to errors, especially in dynamic scenarios where the receivers may be moving. ● Simulation results are based on idealized assumptions and may not fully capture the real-world challenges and complexities of mmWave UAV communication systems. | A CV-aided joint optimization scheme of flight trajectory and power allocation for mmWave UAV communication systems. | [116] |
Power control and beam alignment | LSTM |
● Proposes a novel learning framework for beam selection and power control in mmWave massive MIMO communications. ● Addresses the missing data problem and employs LSTM for temporal processing of inputs. ● Designs a learning agent to predict the proper transmit power based on the required transmission rate. |
● The proposed framework is only evaluated through simulations and not tested on real-world data. ● The proposed framework requires accurate ray-tracing channels for training, which may not be easily available. ● The complexity of the proposed framework may be high due to the use of deep learning techniques. ● The proposed approach assumes that the user locations are known, which may not be the case in some scenarios. | Proposal of a DL framework for beam selection and power control in massive MIMO mmWave communications to optimize transmit power and beam selection for users with unknown channel state information. | [117] |
Beam change prediction | LSTM |
● The proposed scheme uses LSTM-enabled models to predict whether a beam change is likely to occur during the next measurement cycle. ● Training and test data are generated using 5G-NR-compatible hardware in an outdoor environment. |
● Only a single outdoor scenario is measured. ● Low mobility, as the measurements are performed during walks. ● Assumes a specific beamforming algorithm, and it is not clear how the proposed approach would perform with different beamforming algorithms, such as time-division multiplexing, frequency-division multiplexing, and hybrid beamforming. | The LSTM-based beam change prediction scheme can achieve over a 58% power reduction regarding beam management compared to deployed commercial schemes. | [109] |
Beam alignment | DNN |
● Reduction of the overhead compared to ES and improves the accuracy compared to hierarchical beam search. ● Uses a uniform planar array on both sides of the link, with the goal of analyzing the effects of rotation in 3D space. |
● The solution presents high computational complexity. ● It relies on precise location and orientation information, which may not always be available or accurate in practice. ● The method could be sensitive to errors or inaccuracies in the contextual information. | This approach proposes using contextual information (position and orientation of user) for the initial beam alignment procedure through deep learning techniques. | [118] |
Challenges | Algorithm | Highlight (Pros) | Limitations (Cons) | Key Contribution | Ref. |
---|---|---|---|---|---|
Beam selection and blockage prediction | Kernel-based KNN | Employs sub-6 GHz CSI to predict vehicle’s positions and, consequently, pre-activate the target BS as a way to speed up handovers preemptively. |
● As it is a lazy learner algorithm, KNN requires the whole dataset to be stored in memory, consuming a large portion of it if the dataset is large. ● The cost of calculating the distance between the new input and each existing example is huge, and sub-optimal solutions may degrade the performance of the algorithm. |
● Uses sub-6 GHz CSI and a kernel-based ML algorithm to predict vehicles’ positions. ● Use historical handover data and the KNN algorithm to accelerate handovers without complicated target selection and beam training processes. | [139] |
Handover success prediction | XGBoost |
● Performs preemptive handover procedures based on estimates of the mmWave channel conditions taken from collocated LTE cell measurements in the sub-6 GHz band. ● The proposed approach has the potential to decrease latency and increase the QoS and QoE. |
● The proposed solution only works if the mmWave and sub-6 GHz cells are collocated. ● The XGBoost algorithm is required to be retrained if any of the cells operate at different frequencies. |
● Introduces the concept of partially blind handovers. ● Employs XGBoost to predict handover success rate from sub-6 GHz to mmWave frequencies. ● Shows that the combination of XGBoost and partially blind handovers improves the handover success rate. | [138] |
Throughput estimation | AROW |
● Employs a time- and memory-efficient online ML algorithm. ● The BS does not need to transmit any control frame, reducing the overhead and increasing throughput. |
● The experiments are carried out with static mmWave BS and devices, which makes the results less useful. ● Online learning algorithms suffer from noisy updates, which might affect the proposed solution’s performance. | Estimates mmWave throughput using depth images and the AROW algorithm. | [133] |
Blockage prediction and preemptive handover | DRL | Uses received power signals and video from depth cameras to train a DRL agent to overcome the computational complexity of learning the optimal handover policy, decreasing handover time. |
● Only two base stations are used, and the experimental setup is rather contrived, which makes the results complicated to be extrapolated to other cases. ● Requires some time for the DRL algorithm to converge as it learns from trial and error attempts. ● The experimental setup is rather contrived. | Shows that blockage prediction is improved by augmenting the input to the DRL agent with video from depth cameras. | [134] |
DRL | Improves blockage prediction and handover reaction time by using depth images from multiple cameras. |
● Blockage caused by pedestrians being out of the camera’s coverage is hard to avoid, requiring a greater number of cameras to be solved. ● Assumes a specific camera and sensor configuration, and unclear how the approach performs with different configurations (e.g., resolution, field of view, noise). | Employs DRL with received signal powers and images from multiple cameras as states to predict blockage and proactively initiate handovers. | [137] | |
GRU |
● Decreases the latency caused by handovers as the current serving BS proactively knows the next serving BS, and then, it can start off the handover procedure before the UE loses the connection with the serving BS. ● Does not require cooperation among multiple BSs, which decreases the overhead associated with coordinated transmissions, consequently reducing power consumption. |
● The proposed model does not account for mobile blockages, only working for stationary blockages. ● Requires a relatively large dataset to achieve reasonable accuracy. ● Assumes specific antenna array configuration, and unclear how the approach performs with different configurations (e.g., uniform linear, circular, non-uniform). | Presents a blockage prediction and proactive handover solution that reduces latency and increases the system reliability in high-mobility applications without requiring high cooperation overhead of coordinated transmission. | [129] | |
Load-balancing handover | DDPG | Maximizes the sum-rate of all UEs moving along different trajectories while minimizing the number of handover and outage events. |
● Does not consider interference from other active UEs, only from other BSs. ● Assumes the UEs’ trajectories are deterministic (perfect mobility prediction). ● The decision process requires estimating the channel capacity and its backup BSs, thus requiring the CSI between the user and multiple BSs. | Maximizes the sum data rate of all users and minimizes the number of handovers and outage events using the DDPG algorithm. | [144] |
Beam gain maximization | CMAB |
● Traditional 3GPP signaling can be used. There is no need for special measurements or new signals. However, information such as the location, speed, and antenna setup can be used for context. ● Link beam performance gain of 0.3 to 0.7 dB compared to the methods in practical propagation environments. |
● A centralized RL agent is required to handle measurement reports from UEs. ● The UE mobility model used for the numerical results is simple (UEs only move on vertical lines). | The handover mobility optimization considers current 5G deployment aspects and uses current 5G signaling. | [145] |
Joint handover and beamforming optimization | Q-learning |
● Channel estimation uses a set of location-based path skeletons, which are defined according to the channel gain, AoA, and AoD. ● Pilot signals are sent only through the path skeleton sets, thus reducing overhead. ● Minimizes the number of handovers by using Q-learning to decide the best backup BS for each UE location and using a link quality threshold to trigger the handover. |
● Assumes perfect trajectory information. ● Requires keeping and updating a path skeleton database, which can be costly for a dense and highly mobile scenario. ● Only a few UE location points are considered for the UE trajectories. | Beamforming can be performed using a low number of pilots due to the use of path skeletons. Handover optimization uses Q-learning to determine the best backup BS for handover based on each UE location and trajectory. | [146] |
Minimization of handovers | MAB |
● Uses received signal strength information collected from the surrounding environment to obtain a coarse UE location estimate to feed the RL algorithm. ● UE location information allows for better trajectory and LoS blockage prediction. ● The proposed method achieves a lower handover number and higher average time of connection in different simulation environments compared to other RL-based handover methods. |
● Low mobility, as UEs are simulated always moving at 1 m/s. ● Does not present data rate results. ● Assumes a specific mobility model, and it is not clear how the proposed approach would perform with different mobility models. ● Assumes a specific type of blockage, and it is not clear how the proposed approach would perform with different types of blockages. |
● Achieves a lower number of handovers than other methods using current 3GGP signaling (i.e., RSS). ● Does not require accurate location and trajectory information. | [131] |
DRL |
● SMART’s computational complexity is much lower than that of the brute-force algorithm to calculate the optimal solution. ● The algorithm can be implemented in a distributed way. |
● The UE may not stay around a specific BS for a sufficient time. Therefore, it cannot have enough historical information to estimate the reward accurately. ● It is not always possible to select the BS with the highest reward. | Reduces the number of handovers and maintains the user’s QoS. | [147] | |
DRL | RHando-F and RHando-S adapt their policies to the channel fading characteristics, providing the robustness of the proposed framework. |
● The method selection is not discussed. ● The two proposed methods perform differently depending on the number of connection requests. | Reduces the number of handovers and increases the average network throughput. | [148] | |
Handover success rate maximization and power allocation | DRL |
● Tackles the joint problem of handover minimization and power allocation. ● The proposed solution addresses instability and vicious competition issues, which are common to decentralized cooperative multi-agent approaches. ● They employ the counterfactual baseline to mitigate the credit assignment problem, achieving better performance. |
● Higher overhead since information such as individual UE states must be sent to the central model. ● Overhead is also increased due to the transmission of policies to the individual UEs. ● Assumes a specific user behavior pattern, and it is not clear how the proposed approach would perform with different user behavior patterns. ● The proposed approach involves a complex DRL algorithm, which may require significant computational resources and may not be practical for implementation in real-world systems. | Employs a fully cooperative multi-agent DRL approach to optimize handover success and power allocation jointly. | [140] |
Maximization of handover success rate and user localization | DL |
● The proposed solution improves the network’s throughput, reduces signaling overhead, and improves the overall network’s energy efficiency. ● Performs better than 3GPP models under the presence of uncertainty. |
● As it employs a DL model, it requires a large training dataset, which requires a more prolonged training phase. ● Does not consider delayed (outdated) channel information. ● As it employs the RSRP as the input to the DL model, it loses the phase information, which might negatively impact the performance of the handover and localization mechanism. | Usage of DL with users’ RSRP signals as the input to implement a handover and localization mechanism. | [142] |
Maximization of handover success rate | XGBoost |
● The proposed solution minimizes the signal overhead and improves the success rate of handovers. ● Reduces energy consumption due to reduction of signaling overhead. |
● XGBoost is very sensitive to outliers since every classifier is forced to fix the errors in the predecessor learners. Therefore, pre-processing is required, which might increase the proposed solution’s computational complexity ● XGBoost is hardly scalable. | Usage of XGBoost and the CSI to implement a handover mechanism. | [143] |
Handover prediction | DRL |
● The multi-agent solution is based on UEs’ trajectories. ● Agents share their policies, speeding up the learning process. ● The agents use image-like states, presenting the location of UEs, BSs, and obstacles at a given time, as the input to the DRL models. ● Differently from other works, they consider the issue that UEs might handover to the same BS, which decreases the system throughput. The proposed multi-agent solution ensures handover minimization and the system’s throughput maximization. |
● It is based on the trajectories of UEs. Therefore, it requires the transmission of such information, which increases the overhead. ● As it is based on the trajectories of users, it cannot be applied to the initial access phase. ● It is an offline learning framework that requires data to be collected before any training is performed. ● As the agents share policies, it might increase the transmission overhead, decreasing the network’s performance. | Multi-agent DRL approach that employs image-like states as the input and takes the maximization of the system’s throughput into consideration as well. | [141] |
Q-learning |
● Improves handover decisions by predicting human blockages based on pedestrians’ locations and velocities. ● Maximizes the throughput of users. |
● If the number of access points increases considerably, Q-learning will not be able to learn an optimal handover policy. ● The states, namely location and velocity, are discretized, which discards part of the information conveyed by them. | Usage of pedestrians’ locations and velocities to maximize their throughput by predicting the necessity of handovers. | [136] | |
Proactive handovers | DRL | Employs DRL to map images into handover decisions, improving the QoS perceived by users since handovers are proactively triggered. |
● Since it is a DRL-based solution (learns by trial and error), it may present a long learning curve until convergence, which might hinder its deployment. ● As it uses images, it requires a relatively large number of images to achieve reasonable performance. ● The delay to obtain an image might impact the performance of the proposed framework. | Usage of camera images to proactively trigger handovers. | [135] |
Paper | ML Technique | Antenna Type | Frequency | Complexity Analysis |
---|---|---|---|---|
[157] | DRL | 4 × 2 MIMO | Not mentioned | ◯ |
[163] | DL | 64 Antennas | Not mentioned | ⚫ |
[168] | DL | 64 antennas | Not mentioned | ◯ |
[169] | DL | 36 BS antennas, 9 UE antennas | mmWave | ⚫ |
[170] | DL | 64 antennas | 28 GHz | ⚫ |
[171] | DL | 64 antennas | Not mentioned | ◯ |
[172] | DL, RL | 32 × 8 UPA | mmWave | ◯ |
[174] | AHC | 8 × 8 UPA transmitter, 2 × 2 UPA receiver | mmWave | ◯ |
[175] | DL | 64 ULA | Not mentioned | ◯ |
[176] | DL | 64 ULA | mmWave | ◯ |
[178] | DL | 4 antennas | Not mentioned | ◯ |
[179] | DL | 64 ULA | 3.5 GHz | ◯ |
[180] | DL | 36 antennas | mmWave | ⚫ |
[181] | DL | 64 Antenna | mmWave | ⚫ |
[182] | DL | 64 ULA | mmWave | ◯ |
Challenges | Algorithm | Highlight (Pros.) | Limitations (Cons.) | Key Contributions | Ref. |
---|---|---|---|---|---|
Beam prediction under adversarial attacks | DL | The proposed counterattack can be used against a variety of different adversarial ML attacks. |
● To be effective, the attacker must have access to the gradient of the loss function for a given input instance, which in turn implies having access to the model’s weights, which is often unfeasible. ● It focuses on the fast gradient sign method as the adversarial attack, but does not consider other stronger attack techniques. ● The evaluated threat model is limited, and the proposed defense may not generalize well to more advanced attacks. ● The proposed adversarial learning approach is quite generic and not tailored to the unique characteristics of 6G mmWave systems. ● Domain-specific insights are lacking, limiting the practical value. | Proposes a mitigation method that uses the gradients of the victim’s model to retrain it with adversarial samples and their respective labels and mitigate adversarial attacks, consequently improving the security. | [188] |
Proposes two methods for counterattacking adversarial attacks: adversarial training and defensive distillation. | [189] | ||||
● Studies how adversarial attacks confuse trained DL models used for mmWave beam prediction. ● Proposes a method to mitigate adversarial attacks using iterative adversarial training. | [190] |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Brilhante, D.d.S.; Manjarres, J.C.; Moreira, R.; de Oliveira Veiga, L.; de Rezende, J.F.; Müller, F.; Klautau, A.; Leonel Mendes, L.; P. de Figueiredo, F.A. A Literature Survey on AI-Aided Beamforming and Beam Management for 5G and 6G Systems. Sensors 2023, 23, 4359. https://doi.org/10.3390/s23094359
Brilhante DdS, Manjarres JC, Moreira R, de Oliveira Veiga L, de Rezende JF, Müller F, Klautau A, Leonel Mendes L, P. de Figueiredo FA. A Literature Survey on AI-Aided Beamforming and Beam Management for 5G and 6G Systems. Sensors. 2023; 23(9):4359. https://doi.org/10.3390/s23094359
Chicago/Turabian StyleBrilhante, Davi da Silva, Joanna Carolina Manjarres, Rodrigo Moreira, Lucas de Oliveira Veiga, José F. de Rezende, Francisco Müller, Aldebaro Klautau, Luciano Leonel Mendes, and Felipe A. P. de Figueiredo. 2023. "A Literature Survey on AI-Aided Beamforming and Beam Management for 5G and 6G Systems" Sensors 23, no. 9: 4359. https://doi.org/10.3390/s23094359
APA StyleBrilhante, D. d. S., Manjarres, J. C., Moreira, R., de Oliveira Veiga, L., de Rezende, J. F., Müller, F., Klautau, A., Leonel Mendes, L., & P. de Figueiredo, F. A. (2023). A Literature Survey on AI-Aided Beamforming and Beam Management for 5G and 6G Systems. Sensors, 23(9), 4359. https://doi.org/10.3390/s23094359