Budgeted Thompson Sampling for IRS Enabled WiGig Relaying
Round 1
Reviewer 1 Report
The intelligent reconfigurable surface (IRS) is a promising technology to extend future B5G and 6G wireless communication coverage ranges. This paper proposes a self-learning-based budgeted Thomson sampling approach to reduce the beamforming training time. To have efficient beamforming, it is desirable to use a cost-effective algorithm to determine phase values for IRS elements. However, the author seems to misunderstand the operating principle between IRS and antenna array for beamforming. Antenna arrays use phase shifters to control the phase change for each antenna element to have beamforming. While, IRS controls the phase of each unit cell by the reconfigurable meta-atom design, which consumes less energy.
Author Response
Thank you for your feedback and for bringing up the issue of the operating principle between intelligent reconfigurable surfaces (IRSs) and antenna arrays for beamforming. We apologize for any misunderstandings in our previous explanation and appreciate the opportunity to provide a clearer explanation. You are correct that IRSs control the phase of each unit cell by the reconfigurable meta-atom design, while antenna arrays use phase shifters to control the phase change for each antenna element. This difference is a significant advantage of IRSs over antenna arrays, as the reconfigurable meta-atom design of IRSs consumes less energy compared to the use of phase shifters in antenna arrays.
Our paper proposes a self-learning-based budgeted Thomson sampling approach to select the supreme IRS board while considering the consumed beamforming training time. Our goal is to find a cost-effective algorithm that can choose the best surrounded IRS relay with estimating the phase values for the IRS elements efficiently, which is critical for improving the performance of IRSs in wireless communication systems. We try to make the WiGig BS choose the best surrounded IRS board using multiarmed techniques to adapt dynamic communication environment without any prior channel state information. Previous work confirms our theoretical findings such as
- M. Mohamed, S. Hashima, K. Hatano and M. M. Fouda, "Cost-Effective MAB Approaches for Reconfigurable Intelligent Surface Aided Millimeter Wave Relaying," in IEEE Access, vol. 10, pp. 81642-81653, 2022, doi: 10.1109/ACCESS.2022.3195303.
- M. Mohamed, S. Hashima and K. Hatano, "Energy Aware Multiarmed Bandit for Millimeter Wave-Based UAV Mounted RIS Networks," in IEEE Wireless Communications Letters, vol. 11, no. 6, pp. 1293-1297, June 2022, doi: 10.1109/LWC.2022.3164939.
- Du, J. Zhang, J. Cheng and B. Ai, "Millimeter Wave Communications With Reconfigurable Intelligent Surfaces: Performance Analysis and Optimization," in IEEE Transactions on Communications, vol. 69, no. 4, pp. 2752-2768, April 2021, doi: 10.1109/TCOMM.2021.3051682.
We appreciate your feedback and hope that our revised explanation accurately reflects the operating principle between IRSs and antenna arrays for beamforming. Thank you for the opportunity to provide more information about our study.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Please the attached comments.
Comments for author File: Comments.pdf
Author Response
Reviewer#2: This paper proposed a self-learning-based budgeted Thomson sampling approach for IRS relay probing (BTS-IRS) to handle the supreme IRS selection with the lowest BT time cost. Generally, the authors have done a solid work, however, some parts are not well written. Both the motivations and contributions should be further clarified.
Author response: We would like to deeply thank the respected reviewer for the time and efforts she/he spent in reviewing our humble work especially this reviewer gives us wonderful deep insight comments/concerns.
Reviewer#2, Concern # 1: The first critical issue is the different concepts between IRS and relay. There have a lot of researches to distinguish the IRS and relay techniques, where the former only reflect the incident signal by adjusting its phase shift and possible signal amplification, while the latter one would regenerate the signal with the hardware of RF chains. Thus, it is not suggested to use “IRS relay”.
Author response: We would like to deeply thank the respected reviewer for this valuable and constructive comment. Thank you for pointing out the important distinction between IRS and relay techniques. You are correct that IRS only reflects the incident signal by adjusting its phase shift and possible signal amplification, while a relay regenerates the signal using RF chains such as amplify and forward and decode and forward types as mentioned in the manuscript. However, our intention behind IRS relay term that as mentioned in the system model section and from fig.1 we have IRS panels that relays the data from the WGBS to the mobile users via non line of sight (NLOS) as direct link is blocked. Hence, we choose the best probed one to relay information to the users. This means that, the IRS panels are the relays in our scenario and we are trying to find the best one. Furthermore, some previous work has the same term on their titles as follow:
- V. Nguyen, T. P. Truong, T. M. T. Nguyen, W. Noh and S. Cho, "Achievable Rate Analysis of Two-Hop Interference Channel With Coordinated IRS Relay," in IEEE Transactions on Wireless Communications, vol. 21, no. 9, pp. 7055-7071, Sept. 2022, doi: 10.1109/TWC.2022.3154372.
- Yuan, D. Wu, Y. Huang and C. -L. I, "Reconfigurable Intelligent Surface Relay: Lessons of the Past and Strategies for Its Success," in IEEE Communications Magazine, vol. 60, no. 12, pp. 117-123, December 2022, doi: 10.1109/MCOM.003.2200193.
- Su, X. Pang, S. Chen, X. Jiang, N. Zhao and F. R. Yu, "IRS-UAV Relaying Networks for Spectrum and Energy Efficiency Maximization," ICC 2022 - IEEE International Conference on Communications, Seoul, Korea, Republic of, 2022, pp. 2834-2839, doi: 10.1109/ICC45855.2022.9839112.
Thus, the name is a mater of terminology rather than a scientific concept.
Author action: To cope with this valuable comment, we updated the Introduction part of the revised manuscript as follows:
Revised Manuscript
Introduction
Reviewer#2, Concern #2: The full name of six-generation (6G) should move to its first appearance in the
first paragraph in Introduction.
Author response: We would like to deeply thank the respected reviewer for this valuable and constructive comment. In the revised version we corrected this issue.
Author action: To cope with this valuable comment, we updated the conclusion of the revised manuscript as follows:
Revised Manuscript
Reviewer#2, Concern #3: It is suggested to introduce the following recent works in IRS [R1]-[R2] and
beamforming [R3]-[R4] fields to highlight the state-of-the-art of this paper.
Author response: We would like to deeply thank the respected reviewer for this valuable and constructive comment. In the revised version we added the citations in related work part.
Author action: To cope with this valuable comment, we updated the conclusion of the revised manuscript as follows:
Revised Manuscript
References
Reviewer#2, Concern #4: The motivation is not well clarified, which is important to demonstrate why the authors investigated this work. The contributions should be also improved to exploit the possible applications to real world.
Author response: Thank you for your feedback and for pointing out the importance of clarifying the motivation behind our work. We appreciate the opportunity to better explain our motivation and contributions.
The motivation behind our work is to investigate the problem of WiGig aided IRS relaying. In recent years, both WiGig and IRSs have received significant attention as promising technologies for improving the performance of wireless communication systems, particularly in terms of increasing spectral efficiency and reducing energy consumption. However, the performance of IRSs is highly dependent on the choice of precoding techniques and the training time required to determine the optimal precoder weights. Hence our goal is to find the best surrounding IRS panel to rely information from WGBS to mobile user using it.
Our contributions are summarized as follows:
- The WiGig IRS relay probing problem is mathematically formulated as an optimization problem in order to choose the supreme IRS relay that maximizes the attainable data rates of WiGig communication linkages. Still, the selection process is constrained by the time consumption of the BT procedure of the chosen IRS panel.
- WiGig-IRS relay probing issue is reformulated as online budgeted MAB, where budgeted Thompson sampling (BTS-IRS) with guaranteed time-cost performance is envisioned to handle it.
- The proposed BTS-IRS algorithm demonstrates superior performance compared to other benchmark schemes through numerous numerical simulations in various scenarios.
- Numerous numerical evaluations were carried out to illustrate the superiority of the envisioned BTS scheme compared to other benchmarks such as the brute force solution, the classical TS algorithm and random selection of IRS relays. Moreover, BTS-IRS is compared to previous cost-effective UCB-based MAB schemes. The BTS-IRS scheme exhibited an average spectral efficiency that was superior to other benchmark schemes, with comparable BT time computation.
We hope that our explanation clarifies the motivations and contributions of our work.
Author action: To cope with this valuable comment, we updated the Introduction of the revised manuscript as follows:
Revised Manuscript
Reviewer#2, Concern #5: Why the authors considered the beamforming training time? Typically, for selecting the best IRS or relays, the existing works usually focused on the channel estimation and the channel quality and most of them have not take the BT into account. The authors should further explain the reasons.
Author response: We would like to deeply thank the respected reviewer for this valuable and constructive comment. The beamforming is an important issue in mmWave communications and is a realistic cost in our setup as reducing the BT time will reduce the latency hence improve the channel quality in advance. Also, usually, BT time includes channel estimation time for adjusting the best TX/RX beams. Thus, BT time is more generalized metric to complexity of RIS relaying especially when using mmWave. Multiarmed bandits can handle this issue (choose the best nearby IRS panel that maximizes the data rate and minimize the BT time) without any prior information about the channel state information, which highly relaxes channel estimation complexity issues. Moreover, the channel quality is measured in our paper using spectral efficiency metric which we try to maximize while minimizing the BT time.
In our paper, we considered the beamforming training time because it is a critical factor that can significantly impact the performance of intelligent reconfigurable surfaces (IRSs) systems. The beamforming training time represents the amount of time required to train the system and determine the optimal precoder weights for a given channel state. In real-world wireless communication systems, the channel can change quickly, and the beamforming weights must be updated frequently to ensure optimal performance. The beamforming training time can have a significant impact on the overall system performance, as longer training times can result in reduced spectral efficiency and reduced system throughput. Therefore, the beamforming training time is a crucial factor that must be considered when designing and evaluating RIS systems. By taking the beamforming training time into account, we can better understand the practical limitations and challenges of implementing IRS systems in real-world scenarios.
Reviewer#2, Concern #6: Why the authors considered analog precoder vector instead of digital or hybrid
one?
Author response: We would like to deeply thank the respected reviewer for this valuable and constructive comment. In our study, we considered analog precoding as it provides several benefits over digital and hybrid precoders in the context of intelligent reconfigurable surfaces (RISs). One of the main reasons for considering analog precoding is the reduced complexity compared to digital or hybrid precoders. Analog precoders can be implemented with low-complexity circuits, which makes them suitable for low-cost and low-power RIS systems. Another advantage of analog precoding is the ability to implement continuous phase shifts, which can improve the performance of the RIS system. In contrast, digital precoders are limited to discrete phase shifts, which can result in reduced performance. Additionally, analog precoding can be easily adapted to the environment in real-time, which is important in dynamic wireless environments. In conclusion, the choice between analog, digital, and hybrid precoders in the context of IRSs depends on the specific requirements of the system and the trade-off between performance, complexity, and cost. In our study, we considered analog precoding as it provides a good balance between performance and complexity for the specific scenario we analyzed. We hope that our explanations have clarified the motivations behind our choice of analog precoding.
Author action: To cope with this valuable comment, we updated the conclusion of the revised manuscript as follows:
Revised Manuscript
Reviewer#2, Concern #7 The reviewer is not familiar with the BMAB method, which looks like it is suitable for any selection problem. Besides, mathematical processing with IRS is not complex. Thus, the authors should highlight the novelty except for the existing BMAB method.
Author response: Thank you for your feedback and for bringing up the issue of the novelty of the budgeted multi-armed bandits (BMAB) method in the context of WiGig aided IRS relay probing. We understand the importance of highlighting the novelty of our work and appreciate the opportunity to provide more information. While the BMAB method is a well-known approach for solving selection problems, our work represents a novel application of the BMAB method to the problem of beamforming training time optimization in WiGig-IRS systems. To the best of our knowledge, this is the first time the BTS-IRS method has been applied to WiGig IRS relay probing with anticipating the beamforming training time as time costIn our work, we use the BTS-IRS method to efficiently determine the optimal IRS relay with rational beamform training time. This is particularly important in wireless communication systems, where the channel can change quickly, and the beamforming weights must be updated frequently to ensure optimal performance. Our work represents a significant step forward in terms of optimizing the beamforming training time. By using the BMAB method, we can achieve a balance between performance and cost (beamform training time), which is critical in real-world wireless communication systems.
We hope that our explanation highlights the novelty of our work and the significance of our contribution. Thank you for the opportunity to provide more information about our study.
Reviewer#2, Concern #8 The reviewer has no idea why the Mean BT time cost would change with the transmitter power, which need further explanation.
Author response: We would like to deeply thank the respected reviewer for this valuable and constructive comment. In the revised version we corrected this error coming from low number of iterations.As the respected reviewer mentioned in his comment, BT time cost should be constant against transmitted power.
Author action: To cope with this valuable comment, we updated this figure in the revised manuscript as follows:
Revised Manuscript
Finally, we hope if we could clearly answer all inquiries and questions given by the respected reviewers, and we would like to deeply thank them for the time and efforts they spent in reviewing our paper to enhance its quality and its technical content. Also, we hope if we could convince the reviewers and the associate Editor to change their decision and accept our revised paper to be published in the prestigious highly impacted Journal of electronics.
Thanks, with our best regards.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
As for choosing the best surrounded IRS board for wireless communication, I expect more explanation on the working principle of the algorithm:
1. IRSs are actively tunned to adapt the communication channel, so did the author consider this variate？
2. What is the background principle of the experiment result? Could the author explain it in more detail?
Author Response
Reviewer#1: Concern #1 IRSs are actively tunned to adapt the communication channel, so did the author consider this variate？
Author response:
We would like to deeply thank the respected reviewer for the time and efforts she/he spent in reviewing our humble work especially since this reviewer gives us wonderful deep insight comments/concerns.
One of the key advantages of using an IRS is its ability to actively manipulate the signal, and this can include tuning the reflectivity coefficients of the IRS elements to adapt to the changes in the channel conditions. Yes, we considered the adaptivity of IRSs in their work. In the paper, we highlight that the performance of WGBS is heavily contingent upon the selection of a suitable IRS relay considering the training time required to determine the optimal precoder weights. Therefore, we proposed a self-learning-based budgeted Thompson sampling approach for IRS relay probing (BTS-IRS) to address the time-consuming nature of beamforming training (BT) for IRSs. The proposed BTS-IRS algorithm selects the optimal IRS relay based on the payoff and costs posterior distributions, considering the BT time cost of probing the IRS relay. The authors have also evaluated the performance of BTS-IRS in various scenarios through extensive numerical simulations. Therefore, the authors have considered the adaptivity of IRSs in their work.
In our paper, we focus on the selection of an optimal IRS relay for the maximum attainable data rate as the direct link between WGBS and UE is blocked. This is a complex problem, as it requires the tuning of the phase shifts for both the WiGig base station and the IRS relay. IRSs are indeed actively tuned to adapt the communication channel, and we take this into account by incorporating the beamforming training (BT) time cost of probing the IRS relay into our budgeted Thompson sampling approach (BTS-IRS). BTS-IRS is a self-learning-based method that addresses the challenge of selecting the optimal IRS relay while minimizing the BT time cost. In this approach, we model the WiGig-IRS relay probing problem as a budgeted multi-armed bandit (MAB) methodology, where the WGBS, IRS boards/relays, throughput, and BT are the bandit player, arms, reward, and cost, respectively. We incorporate the BT time cost of probing the IRS relay into the main BTS formula, where both payoff and cost posterior distributions are sampled separately, their ratio is estimated, and the arm/IRS relay with the highest ratio is chosen. Through extensive numerical simulations in various scenarios, we show that our proposed BTS-IRS algorithm outperforms other benchmark schemes, such as the brute force solution, the classical Thompson sampling algorithm, and random IRS relay selection. We also compare our approach to previous cost-effective UCB-based MAB schemes. The BTS-IRS scheme exhibits superior average throughput performance over other compared benchmark schemes, with comparable BT time computation.
Author action: To cope with this valuable comment, we updated the conclusion of the revised manuscript as follows:
Revised Manuscript
- Introduction
Page 2: footnote
Reviewer#1: Concern #2 What is the background principle of the experiment result? Could the author explain it in more detail?
Author response: We would like to deeply thank the respected reviewer for the time and efforts she/he spent in reviewing our humble work especially this reviewer gives us wonderful deep insight comments/concerns. In terms of the background principle of the experiment result, Figure 1 explains the setup scenario. The proposed BTS-IRS approach utilizes a budgeted Thompson sampling algorithm that incorporates the BT time cost of probing the IRS relay into the main BTS formula. This enables the selection of the IRS relay with the highest payoff/cost ratio, and hence, reduces the overall time required to select the optimal IRS relay. The simulation results show that the BTS-IRS approach achieves improved performance in terms of BT time consumption/cost, spectral efficiency, and attainable data rate compared to other benchmark schemes.
Author action: To cope with this valuable comment, we updated the Introduction part of the revised manuscript as follows:
Revised Manuscript
- Numerical Analysis
Reviewer 2 Report
The authors have well addressed my concerns, no further comments.
Author Response
We would like to deeply thank the respected reviewer for the time and efforts she/he spent in reviewing our humble work especially since this reviewer gives us wonderful deep insight comments/concerns.
Round 3
Reviewer 1 Report
I appreciate the authors' explanation, now it makes much more sense.
