Questions & Answers
If you have any questions on the topic, please feel free to send your question to y.mao16@imperial.ac.uk. We will collect all questions and post the answers here.
What are the major benefits of RSMA?
The most appealing capability of RSMA is that it is powerful to manage multi-user interference. Therefore, it is suited to any environment with multi-user interference. Based on the existing literature of RSMA, we here summarize the major advantages of RSMA:
Universal: RSMA is a more general multiple access framework that outperforms and unifies orthogonal multiple access (OMA), space division multiple access (SDMA) based on multi-user linear precoding (MU-LP) and multi-antenna power-domain non-orthogonal multiple access (NOMA) as sub-schemes.
Flexible: RSMA is suited to all user deployments (with a diversity of channel directions, channel strengths) and network loads (underloaded and overloaded regimes). It implies that RSMA is capable of managing all different kinds of interference flexibly. RSMA automatically reduces to other multiple access techniques according to the channel conditions, i.e., it reduces to SDMA when user channels are orthogonal in the underloaded multiple-input single-output broadcast channel (MISO BC) with perfect channel state information at the transmitter (CSIT). When the channels are aligned with certain channel strength disparities, it automatically boils down to power-domain NOMA. For other channel conditions, RSMA takes advance to the common streams and achieves a better interference management by partially decoding the interference and partially treating the remaining interference as noise.
Robust: RSMA is robust to CSIT inaccuracy. The impairment of CSIT may due to i.e., quantized feedback, pilot contamination, channel estimation error, mobility, latency, subband feedback. As RSMA is primarily motivated by multi-antenna deployments with multi-user interference coming from imperfect CSIT, it compensates the degree-of-freedom (DoF) loss of other multiple access techniques in imperfect CSIT and is therefore, less sensitive to CSIT inaccuracy.
Spectrally efficient: The spectral efficiency of RSMA is always larger than or equal to that of existing multiple access techniques. Considering a MISO BC without quality of service (QoS) constraints, the rate region of RSMA comes much closer to the optimal dirty paper coding (DPC) region than SDMA and NOMA when CSIT is perfect. When CSIT becomes imperfect CSIT, linearly precoded RSMA is able to achieve a larger rate region than complex DPC in multi-antenna BC. As RSMA achieves the optimal DoF in both perfect and imperfect CSIT, it optimally exploits the spatial dimensions and the availability of CSIT. This contrasts with SDMA and NOMA that are suboptimal.
Energy efficient: As RSMA is more general than SDMA and NOMA, its energy efficiency is also larger than or equal to that of existing multiple access techniques in a wide range of user deployments.
Enhancing QoS and fairness: RSMA exhibits a more explicit performance gain over other multiple access techniques when there is a QoS rate constraint for each user or when a higher weight is allocated to the user with a weaker channel condition. Therefore, the ability of a wireless network architecture to partially decode interference and partially treat interference as noise leads to enhanced QoS and user fairness.
Reducing complexity: The performance gain of RSMA can come with a lower transmitter and receiver complexity than multi-antenna NOMA. In contrast to multi-antenna NOMA that requires user grouping, ordering, switching (between NOMA and SDMA) at the transmit scheduler and multiple layers of SIC at the receivers, 1-layer RS without any user ordering, grouping or dynamic switching at the transmit scheduler and with only one layer of SIC at each receiver is capable of achieving significant performance gain over. In contrast to SDMA that requires user pairing to pair users with semi-orthogonal channels, RSMA is suited to all channel conditions and it does not require complex user scheduling and pairing. Moreover, RSMA is capable of further reducing CSI feedback overhead in the presence of quantized feedback.
What are the major research challenges of RSMA?
The study of RSMA is still in its infancy. There are still many challenges and open issues remain to be addressed. RSMA is a goldmine of research problems for academia and standard specification issues for industry. The multifarious attractive and potential research directions of RSMA are summarized in the figure below:
The message of each user is split into one or multiple common parts and a private part. How to design the message splitter and message combiner at the transmitter?
This question is related to the implementation of RSMA, which will be investigated towards the standardization of RSMA. We here illustrate an example of how to split user messages and combine the common messages in a two-user rate-splitting framework below:
How does the optimized rate allocation for the common stream guide the practical message split at the transmitter?
As is illustrated in the two-user RSMA physical layer design figure below [Ref 1]:
The optimized rate of the common stream (which is the sum of the optimized rate allocation \bar{C}_1 and \bar{C}_2) will determine the modulation scheme and the block length based on the adaptive modulation coding algorithm. Therefore, the total number of information bits that can be transmitted by the common stream for that block length is determined. These information bits are then allocated proportionally among users based on the optimized \bar{C}_1 and \bar{C}_2. For example, if there are in total 100 information bits in the common stream, and \bar{C}_1=0.6 bit/s/Hz, \bar{C}_2=0.4 bit/s/Hz, then 60 bits are allocated to user-1 while other 40 bits are allocated to user-2.
[Ref 1] O. Dizdar, Y. Mao, W. Han and B. Clerckx, "Rate-splitting multiple access for downlink multi-antenna communications: Physical layer design and link-level simulations," 31st Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), London, United Kingdom, 2020.
What is Rate-Splitting Multiple Access (RSMA)? What is the basic principle of RSMA?
The definition of RSMA is to serve multiple users based on the concept of rate-splitting (RS). The concept of RS is to split user messages at the transmitter and rely on superposition coding for the transmission of the split messages. Note that RSMA is not limited to successive interference cancellation (SIC) receivers. It is suited for other types of decoder (i.e., joint decoder, turbo decoder).
The basic principle of RSMA is to softly bridge the two extremes of fully decoding the interference and fully treating the interference as noise by partially decoding the interference and partially treating the interference as noise.
What is the status of standardization of RSMA?
The standardization of RSMA has not been taken by 3GPP yet. The machinery required for RSMA is already being studied, discussed and developed. Some current work items in 3GPP, i.e., multi-user multiple input multiple output (MU-MIMO)/coordinated multi-point (CoMP), multiuser superposition transmission (MUST), network-assisted interference cancellation and suppression (NAICS), multicast functionality can be leveraged for RSMA as well.
As the common stream needs to be decoded by multiple users, will it cause any privacy/security issues, i.e., some users may not want to expose their information publicly?
Note that decoding the common stream at the physical layer does not imply the sharing of data as encryption could be implemented at higher layers and decrypting can be performed using user-specific codes.
In 1-layer RS, why do we combine the common parts of user messages into a single common message?
By combining the common parts of multiple users into a single common message, the transmitter and receiver design is much simplified. For example, in the K user case, if the message of each user is split into two parts without message combining, the transmitter has to encode 2K messages and design 2K precoders. At each user, more layers of SIC are needed and the decoding order has to be optimized at the transmitter. In contrast, the transmitter of 1-layer RS only encodes and precodes K+1 streams and only a single layer of SIC is needed at each user without any issue of decoding order design. Therefore, the complexity of 1-layer RS is much reduced.
Is there any situation where the common message is decoded successfully but the private message is failed to be decoded at a certain user?
Most of literature works make use of Shannon theoretic concept. Under Gaussian signalling and infinite block length, the ergodic rates are achievable, and no SIC error occurs. If finite block length is considered, error would occur and SIC would be imperfect. Under imperfect SIC, it is possible that the common message is decoded successfully but the private message is failed to be decoded at a certain user. Readers are referred to the recent paper [Ref 1] where finite block length and SIC error are incorporated.
[Ref 1] O. Dizdar, Y. Mao, W. Han and B. Clerckx, "Rate-splitting multiple access for downlink multi-antenna communications: Physical layer design and link-level simulations," 31st Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), London, United Kingdom, 2020.
The generalized RS strategy seems to be much more complex than NOMA and SDMA specially when the number of users is large. Why is it still important?
We agree that the generalized RS strategy specified in [Ref 1] can be complicated to implement especially when the number of users is large. However, as a generalized framework of RS, it embraces SDMA, power-domain NOMA, multicasting, OMA as special cases. It brings a novel method of softly bridging existing multiple access techniques without using naive hard switching.
Moreover, the generalized RS is applicable to the scenarios with relatively small K and it achieves non-negligible performance gain over existing multiple access techniques. Hence, the transmitter could schedule a small number of users in each resource block.
Another use of the generalized RS is to show the performance of other low-complexity RS strategies such as 1-layer RS and 2-layer hierarchical RS (HRS). Low-complexity RS strategies can achieve closer performance to the generalized RS while their complexities are much lower than the generalized RS and power-domain NOMA strategies. This implies that by departing from the extremes of treating interference as noise and fully decoding interference, it is possible to find alternative strategies that perform well and have a relatively low complexity. It helps us draw the conclusion that 1-layer RS is a good alternative to the generalized RS.
Therefore, the generalized RS is a significant strategy in the framework of RSMA.
[Ref 1] Y. Mao, B. Clerckx, and V.O.K. Li, "Rate-splitting multiple access for downlink communication systems: bridging, generalizing, and outperforming SDMA and NOMA," EURASIP Journal on Wireless Communications and Networking, 133 (2018).
What are major differences between doing rate-splitting for only one user and doing rate-splitting for multiple users?
If the objective is to maximize the sum rate or energy efficiency (defined by sum rate dividing the sum transmit power) subject to transmit power constraint only, the selection of doing rate-splitting for one user (as in [Ref 1]) or more users (as in [Ref 2]) at the transmitter has no impact on the performance. The major question towards this problem is whether rate-splitting is required or not? If the answer is in the affirmative, the question would become: “how much of the total (sum) information should be carried by the common message regardless of how the common message is split?”
However, if user fairness is considered, i.e., when the objective is to maximize the weighted sum rate or/and subject to the QoS rate constraints for each user, different selection schemes of users doing rate splitting at the transmitter will influence the final performance. A better method is to do rate-splitting for all users so as to provide rooms for allocating the rate of the common stream among users.
[Ref 1] H. Joudeh and B. Clerckx, "Sum-rate maximization for linearly precoded downlink multiuser MISO systems with partial CSIT: A rate-splitting approach," in IEEE Transactions on Communications, vol. 64, no. 11, pp. 4847-4861, Nov. 2016.
[Ref 2] Y. Mao, B. Clerckx, and V.O.K. Li, "Rate-splitting multiple access for downlink communication systems: bridging, generalizing, and outperforming SDMA and NOMA," EURASIP Journal on Wireless Communications and Networking, 133 (2018).
At each user, how to recover the original message from the decoded common and private messages?
When the message of a single user is split at the transmitter, the common message only contains the common part of one user. Additional control signaling is only needed between the transmitter and that user to indicate how to recover the original message.
When the messages of multiple users are split and the common parts of these users are merged into one common message at the transmitter, each user first decodes the common message and then retrieves the part intended for itself from the decoded common message. Additional control signaling is required between the transmitter and all users to indicate how to the split and recover the original messages at each user.
The performance improvement of RSMA at low SNR is very small. Does it mean that RSMA is not suitable for microcell or femtocell which has relatively low transmit power?
The performance gain of RSMA is indeed prominent at higher SNRs. Please note that such performance improvement already comes when SNR is larger than 15 dB. SNRs ranging from 15 dB to 30 dB are in fact finitely high or moderate.
Please note that cooperative RS (CRS) with user relaying has been shown to achieve performance gain over other multiple access techniques at lower SNR, i.e., SNR=5 dB or 10 dB [Ref 1, Ref 2]. Therefore, for microcell or femtocell, CRS scheme could be more beneficial.
[Ref 1] J. Zhang, B. Clerckx, J. Ge and Y. Mao, "Cooperative rate splitting for MISO broadcast channel with user relaying, and performance benefits over cooperative NOMA," in IEEE Signal Processing Letters, vol. 26, no. 11, pp. 1678-1682.
[Ref 2] Y. Mao, B. Clerckx, J. Zhang, V. O. K. Li and M. A. Arafah, "Max-min fairness of K-user cooperative rate-splitting in MISO broadcast channel with user relaying," in IEEE Transactions on Wireless Communications, vol. 19, no. 10, pp. 6362-6376, Oct. 2020.
It seems that RSMA can only be employed when there are common messages to be transmitted. What if there are only private messages?
The principle of RSMA is to split user messages into common and private parts. The common parts are combined into a common message to be decoded by all users while the private parts are independent private messages to be decoded by the corresponding user only. The amount of the message of each user to be split into the common message or the private message will be adaptively designed according to the user channel conditions. When the user channels prefer private message-only transmission, i.e., when the user channels are orthogonal with perfect CSIT, there is no message split for each user.
The common message is not a message that is originally intended for all users. It is required to be decoded by all users but is not necessarily intended for all users. It is created from the unicast messages