Joint selection of the MCS’s and power allocation coefficients in the two-user downlink PD-NOMA system

. Power Domain Non-Orthogonal Multiple Access (PD-NOMA) is a perspective multiplexing technique for future cellular networks. Nevertheless, it is poorly studied and not applied in the existing systems due to the complexity of PD-NOMA signal processing, resource scheduling, and power allocation. The issue is that a modulation and coding scheme (MCS) selection, including power allocation, is a cooperative procedure considering the channel state information of each multiplexed user. It can be solved by enumerating all possible multiplexing combinations but at the expense of the high computational complexity. In our work, we propose a composed table with the joint MCS’s, which can be selected by the base station (BS) for the user multiplexing in a downlink PD-NOMA system based on their signal-to-noise (SNR) ratios. It allows selecting two MCS’s with two power allocation coefficients for both users and guarantees the 10% block error rate (BLER) performance in the additive white Gaussian noise (AWGN) channel. The joint MCS selection method is based on a max-rate scheduling strategy and provides system capacity maximization ignoring fairness between users. The proposed table is given in the Appendix.


Introduction
PD-NOMA is one of the perspective techniques for user multiplexing in future networks. It has been shown [1][2][3][4][5] that PD-NOMA allows increasing the system capacity in the multiuser communication systems such as cellular (3GPP long term evolution, LTE) and wireless local area networks (IEEE 802.11). The basic PD-NOMA idea is the nonorthogonal multiplexing of several users in a power domain of the same time-frequency resource segment by the superposition of their signals with different power weights. Thus, there is a controlled co-channel interference that is cancelled at the receiver side by the successive interference cancellation (SIC) method [6]. The issue is that user signals are dependent on each other, and it should be considered during resource scheduling, cooperative MCS selection, and multiplexing.
To achieve the maximum system capacity, it is necessary to use the MCS that has the highest spectral efficiency (SE) and guarantees the required error rate performance within the given channel conditions. Therefore, adaptive modulation and coding (AMC) has been proposed [7,8]. It is a simple approach that selects the most efficient MCS according to the propagation channel quality. The AMC is widely used in actual wireless networks with orthogonal multiple access (OMA). Thus, in conventional OMA systems, MCS's for different users are independently selected because they are orthogonal to each other. Unlike OMA, in PD-NOMA systems, the users are non-orthogonal in the power domain, and it is a reason for the joint MCS selection. Besides, the power allocation coefficients should be calculated by taking into account the channel quality of each user. It can be carried out by enumerating all possible multiplexing configurations and selecting the most effective of them (joint MCS's and power allocation coefficients) at the expense of the high computational complexity.
Many researchers have developed power allocation schemes for PD-NOMA within the conditions of the Shannon system model, considering the infinite number of MCS levels. However, during the literature searching, we have found the only paper devoted to the joint MCS selection in PD-NOMA systems using finite number of MCS levels from the practical system and considering the efficiency of used error-correcting methods. In [9], the authors propose their method of a joint power allocation and MCS selection considering max-min user fairness and determining the power allocation coefficient according to the MCS level of the user with a poor channel. It allocates the maximum power shared to the user with better channel conditions while guaranteeing the decoding of the second user. However, their algorithm is iterative, so it has high computational complexity.
In our work, we propose a table that makes it possible to select the pair of MCS levels depending on the user SNR values considering the max C/I scheduling strategy. It maximizes the system throughput instantly by selecting two MCS's that have the highest sum SE. This approach has a very low computing complexity and may be a temporary solution for the AMC in a two-user PD-NOMA system until an optimal approach is found.

Downlink PD-NOMA system
We are using the PD-NOMA system model with multiplexing of two users in the power domain. According to its principles, the group signal is a superposition of two user signals with different power weights. The user with a weak propagation channel gets the largest power share compared to the user with a strong propagation channel. We sorted users in the ascending order of their channel quality and identified the weak user (UE1) with power allocation coefficient 1 p and the strong user (UE2) with

Proposed joint MCS Table
In this section, we have described the joint MCS table design process. Our goal is to search for the most efficient joint MCS combination, which guarantees 10% BLER performance and provides the maximization of the system capacity, considering the user's SNR values. Our approach to the search includes the enumeration of all possible combinations for each joint user's SNR values and their verification with the help of a simulation. The combinations that do not guarantee 10% BLER are eliminated from a valid range of searches. Then, one with the most SE is selected from the remaining combinations.

Used MCS levels
A downlink system capacity significantly depends on a user non-orthogonal multiplexing technique at the BS side, which includes the joint MCS selection and power allocation between users. Commonly, all possible MCS's are given in a special table in a technical specification. In our system model, we use a bank of the finite number of MCS levels given by [10]. Their main characteristics are presented in Table 1.

MCS Table in LTE Downlink..
Each MCS level is indexed by m and based on a square Q-QAM and Turbo encoding with the code rate R. The SNR threshold with 10% BLER is used for switching m together with changing the channel state. It allows increasing the SE of transmission due to improving the channel quality.

Simulation model
The structure of the simulation model is shown in Fig. 1 and all possible power allocation coefficients controlled by 1 p in the range from 0.5 to 1 with the 0.05 step. Next, we initiate the simulation of a data transmission process and estimate a BLER performance. Further, the 1 m , 2 m , 1 p combination that guarantees 10% BLER performance and maximizes the system capacity is selected and added in the table for the SNR thresholds 1 W , 2 W . In case the identical sum SE is given by several combinations, the priority is given to the weak user, and the combination that simultaneously maximizes throughput and the SE of UE1 is added in the table. Fig. 1. The structure of the simulation model. p , so only five additional bits are required to transfer the control information about the power shared between the users.

Joint MCS's selection
Exactly as for the AMC in the OMA, in the PD-NOMA, the BS should select the MCS combination with the highest sum SE for the given users' SNR values, considering the thresholds 1 W , 2 W . These methods are well described in [10]. For example, if UE1 and UE2 possess 1