With the development of 5G applications, the demand for high-quality services has increased rapidly. The mean opinion score (MOS) and user fairness have become important performance indicators for measuring the quality of 5G applications. Orthogonal multiple access-based visible-light communication (VLC-NOMA) can provide high capacity and speed to satisfy the requirements of 5G applications. However, according to Shannon’s formula, whether the user's achievable rate can satisfy its minimum and recommendation rate requirements is an important factor affecting the MOS value in VLC-NOMA. Therefore, developing an effective resource allocation method is an effective way to improve the MOS value and user fairness of indoor VLC-NOMA networks.

A joint optimization method for adaptive user grouping and power allocation (JOAUGPA) is proposed to improve the MOS and user fairness of large-scale equipment access in VLC-NOMA. In JOAUGPA, adaptive user grouping based on hierarchical agglomerative clustering (HAC) is proposed to group users into NOMA groups, which combines NOMAs that satisfy the criteria for hierarchical clustering and NOMA grouping to increase the MOS value of the system. In addition, a power-adaptive adjustment (PAA) method is designed to improve the system MOS value and user fairness while satisfying the minimum rate and MOS requirements of users. The proposed PAA can adaptively allocate the remaining power of the NOMA group to different users based on changes in the MOS value of the NOMA group. Its objective is to maximize the MOS value of the NOMA group.

By simulating a room with 5 m×5 m×3 m with one light emitting diode (LED) on the ceiling as a VLC AP and randomly distributing multiple users at a height of 0.85 m from the ground, the Monte Carlo method is used to verify the performance of proposed sub algorithms HAC and PAA in JOAUGPA algorithm in a given simulation parameter scenario (Table 1 and Table 2). The comparison methods for HAC include adaptive user grouping (AUG), random user grouping (RUG), traditional user grouping (CUG), and user grouping based on linear search (UGLS) algorithms. The PAA comparison methods include fixed power allocation (FPA) and gain ratio power allocation (GRPA) algorithms. The simulation results show that compared with other NOMA grouping methods, the proposed JOAUGPA method can obtain a higher total system MOS value. When M=12, the proposed JOAUGPA achieves a network total MOS increase of 3.25% compared with the UGLS+PAA algorithm, 4.69% compared with the AUG+PAA algorithm, 5.86% compared with CUG+PAA, and 6.88% compared with RUG+PAA (Fig. 1). When the number of users is 10, as the minimum user equipment (UE) demand rate and recommendation rate increase, the proposed JOAUGPA network has the least decrease in total MOS value (Fig. 2). The proposed PAA method obtains larger MOS values and minimizes the MOS value reduction compared with the GRPA and FPA power allocation methods when they adopt the same NOMA grouping (Fig. 3). Compared with other NOMA grouping methods and power allocation methods, the proposed JOAUGPA algorithm has the least decrease in Jain’s fairness index (JFI) as the number of users increases (Figs. 5 and 6).

To improve the MOS value and user fairness of indoor VLC-NOMA downlink transmission systems, a joint optimization method for adaptive user grouping and power allocation is proposed in the paper. Using HAC to form the NOMA group and the PAA method, the MOS value of VLC-NOMA can be maximized, and user fairness can be improved while ensuring the user's MOS threshold and successive interference cancellation (SIC) decoding constraints. With the development of indoor Internet applications, the proposed JOAUGPA can improve the high-speed data transmission of indoor users, increase their quality of experience (QoE), and promote the development of mobile Internet applications.