On the Fundamentals of Power Allocation Strategies for Non-Orthogonal Multiple Access Downlink Systems
Abstract
The fundamentals of non-orthogonal multiple access (NOMA) power allocation strategies for downlink wireless transmissions are investigated, based on improving the performance over downlink transmissions using orthogonal multiple access (OMA). First, the scenario where the base-station possesses the perfect channel state information (CSI) is considered. For this scenario, the power allocation region is defined for a two-user downlink NOMA system. The expressions for the ergodic capacities and outage probabilities are derived for the case when the two users are pair randomly, each user with independently identically distributed (i.i.d.) fading channel gains. These results are then extended to the case where a cell-center user and cell-edge user are paired together, and a closed-form expression for the gap in ergodic capacity between NOMA and OMA is derived when SNR is large. This scenario is then extended to the general multi-user case, and it is proved that there always exists a power allocation strategy which allows all users to achieve a higher capacity when compared to OMA. In the second scenario, the effects of users with cached files on power allocation are investigated. When a user with weak channel condition has cached a file which a user with strong channel condition is requesting, these two users downlink transmissions are paired. An approximation of the optimum power allocation is derived, and the union-outage probability of this system is shown to improve over conventional NOMA and OMA. In the final scenario, the complete description of power allocation strategies is derived for more realistic wireless systems, where it is not assumed that the base-station possesses perfect CSI. Based on the target rates of each user, the fundamental properties of the power allocation strategy are derived, and the approach for selecting a strategy which improves the outage probability of each user is outlined. The existence of the power allocation strategies and their optimal energy efficiency is proved to be functions of the target rates and the OMA system parameters which they are compared to.