Performance Analysis and Optimization of NOMA URLLC Systems With Imperfect CSI

By incorporating non-orthogonal multiple access (NOMA), the fairness and spectral efficiency of short-packet transmissions with respect to orthogonal multiple access (OMA) for ultra-reliable and low-latency communication (URLLC) can be significantly improved and make it attractive for the emerging Internet of Things. This paper investigates the two-user downlink NOMA for frequency division duplex URLLC systems, where the transmission latency is measured by blocklength. In light of the challenging demand for accurate channel state information (CSI) with limited pilot in a short frame, we consider the practical downlink channel estimation. We derive the closed-form approximation on the effective system throughput given power allocation coefficients and pilot length. To strike the tradeoffs among performance, complexity, and signaling overhead, we aim to jointly optimize the power allocation coefficients and pilot length to maximize the average effective system throughput with imperfect CSI at each user only, subject to the reliability requirement of each user, the transmission latency constraint, the total transmit power constraint, as well as the minimum required data rate of the users with the lower channel gain (i.e., the far user). We first develop the sufficient and necessary conditions in explicit closed-form for the formulated optimization problem to be feasible, and also analytically characterize the feasible region in explicit closed-form. These analytical results facilitate devising a low-complexity algorithm to obtain the optimal solution through statistical CSI-based static optimization. The numerical results of many different sets of system parameters are provided to validate the obtained analytical results and to show how the optimized pilot length varies with some main parameters.