Power Allocation Schemes for Uplink Massive MIMO System in the Presence of Imperfect CSI

• Published in: IEEE transactions on signal processing Vol. 68; pp. 5968 - 5982
• Main Authors: Yu, Xiangbin, Du, Yuheng, Dang, Xiao-yu, Leung, Shu-Hung, Wang, Hui
• Format: Journal Article
• Language: English
• Published: New York IEEE 2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adaptive systems; Antennas; Complexity; Complexity theory; Constraints; energy efficiency; Estimation error; Fading channels; imperfect CSI; Lagrange multiplier; Massive MIMO; Mathematical programming; Maximum power; MIMO (control systems); MIMO communication; Multilayers; Multiple objective analysis; Optimization; power allocation; proportional fairness; spectral efficiency; Uplink
• ISSN: 1053-587X; 1941-0476
• DOI: 10.1109/TSP.2020.3029404

In this paper, spectrum efficiency (SE) and energy efficiency (EE) performances of uplink massive multiple-input multiple-output (MIMO) systems with continuous-rate adaptive modulation and imperfect channel state information (CSI) are investigated over Rayleigh fading channels. Given a target BER, closed-form SE and EE expressions are derived. With these results, a constrained non-concave optimization problem of power allocation (PA) to maximize the SE is firstly formulated subject to the maximum transmit power per user. A near-optimal PA scheme with the concave-convex procedure (CCCP) method is then developed, and it can offer nearly optimal performance close to that of the optimal scheme with exhaustive search, but require lower complexity. Since this scheme needs more iterations, we propose a suboptimal scheme with reduced complexity and slight performance loss, which utilizes the property of a large number of receive antennas to approximate the objective function as a concave one, thereby reducing the number of iterations. Secondly, the PA problem for EE maximization under the same constraints is addressed. Based on the CCCP method and fractional programming theory, we propose a near-optimal PA scheme to solve this problem, and it has the EE performance very close to the optimal exhaustive search scheme. However, multi-layer iterations make this scheme have relatively higher complexity. For this reason, we propose a low-complexity suboptimal scheme with small EE loss by using variable transformation and a large number of antennas. Based on the above results, we simultaneously optimize both EE and SE. A multi-objective optimization problem (MOP) subject to the maximum power constraint and proportional data rate constraint is formulated. By means of the weighted sum method, the complicated MOP is transformed into a simple single-objective optimization problem. Then a PA scheme is proposed by using the Lagrange multiplier method to balance the EE and SE effectively. Simulation results validate the effectiveness of the proposed PA schemes in term of high SE and EE, and illustrate the fundamental tradeoff between EE and SE for different parameter settings.