Research On Energy Efficiency Optimization In MIMO Systems Under Imperfect CSIT

Energy efficiency (EE) has been one of the key performance indicators (KPI) of the fifth mobile communication systems (5G). Improving EE of wireless communication systems can considerably not only reduce the capital expenditure (CAPEX) and operation expenditure (OPEX) of mobile network operators (MNO), but also reduce the CO2emission of the information communication technology (ICT) to realize the low-carbon goal of the society. Besides, multiple input multiple output (MIMO) has become one of the key techniques of many wireless communication systems including5G, significantly improving the spectral efficiency (SE). Noting that the channel state information at transmitter (CSIT) of practical MIMO systems is usually imperfect, it is necessary to study the EE optimization schemes in MIMO systems under imperfect CSIT.This thesis focuses on the researches of energy efficiency optimization schemes in downlink MIMO systems under imperfect CSIT. Since precoding is the basic approach to obtain diversity and multiplexing gain, this thesis firstly studies the energy-efficient precoder design in multiuser MIMO (MU-MIMO) systems with bounded CSIT error. Besides, considering that there is a tradeoff between the training cost and energy efficiency, this thesis studies EE optimization schemes in training-based MU-MIMO systems and multicell MIMO systems, respectively.Firstly, precoding is an important approach to improve the performance of MU-MIMO systems. Thus, this thesis studies the energy-efficient precoder design in MU-MIMO systems with bounded CSIT uncertainty. In order to solve this non-convex fractional programming, this thesis first uses the relationship between the user rate and the Minimum Mean Square Error (MMSE) as well as the min-max inequality to transform the original problem into its lower bound problem with a max-min and fractional form. Then this fractional problem is rewritten with a parameterized subtractive form via the fractional programming theorem. This thesis further uses Lagrangian duality to transform the max-min problem into a max-max form and proposes an iterative algorithm with guaranteed convergence to solve it. Numerical results show that our proposed algorithm can obtain higher energy efficiency than existing schemes in MU-MIMO systems.Secondly, base stations (BSs) obtain CSIT with some training cost. Extremely-low training power leads to inaccurate CSIT, which would reduce EE. Extremely-high training power makes the rate improvement not compensate for the power consumption from the training, which would also reduce EE. Thus there is a tradeoff between the training power and energy efficiency, which needs further researches. This thesis studies energy efficiency optmization in training-based MU-MIMO systems and focuses energy-efficient power optimization problem. Considering the CSIT unavailable fact, this thesis proposes a two-step optimization scheme which consists of the ergodic EE and instant EE optimization. In the ergodic EE optimization, we firstly derive a tighter lower bound of the achievable rate, based on which the original problem is transformed into the optimization of the ergodic EE’s lower bound. This thesis proves this problem is jointly quasi-concave with respect to training power and data power, and then an alternating optimization algorithm is further proposed to solve it. In the instant EE optimization, transmitters transmit the training with the power obtained in the ergodic EE optimization, which is used for channel estimation and result in imperfect CSI. This thesis predicts the instant EE according to this imperfect CSI and further proposes an instant EE optimiztion algorithm to optimize the data power. Simulation results show that our proposed two-step EE optimization algorithm can improve EE in MU-MIMO systems compared to the SE scheme and the one-step scheme only with the ergodic EE optimization.Moreover, single cell processing (SCP) and coordinated beamforming (CBF) are two classical transmission modes in multicell MIMO systems. Different transmission modes may result in different EE in the case of all possible user positions. Thus, this thesis focuses EE optimization in training-based multicell MIMO systems, considering the minimum rate requirement. This optimization problem is a multiple-variable hybrid fractional programming with respect to the training power, data power and transmission mode. This thesis proposes the approximation expressions of the ergodic EE, based on which we propose a two-step algorithm to optimize the transmit power. In the first step, an alternating optimization algorithm is proposed to solve the energy efficient power allocation problem without the minimum rate requirement. In the second step, if the solutions in the1st step satisfy the minimum rate requirement, the two-step algorithm finishes. If not, the power optimization problem with the minimum rate requirement is transformed into a problem with an objective to minimize the total power consumption and a linear programming algorithm is proposed to it. Both algorithms have guaranteed convergence. After the optmization of the transmit power, this thesis proposes to obtain the energy-efficient transmission mode via the exhaustive search. Simulation results show that our proposed energy-efficient algorithm can get significantly higher EE than the SE scheme with the maximum transmit power when the minimum rate requirement is low. Moreover, the minimum rate requirement has an impact on the transmission mode selection. When the minimum rate requirement is high, the system tends to employ the CBF mode. The reason is that BSs needs to increase transmit power to satisfy the high-rate requirement, which increases the inter-cell interference (ICI) and makes EE limited by ICI. So BSs need to employ CBF mode to suppress ICI and improve EE in multicell MIMO systems.In this thesis, an energy-efficient precoding is firstly proposed in MU-MIMO systems with bounded CSIT error. After that, energy efficiency optimization schemes are proposed in training-based MU-MIMO systems and multicell MIMO systems, respectively. All the researches in this thesis have the important reference value on how to improve EE in MIMO systems under imperfect CSIT.