Abstract
This paper proposes a joint transmitter-receiver design to minimize the weighted sum power under the post-processing signal-to-interference-and-noise ratio (post-SINR) constraints for all subchannels. Simulation results demonstrate that the algorithm can not only satisfy the post-SINR constraints but also easily adjust the power distribution among the users by changing the weights accordingly. Hence the algorithm can be used to alleviates the adjacent cell interference by reducing the transmitting power to the edge users without performance penalty.
Introduction
In a multiuser MIMO (Multiple-Input Multiple-Output) downlink, a base station transmits multiple data streams to several mobile users simultaneously. This architecture introduces two primary challenges: multiuser interference, where signals leak between users, and self-interference, where multiple data streams for a single user interfere with each other. While many solutions address either the transmitter (precoding) or the receiver (filtering) independently, this paper explores a joint design approach. The objective is to minimize the total weighted transmit power at the base station while ensuring each data stream meets a specific Signal-to-Interference-plus-Noise Ratio (SINR) requirement.
Methodology
The core of the proposed solution lies in the concept of Uplink-Downlink Duality. The authors demonstrate that the downlink optimization problem has a dual counterpart in a virtual uplink scenario. In this virtual uplink, the roles of transmitter and receiver are swapped, yet the optimal beamforming matrices remain consistent with the downlink.
To solve the joint optimization problem, which is inherently difficult when considering both filters simultaneously, an iterative algorithm is employed. This algorithm cycles through a four-step loop to achieve convergence.
The process involves fixing power and transmit filters to optimize receive filters in the downlink, then using duality to transform the problem into the virtual uplink to optimize the corresponding filters there. Finally, power allocation is updated, and the cycle repeats until the filters converge.
Experimental Setup
The algorithm's performance was evaluated using Monte Carlo simulations under i.i.d. flat Rayleigh fading channel conditions. The simulation configuration featured a base station with 8 antennas ($M=8$) communicating with multiple users. Each user was equipped with 2 antennas ($N_i=2$) and received 2 data streams ($L_i=2$), utilizing QPSK modulation.
Key Insights and Analysis
A pivotal insight from this work is that duality serves as a bridge, allowing the optimization problem to be solved in the domain that is mathematically most tractable. Furthermore, the introduction of a weight vector ($\mathbf{w}$) provides a mechanism for precise interference management. By adjusting these weights, network operators can control the power distribution among users. For instance, increasing the weight for a user at the cell edge allows the system to maintain their SINR while significantly reducing their transmit power, thereby minimizing interference to adjacent cells.
Results
The iterative algorithm successfully converges to solutions that satisfy all prescribed SINR constraints. A critical observation regarding system capacity is that performance is optimal when the number of base station antennas ($M$) meets or exceeds the total number of data streams ($KL$). As $KL$ approaches or exceeds $M$, managing residual interference becomes significantly more power-intensive.
The effectiveness of weight-based power management is demonstrated in simulations where increasing a specific user's weight from 1 to 20 led to a substantial 10dB reduction in their transmit power.
Crucially, this reduction for the "edge user" resulted in only a negligible 1dB increase in the total system transmit power, proving that neighboring cell protection can be achieved with minimal impact on local system performance.
Conclusion
While the proposed joint design offers significant advantages for power and interference management, it relies on the assumption of perfect channel state information (CSI). In practical scenarios, noisy or delayed CSI could impact performance. Additionally, the current model assumes flat fading; extending this to frequency-selective channels like OFDM would be a valuable direction for future research. Finally, the computational latency of the iterative approach remains an area for further analysis to ensure feasibility in real-time mobile hardware.
Summary updated