Space Division Multiple Access with a Sum Feedback Rate Constraint

TL;DR

This paper introduces a novel SDMA scheme that maintains high throughput in multi-antenna broadcast channels while keeping the total feedback rate constant, effectively balancing capacity and feedback constraints.

Contribution

The paper proposes a new SDMA design with feedback rate constraints, including algorithms for CSI quantization, threshold-based feedback, and joint beamforming and scheduling.

Findings

Achieves near-optimal sum capacity growth despite feedback constraints.

Feedback overflow probability decreases exponentially with increased feedback rate.

Outperforms existing SDMA schemes in sum capacity under feedback limitations.

Abstract

On a multi-antenna broadcast channel, simultaneous transmission to multiple users by joint beamforming and scheduling is capable of achieving high throughput, which grows double logarithmically with the number of users. The sum rate for channel state information (CSI) feedback, however, increases linearly with the number of users, reducing the effective uplink capacity. To address this problem, a novel space division multiple access (SDMA) design is proposed, where the sum feedback rate is upper-bounded by a constant. This design consists of algorithms for CSI quantization, threshold based CSI feedback, and joint beamforming and scheduling. The key feature of the proposed approach is the use of feedback thresholds to select feedback users with large channel gains and small CSI quantization errors such that the sum feedback rate constraint is satisfied. Despite this constraint, the proposed SDMA design is shown to achieve a sum capacity growth rate close to the optimal one. Moreover, the feedback overflow probability for this design is found to decrease exponentially with the difference between the allowable and the average sum feedback rates. Numerical results show that the proposed SDMA design is capable of attaining higher sum capacities than existing ones, even though the sum feedback rate is bounded.