摘要
The downlink zero-forcing beamforming strategy in the case of random packet arrivals is investigated. Under this setting, the relevant fairness criterion is the stabilization of all buffer queues which guarantees a bounded average delay for all users. It has been shown that allocating resources to maximize a queue-length-weighted sum of the rates is a stabilizing policy. However, the high complexity of user selection and the feasible rates determination for optimal scheme may prevent the real-time scheduling operation. Two low complexity algorithms are provided taking the channel state, queue state and orthogonality into account. In particular, the authors pick the first user with the largest product between channel gain and queuing length, and select the remaining users to construct candidate user set based on the greedy user selection method or channel orthogonal user selection method. Then, the power and rate allocation for the selected users are implemented based on the modified water-filling method. The complexity of the proposed algorithms is analyzed. The average delay and average throughput are studied in homogeneous scenarios and heterogeneous scenarios, respectively. Simulation results show that the proposed algorithms can take full advantage of the multi-user diversity gain and provide average delay （or throughput） and fairness improvement compared with channel-aware-only schemes.
The downlink zero-forcing beamforming strategy in the case of random packet arrivals is investigated. Under this setting, the relevant fairness criterion is the stabilization of all buffer queues which guarantees a bounded average delay for all users. It has been shown that allocating resources to maximize a queue-length-weighted sum of the rates is a stabilizing policy. However, the high complexity of user selection and the feasible rates determination for optimal scheme may prevent the real-time scheduling operation. Two low complexity algorithms are provided taking the channel state, queue state and orthogonality into account. In particular, the authors pick the first user with the largest product between channel gain and queuing length, and select the remaining users to construct candidate user set based on the greedy user selection method or channel orthogonal user selection method. Then, the power and rate allocation for the selected users are implemented based on the modified water-filling method. The complexity of the proposed algorithms is analyzed. The average delay and average throughput are studied in homogeneous scenarios and heterogeneous scenarios, respectively. Simulation results show that the proposed algorithms can take full advantage of the multi-user diversity gain and provide average delay （or throughput） and fairness improvement compared with channel-aware-only schemes.
基金
supported by the National Natural Science Foundation of China (6077212, 60772166)
