Geometric average-based fast-converging proportional fair scheduling for LTE downlink transmissions

Abstract

Long Term Evolution (LTE) wireless communication system has adopted Orthogonal Frequency Division Multiple Access (OFDMA) technology to provide downlink transmissions to multiple User Equipment (UE) in a single transmission time interval. In OFDMA, UEs provide quantized channel quality information, called channel quality indicator (CQI), to the base station. CQI feedback is then used at the base station to choose which UE to be scheduled and which modulation and coding scheme (MCS) for the selected UE to be used in downlink transmissions. Each UE sends averaged CQI over all subcarriers instead of sending the CQI of each subcarrier to minimize feedback overhead. As a result, only one MCS is applied at the base station for multiple subcarriers intended for the same UE. This can cause significant throughput degradation, especially in low SNR regions in which the actual CQI of each subcarrier may vary significantly. Our experimental results show that the two well-known scheduling algorithms, namely Best CQI and Proportional Fair Scheduling (PFS), fail to provide spectrum efficiency and/or fairness among UEs. These algorithms tend to provide a large fraction of bandwidth to a single UE; hence, the UE fails to demodulate transmitted data to which only one modulation scheme was applied. In this paper, we propose a novel scheduling algorithm that achieves better throughput performance, while providing more fairness among UEs than existing solutions in both high and low SNR environments. The proposed algorithm improves cell throughput by 66% over the PFS when UEs experience various SNR values. When UEs undergo similar SNR values, the system throughput can be improved by more than 120% over BestCQI and by almost 60% over PFS. Extensive simulations validate that our proposed solution also significantly enhance Block Error ratio (BLER) in addition to network throughput and fairness.