GCC: Group-Based CSI Feedback Compression for MU-MIMO Networks

Multi-user Multiple Input Multiple Output networks (MU-MIMO) adopts beamforming to enable Access Point (AP) to transmit packets concurrently to multiple users, which brings formidable overhead. The overhead of collecting Channel State Information (CSI) feedback matrix may even overwhelm real data transmission when the scale of network is large, which incurs unsatisfactory performance and huge waste of resources. In this paper, we address this urgent problem with GCC, a Group-based CSI feedback Compression scheme for MU-MIMO networks, which enables users to feedback their CSI in terms of group determined by their location. Instead of traditional per-packet scheme, GCC limit the quantity of CSI feedback in each transmission round regardless of the size of network by allowing the location-related users to share a CSI matrix. We use a novel metric to do the tradeoff between throughput and capacity loss of the system. We realize GCC in different scenarios and compare it with exist works, evaluation result reveals that GCC can achieve much higher throughput and is robust to various situations.