Heterogeneous Doppler Spread-Based CSI Estimation Planning for TDD Massive MIMO

Abstract

Massive multi-input multi-output (MIMO) has been recognized as a key technology for fifth generation (5G) networks. Indeed, it enables to meet the challenging requirements of increased coverage, capacity and massive connectivity. Nevertheless, its performance is limited by channel estimation overhead which scales in time division duplexing (TDD) systems with the number of active users, consequently limiting the system’s efficiency. In this paper, we address the bottleneck of channel estimation in TDD Massive MIMO through optimized lean carrier. We propose an adaptive uplink training scheme that exploits the heterogeneous Doppler spreads of the different users in order to reduce the periodicity of uplink sounding signal transmission. The idea is to enable the network to plan its uplink training decisions for long time periods while taking into consideration user mobility. To this end, we formulate a two time-scale control problem that takes into account the different rates of the wireless channel and location changes. In the fast time scale, an optimized uplink training policy is derived based on estimated user locations. In the slow time scale, positioning decisions are optimized. Simulation results show that the optimized training policies provide considerable improvement in the system efficiency even with partial location knowledge.

Keywords

• Massive MIMO
• Doppler spread
• CSI estimation planning
• machine learning
• POMDP