Survey on positioning information assisted mmWave beamforming training

Abstract

mmWave communication is considered a major player for 5 G and beyond networks due to its wide unlicensed bandwidth. However, it comes with some challenges such as propagation and penetration losses. To overcome these issues, antenna beamforming is adopted. But it is difficult to estimate massive mmWave multiple input multiple output channel for active beamforming, so training using switched antenna arrays with structured codebook is used to perform mmWave beamforming training (BT). The conventional exhaustive search BT scheme wastes a long time and power, especially in pencil beams scenarios, hence, for a better communication, these issues have to be handled. Depending on available out off band information and motivated by the idea that mmWave is a location driven communication network, several mmWave BT schemes had been proposed in literature based on positioning information (PI). In this paper, we will make a survey on those works, clarifying how different positioning services contribute to reduce and relax BT complexity in mmWave communication. In this context, we will divide those studies into two main categories: (1) Straightforward positioning based BT schemes, which use PI for minimizing the searching space of a predefined codebook or reducing the complexity of channel estimation, (2) Positioning-based BT schemes using machine learning approaches, which classified as, first, statistical and probabilistic learning based schemes, secondly, other ML approaches based schemes. Moreover, we will discuss the effect of positioning and orientation errors, existence of obstacles, user mobility and storing information on the performance of BT. In addition, a comparison between all studies, will be presented considering several aspects such as the implementation cost. Finally, challenges facing these schemes will be discussed and several solutions will be suggested as possible future works.