Capacity Analysis And Key Technologies Of Amorphous Cellular Networks

In face of the multiple communication scenarios of the future wireless communication sys-tems,such as wide area coverage,extreme throughput and low-cost massive connectivity,some breakthroughs in the wireless transmission technology and network architecture are urgently needed in the fifth generation mobile communications technology?5G?.By spreading a large number of remote radio heads over a wide area,the amorphous cellular network are able to im-plement network in a flexible manner,namely,”user-centric”network.In amorphous cellular network,the network structure is dynamically adjusted according to traffic needs and charac-teristics,which perfectly copes with the various scenarios in the future wireless communication systems.In view of this,amorphous cellular network has great potential to become the future network architecture.However,there are still technical bottlenecks in the architecture,such as,complicate transmission environment,high hardware costs,dynamic clustering and so on.In this paper,we aim to provide detailed analysis to these key problems,and investigate different networking manners according to communication scenarios in amorphous network.First,two important mathematical tools are introduced:finite-dimension random matrix theory?F-RMT?and large-dimension RMT?L-RMT?.Amorphous networks cooperate distribut-ed antennas to form the cellular,which presents the properties of distributed structure and mul-tiple input multiple output?MIMO?wireless cahnnel.In consideration of the close relationship between MIMO channels and random matrices,by leveraging F-RMT,the probability density function of the unordered eigenvalue of the Wishart matrix for uncorrelated Rayleigh,semi-correlated Rayleigh,and uncorrelated Riciean channels are deduced,as well as its arbitrary order moment.When the matrix dimension is sufficiently large,using the results of L-RMT large-dimension random matrices theory,the deterministic equivalent results of the maximum combining ratio?MRT?,zero-forcing?ZF?precoding,and regularized zero-forcing?RZF?pre-coding strategies are provided.The theoretical results reveal the relationship between channel matrix parameters and MIMO system performance,and provide important theoretical basis for the performance analysis for amorphous cellular networks.Then,from the perspective of improving wide area coverage,a low-cost solution to ap-proach amorphous cellular network is proposed,namely,hybrid distributed massive MIMO?m-MIMO?system.The spectrum efficiency is investigated according to uplink and downlink trans-mission schemes.In the hybrid distributed mMIMO system,via replacing the mixed-resolution radio frequency unit and the fiber-optic hybrid fronthaul network by high-cost,pure ideal ra-dio frequency unit and fiber fronthaul,the deployment cost and the power consumption can be greatly reduced.After deducing the uplink achievable rate with maximum combining ratio receiver,a downlink precoding strategy for a hybrid architecture is designed.Based on the re-sults of previous finite-dimensional RMT and large-dimensional RMT,per-user achievable rate is studied when apply maximum combining ratio transmission?MRT?,zero-forcing?ZF?pre-coding,and regularized zero-forcing?RZF?precoding on the access link,respectively.Based on the theoretical results,the optimal ratio of high-to-low resolution radio frequency unit is de-rived for achieving optimal energy efficiency.The simulations demonstrate that the proposed architecture can achieve same-level throughput of ideal system configuration with much high-er energy efficiency,and the optimal ratio of high-to-low resolution radio to achieve optimal energy efficiency is derived from simulation results.Next,regarding to the extreme throughput requirement,an ultra-dense C-RAN based“user-centric”network is proposed,where each user connects to a few nearby remote radio heads to form its own cell.Utilizing the results from previous finite-dimensional random matrix theory,we study the ergodic capacity of such amorphous cellular networks at high signal-to-noise ra-tios where remote radio heads are modeled by a Poisson point process.Specifically,we derive tractable approximations of the ergodic capacity at high-SNRs for arbitrary antenna configu-rations,and tight lower bounds for the ergodic capacity when the numbers of antennas are the same at both ends of the link.The impact of the system parameters on the ergodic capacity is investigated.By leveraging our analytical results,we propose two efficient scheduling algo-rithms for RRH selection for energy-efficient transmission.The numerical results demonstrate that a large number of deployed RRHs can be waived while still maintaining a desirable energy efficiency level by applying our scheduling algorithms.Finally,for the low-cost massive connectivity scenarios,based on the hybrid distributed massive MIMO,an“user-centric”based amorphous cell is designed as well as the correspond-ing access solution.Such solution aims to maximize sum rate,which consists of three steps,including user activity detection,channel estimation and dynamic clustering.Starting from the user activity detection and channel propagation coefficient estimation,a low-complexity alter-native direction multiplier of method?ADMM?is proposed that can solve the problems at same time.Using the large scale fading?LSF?parameters obtained from the estimated channel coeffi-cients,the dynamic clustering algorithm first forms the LSF into observing space and determines the number of clusters N.With N as a priori knowledge,the clustering algorithm then aims to maximize the achievable sum rate using reweighted-WMMSE,in which the power consumption and fronthaul capacity constraints are modeled using a?₁-norm approximation.The simulation results verify that ADMM algorithm can effectively detect user activity,and the accuracy bound in a dense system can be almost achieved with sufficient pilot length.Regards to dynamic clus-tering,simulations show that the proposed algorithm has strong convergence,and significant performance gains can be obtained in various scenarios.