Introduction

12.1.1 Challenges

Owing to the increased level of frequency sharing, node density, interference, and network congestion in heterogeneous C-RANs, obtaining signal processing techniques and dynamic radio resource allocation optimization algorithms are the most important tasks [2]. Multi-user interference, which is a major performance-limiting factor, should be astutely manipulated through advanced signal processing techniques in the physical (PHY) layers. In addition, to a satisfy the quality-of-service (QoS) requirement, it is crucial to study radio resource allocation optimization in heterogeneous C-RANs, which is usually more challenging than that in a traditional cellular network, considering practical issues such as fronthaul capacity limitations, channel state information (CSI) overhead, and the parallel implementation of algorithms [3]. In heterogeneous C-RANs radio resource optimization algorithms should support the bursty mobile traffic data, which is usually delay-sensitive. Most traditional methods are based on heuristics and there is lack of theoretical understanding on how to design delay-aware radio resource allocation optimization algorithms in a time-varying system. Therefore, it is very important to consider random bursty arrivals and delay performance metrics, in addition to the conventional PHY-layer performance metrics, in cross-layer radio resource optimization, which may embrace the PHY, medium access control (MAC), and network layers [4]. A combined framework taking into account both queueing delay and PHY-layer performance is not trivial as it involves both queueing theory (to model the queue dynamics) and information theory (to model the PHY-layer dynamics). The system state involves both the CSI and the queue state information (QSI), and a delay-aware cross-layer radio resource optimization policy should be adaptive to both the CSI and the QSI of heterogeneous C-RANs. Furthermore, radio resource allocation optimization algorithms have to be scalable with respect to network size, while traditional algorithms become infeasible due to the in huge computational complexity as well as signaling latency involved [5]. The situation is even worse for heterogeneous C-RANs because there are more thin RRHs connected to the BBU pool via fronthaul links. Unlike conventional radio resource allocation optimization, which is designed to optimize the resource of a single base station, that for heterogeneous C-RANs involves radio resources from many RRHs or traditional macro base stations (MBSs), and thus the scalability in terms of computation and signaling is also a key obstacle.