1.- BALANCE GENERAL RESUMIDO

In [89] a novel separation architecture based on control and user (or data) plane split is proposed for capacity improvement in LTE networks but also incorporates energy

saving enabling features. In the proposed architecture, macro BSs provide continuous coverage while small cells BSs are introduced within the macro coverage to support high capacity demand. Macro BSs operate at the usual frequencies of LTE (below 2.5 GHz) and features both control and user plane signalling of conventional LTE. However, small cell BSs operate at higher frequencies of 3.5 GHz and higher and only support user plane signalling. Therefore, small cell BSs do not transmit the conventional control signals (e.g. cell specific reference signal and primary/secondary synchronization signals) needed by a mobile station or user equipment (UE) to associate with the small cell BSs. Hence, these small cells are referred to as Phantom cells. As shown in Figure 2.4, a UE can have a dual connection to a macrocell BS and a small cell BS at the same time, receiving control signals from the macrocell and data from the small cell. Macro BSs help transmit control plane signals on behalf of the Phantom cells and the connection of the UEs to Phantom cells are aided by Macro BSs in a sort of master-slave relationship through a new interface, X3. A new discovery signal for UEs to detect Phantom cells is proposed and it is assumed that the signal is time synchronised with the Macro BS to reduce the UE effort at detecting the Phantom cells. In order to save energy, it is expected that some of the Phantom cells can be switched off in networks with dense Phantom cell deployment.

Macro BS

Small cell BS

User Equipment

Figure 2.4 Control and Data Plane Separation

A similar architecture is proposed in [90], albeit with particular emphasis on energy saving rather than system capacity. Control signal transmission is separated from data transmission as well and transmitted by different types of BSs. On one hand, signalling (or control) BSs, optimised for long range and low data rate transmission, transmit control signals and maintain anytime, anywhere coverage utilizing a small

frequency bandwidth. On the other hand, data BSs optimised for short range and high data rate transmission handles data transmission only. As control signals require low data rates, signalling BSs are designed to be highly energy efficient with relatively larger coverage area compared to conventional architectures. In addition, the data BSs are activated on demand and switched off when no user is active in their vicinity. The activation of appropriate data BSs to serve a user request is done at the signalling BSs. Location information and past channel measurements during active sessions of BSs prior to deactivation are suggested for use in the selection process. The sort of simultaneous connection of a user to macrocell and small cell described in [89] above has also recently been standardised by the 3rd Generation Partnership Project (3GPP) [91]. 3GPP is the standard body that provides specification for the operation of 3G technologies and currently does the same for the state-of-the-art 4G LTE/LTE-Advanced networks. The simultaneous connection is referred to as dual

connectivity by 3GPP. Dual connectivity has been described as a process where a user can concurrently utilise radio resources from at least two access points (Master eNodeB and Secondary eNodeB) connected by non-ideal backhaul [91, 92]. A non- ideal backhaul has latency of several milliseconds to tens of milliseconds [93]. The dual connectivity concept has been proposed mainly to improve per-user throughput, provide robust mobility and reduce handover signalling in the small cell layer of such joint macro and small cells scenario described above [91, 92]. Unlike the previous proposals [89, 90], in dual connectivity both macro and small cells can serve as the Master node [94] and thus, the anchor point for control plane signalling. However, when the control plane is handled by the macrocell, mobility robustness is enhanced [92]. This is because loss of user connection due to incomplete handover process can be avoided when control signals, including handover commands, are transmitted by the macro layer [95]. In addition, when the macro layer handles the control signals and provide coverage, some small cells can be switched off at low load to save energy. Such a concept has been exploited in [32] and is discussed in section 2.7.3.

Some studies did not only present the energy efficiency or QoS benefit of the separation architecture HetNet but also evaluated the energy efficiency performance

as well unlike the previous studies above [89, 90]. In [49] a separation architecture is considered in a single cell scenario and achieved with the replacement of a conventional macro BS with a low power, coverage BS (CBS) and several small cell, traffic BS (TBS). While the CBS handles coverage, the TBS handles data services. Based on linear power models, the power consumption of the separation architecture is estimated using numerical computations and shown to achieve significant energy saving (more than 50%) relative to the conventional macro BS approach. Furthermore, closed form expressions are derived for the adaptation of TBS intensity to changes in user traffic intensity and the optimal TBS intensity for fixed user traffic intensity. The optimal TBS deployment and TBS intensity adaptation are shown to achieve almost 60% energy saving compared to the conventional macro BS.

Similarly, the joint optimization of density of BSs, number of antennas and spectrum allocation for energy efficiency in a separation architecture based HetNet of macro BSs and small cell BSs is investigated in [96]. The optimization problem is formulated and solved in two stages. Firstly, the optimal BS density and number of antennas for each BS tier are determined under the assumption of a known spectrum allocation, which is expressed in terms of the share of bandwidth allocated to each BS tier. The optimal spectrum allocation that results in minimum network energy consumption is then determined. The combination of small cell BSs and multiple antennas is shown to provide significant energy saving relative to a single antenna macro BSs only system. However, the most energy efficient of the two concepts is a function of the design target of the system.

Dense deployment of small cells is proposed to meet the high data traffic demands of the future; albeit macrocells are still required to provide umbrella coverage [13]. Both conventional and separation architecture based HetNets of macrocells and small cells are shown to achieve better energy efficiency in comparison with macrocell only deployments in the studies discussed above. However, apart from the energy efficiency gains of introduction of small cells alongside macrocells, there is benefit in adapting the network state in terms of active cells and resources utilised to instantaneous traffic demands. Dense deployments of small cells are believed to be essential to meeting the goals of the high capacity and high data rate demands envisaged in future wireless networks [97] (i.e. 5G and beyond). However, as the

traffic demand will not be at the peak value for most of the time, techniques to dynamically adapt the number of active cells (or BSs) and the resources (bandwidth, timeslot and power) to the current traffic load are essential. Resource management and topology management techniques that have been proposed in the literature for such dynamic adaptation in both conventional and separation architecture based HetNets are discussed in the next section. Although, there have been studies on homogeneous macrocell deployments, only those on HetNets, which is the focus of this thesis, are presented.