Separation Framework: An Enabler for Cooperative and D2D Communication for Future 5G Networks

I. INTRODUCTION

• Dense deployment strains energy efficiency and complicates interference management, as underlay D2D/M2M can restore intra-cell interference alongside inter-cell interference.The article also frames higher capacity and coverage demands as motivation for separation and cooperation.Source: §I · pp. 1–2
• The article surveys separation proposals across energy efficiency, capacity, interference, and mobility, then reviews cooperation and D2D before examining their potential in SARC.It also identifies challenges and opportunities where SARC, CoMP, and D2D intersect.Source: §I · p. 2, §II · p. 7

II. SEPARATION FRAMEWORK: PERFORMANCE MEASURES AND POTENTIAL GAINS

• SARC separates ubiquitous coverage from on-demand data service: cBSs provide coverage while dBSs provide high-rate data, allowing small cells to sleep without coverage holes.Unlike CARC, SARC can avoid handovers for inactive users, reducing signaling overhead.Source: §I · p. 2, §II · pp. 3–4
• Approximately 80% RAN power per BS switch-off is reported for, while reports Phantom Cell outperforms conventional FDD LTE picocells in spectral and energy efficiency.The article notes that these studies focus on energy and spectral efficiency and leave other architectural aspects unaddressed.Source: §II · pp. 4, 6

A. Energy Efficiency

• Switching off underused base stations can reduce power, but in coupled architectures it creates coverage holes; cell-range expansion can increase interference and backhaul demands.These trade-offs motivate changing the coupling of control and data planes.Source: A. Energy Efficiency · p. 7
• SARC is presented as supporting ubiquitous coverage with reduced transmit power and less control signaling per base station.The section links this separation to addressing the coverage problem associated with base-station sleep modes.Source: A. Energy Efficiency · p. 7
• Base stations consume around 80% of overall power, and remain power-hungry even at low load in dense urban deployments with variable traffic.The section notes that systems do not account for short- and long-term traffic variation in their design.Source: A. Energy Efficiency · p. 7
• The EARTH evaluation framework maps radio-frequency output power to total base-station supply consumption and compares cell types, traffic patterns, and deployment areas.It examines short- and long-term traffic variation to assess energy-saving potential.Source: 1) Evaluation Framework · p. 7

• Most wireless-network resources are idle, while supply power scales linearly with transmit/receive chains and base-station RF output power is nearly linear.Source: 1) Evaluation Framework · p. 7
• For macro BSs, power-amplifier consumption scales with load; the scaling is weaker in micro BSs and negligible in pico/femto BSs.A typical 3-sector site at zero load still consumes 50% of peak power.Source: 1) Evaluation Framework · p. 7
• The parameterized linear power model in represents savings from layout changes or operating measures such as reducing transmission power and deactivating unused antennas.Source: 1) Evaluation Framework · pp. 7–8
• The simpler model in does not discuss coverage-hole implications of sleep-mode operation.Source: 1) Evaluation Framework · p. 8
• For cooperative BSs, reports up to 11% fewer enabled BSs and an additional 39% activity-probability reduction at optimal spacing; does not consider user mobility.Source: 2) BS Switch-off · p. 8

• Traffic-profile switching can save energy, but increased transmit power may cause interference, and statistics-based switching best fits near-constant traffic such as nighttime.Source: 2) BS Switch-off · p. 8
• In SARC, the globally covering cBS can reactively or proactively wake sleeping dBSs without increasing their transmit power, avoiding the interference trade-off described for CARC.Source: 2) BS Switch-off · p. 8
• Area-power approaches can scale poorly as small-cell counts rise and rely on fixed-position mixed deployments, limiting their fit to existing deployments.Source: 2) BS Switch-off · p. 8
• The BS switching-on/off SWES algorithm [50] works distributively with reduced computational complexity and is claimed to offer around 50–80% potential savings for a real metropolitan urban traffic profile; proposed extensions include realistic BS power models and HetNets.Source: 2) BS Switch-off · p. 8
• SWES does not account for handed-over users’ QoS; in homogeneous deployments, neighboring BSs may need higher transmit power to maintain it, increasing inter-cell interference.Source: 2) BS Switch-off · p. 8
• The framework in reports 70–80% lower total energy consumption depending on traffic and deployment, but assumes no fast fading and Gaussian-like inter-cell interference.Source: 2) BS Switch-off · pp. 8–9
• Renewable-energy BSs combine renewable and non-renewable sources with an energy control unit, but the reference model in does not cover ECU design complexity.Source: 3) Renewable Energy Resources · p. 9

3) Renewable Energy Resources:

• The renewable-energy discussion includes solar, wind, and fuel-cell sources, alongside the solar- and wind-powered DoCoMo Eco Tower demonstration.Source: 3) Renewable Energy Resources · p. 9
• Smart-grid proposals use Stackelberg games for energy decisions, but BS switch-off can leave coverage holes, while CoMP coverage for all affected users faces backhaul limits and possible blockage or poor QoS.Source: 3) Renewable Energy Resources · p. 9
• SARC’s cBS coverage avoids CARC’s coverage holes and transmit-power increase after BS switch-off, and can serve remote locales off-peak as well as ultra-dense deployments.Source: 3) Renewable Energy Resources · p. 9

B. System Capacity

• Future capacity requirements include 25 Gb/s/Km2 average area throughput, with peak rates of 4.5 Gb/s/cell downlink and 2.5 Gb/s/cell uplink.Source: B. System Capacity · p. 10
• The article identifies spectrum efficiency, spectrum aggregation, and network densification as capacity approaches, describing densification as a globally accepted, cost-effective, and agile option.Source: B. System Capacity · p. 10
• SARC can use higher-frequency dBSs for capacity, but mm-Wave transmission faces poor propagation, blocking/shadowing, and atmospheric and rain effects.Source: B. System Capacity · p. 10
• SARC’s on-demand PID management lets the cBS localize sleeping dBSs, assign their IDs, and assist synchronization, at the cost of centralized management and tight synchronization.Source: B. System Capacity · p. 11
• Using global context, the cBS can select a high-capacity dBS for each UE and initiate wake-up, PID assignment, synchronization, and handover when that dBS sleeps.Source: B. System Capacity · p. 11

C. Interference Management

• Coordination shifts interference management from individual cells to multicell clusters, where inter-cluster interference limits performance.Frequency reuse can improve cell-edge performance but underutilizes spectrum; selective power reuse avoids that cost but offers no significant capacity gain, while invert power frequency reuse trades between them.Source: C. Interference Management · p. 12
• eICIC addresses interference in time, frequency, and power domains, with ABRBs offering finer time-frequency allocation than ABS.ABRBs also provide co-tier and cross-tier interference control; reducing low-power-node transmit power can degrade overall performance, especially for femtocells.Source: C. Interference Management · p. 12
• SARC can manage dense dBS interference through coordinated clusters, while long-range cBS clusters address cBS interference.The cBS and dBSs operate on different bands; cBS cell-edge users still experience inter-cell interference, and dBS interference is higher in ultra-dense deployments.Source: C. Interference Management · p. 12

D. Mobility Management

• Prior mobility studies omit cell sleeping, and the surveyed SDN handover discussions do not provide design guidelines.Other cited work examines macrocell cooperation, handover rates, phantom cells, or railway networks, with scope limitations including unstudied macrocell handovers or fast handovers.Source: D. Mobility Management · pp. 12–13
• SARC’s globally informed cBS can select and wake sleeping dBSs for handover, but this requires more processing and creates a coverage-wide failure risk if the cBS fails.The cBS may estimate sleeping-cell signal strength using distance-dependent path loss and statistical channel conditions; a self-organized backup cBS may mitigate the failure limitation. Channel conditions can make a farther sleeping dBS preferable to the nearest one.Source: D. Mobility Management · p. 13

1) Cell (re) Selection:

• Inactive UEs can move within cBS coverage without handover, while active data service requires selecting a dBS and may require waking a sleeping cell.The cBS can manage selection globally using UE-reported RSRP, but UE measurements exclude sleeping dBSs, so their channel conditions must be predicted.Source: 1) Cell (re) Selection · p. 13

2) Handover Procedures:

• SARC handover may be partial for cell-centre users, requiring only DPHO, while cell-edge users may need both CPHO and DPHO.This can reduce CPHO overhead for cell-centre users, but complete handover is more complex and sleeping dBSs may cause delay.Source: 2) Handover Procedures · pp. 13–14
• For cell-edge users, CPHO applies in active and inactive sessions; active-session handovers are challenging because control and data BSs are decoupled.Inactive-session CPHO can be initiated quickly without delay, according to the supplied passage.Source: 2) Handover Procedures · p. 14
• Before waking a dBS for DPHO, SARC may predict UE mobility to avoid waking cells for short visits and reduce ping-pong effects.The cBS may also need UE localization and channel-condition prediction for sleeping dBSs; the simplest CPHO/DPHO procedure is depicted in Figure 3.Source: 2) Handover Procedures · p. 14, Figure 3 · p. 14

• Cooperation aims for nodes to assist one another to improve coverage probability, interference management, or overall system capacity.Source: §III · p. 14
• Ad-hoc nodes self-organize without a central controller, whereas cellular cooperative nodes depend on serving BSs for control and connection management.Source: §III · p. 15
• D2D is described as infrastructure-based ad-hoc communication, with peers acting as relays or source nodes and little access- or core-network signaling.Source: §III · p. 15

A. CoMP Classification

• CoMP classification distinguishes uplink joint detection from downlink joint transmission, and centralized, decentralized, or distributed architectures.Coordination ranges from none to limited or full, corresponding to intra-cell beamforming, multicell CB, or fully coordinated CoMP.Source: A. CoMP Classification · p. 15, Figure 4 · p. 16

1) Joint Detection:

• In joint detection, cooperating BSs exchange requested-UE signals as IQ samples, soft coded bits, or decoded bits for serving-BS processing or selection combining.The transfer mode determines where equalization, demodulation, decoding, and CRC checking occur.Source: 1) Joint Detection · p. 16

2) Joint Transmission:

• In joint transmission, cooperating BSs exchange CSI and user data, design beamformers, and jointly transmit to the target UE.Coherent transmission is key to achieving maximum performance gain in this mode.Source: 2) Joint Transmission · p. 16

3) Centralized:

• A centralized controller uses CSI from participating base stations to find globally optimal precoding vectors and shares them for interference exploitation.In decentralized joint transmission, no base station has full global CSI, so globally optimal precoding is unavailable and solutions are sub-optimal.Source: 3) Centralized · p. 16, 4) De-centralized · p. 16

4) De-centralized:

• Any participating base station can act as the centralized node in the decentralized approach, without a dedicated central unit.Source: 5) Distributed · p. 16

5) Distributed:

• Non-cooperative CoMP exchanges no information between base stations; each uses limited UE feedback for interference-aware scheduling and IRC reception.Source: 6) Intra-cell Beamforming · p. 16

6) Intra-cell Beamforming:

• Coordinated beamforming exchanges CSI between cooperating base stations to reduce inter-cell interference with small backhaul capacity.Source: 7) multicell Coordinated Beamforming · p. 16

7) multicell Coordinated Beamforming:

• Full cooperation exchanges CSI and user data between cooperating base stations, enabling coordinated and coherent transmission but requiring very large backhaul capacity and strict synchronization.Source: 8) Full Cooperation · pp. 16–17

B. CoMP Clustering

• Cooperation clusters are limited in size by air-interface and backhaul signaling overhead, and may be static or dynamically adapted to channel conditions.Source: B. CoMP Clustering · p. 17
• Serving base stations can send UE RSRP measurements to the CoMP controller, which selects clusters using indicators such as load, delay, complexity, signal strength, or user priority.The approach can use existing 3GPP functions, including ANR and NRTs, for self-organizing clustering.Source: B. CoMP Clustering · p. 17, Figure 5 · p. 17

C. Decive-to-device Cooperation

• The paper frames D2D as a tier in ultra-dense HetNets, with proximity links offering potential capacity and coverage gains and network-assisted modes supporting content exchange.It identifies peer discovery and D2D data communications as two features, and cites Aura-Net as an earlier proximity-aware system.Source: C. Decive-to-device Cooperation · pp. 17–18
• Network-assisted D2D can support proximity services, sharing, relaying, offloading, and coverage or capacity enhancement, while bringing QoE, privacy, consent, battery, and interference challenges.The paper notes that relaying without owner consent or incentives may not be practical.Source: C. Decive-to-device Cooperation · pp. 17–18
• D2D may use licensed inband or unlicensed outband spectrum, and overlay allocation relaxes interference management but reduces frequency reuse and can waste cellular resources.Underlay sharing enables full frequency reuse but requires more sophisticated interference management.Source: C. Decive-to-device Cooperation · p. 18, Figure 6 · p. 18

IV. COOPERATION IN SARC

• The SARC cooperation discussion covers coordinated beamforming, D2D clustering and D2D CoMP, as well as cloud-RAN realization and fronthaul/backhaul constraints.Source: §IV · p. 18, Figure 7 · p. 19

A. Coordinated Beamforming

• In conventional multicell coordination, UE feedback and subsequent backhaul exchange delay interfering CSI, while quantization or noise perturbation can prevent perfect interference removal and yield sub-optimal performance.The paper compares desired and interfering channel exchange for SARC and conventional architecture in a three-cell model.Source: A. Coordinated Beamforming · p. 19
• The CSI model treats feedback delay as Dk and additional backhaul delay as Dk,x, with total delay Dbh = Dk + Dk,x and Dk,x ≥ Dk.The passage describes the channel errors as Gaussian and the channel autocorrelation using the Clarke model.Source: A. Coordinated Beamforming · p. 19
• SARC feeds CSI/ICI directly to the centralized base station, eliminating the cooperating-base-station backhaul delay so beamformers can be designed coherently.The paper notes that sharing those beamformers with distributed base stations over asymmetric backhaul links remains a problem, which it proposes addressing with centralized timing advance.Source: A. Coordinated Beamforming · pp. 19–20
• With global CSI at the centralized base station, joint beamforming can avoid exchanging interference information and adapt more rapidly to channel conditions than distributed design.Perturbation in the shared joint matrix affects the overall matrix, which the paper describes as producing fair system performance relative to distributed beamformer design.Source: A. Coordinated Beamforming · p. 20

B. D2D Clustering

• SARC's centralized context and CSI can support control of intra- and inter-cell interference from underlay D2D, while self-organized clusters and network control are proposed to improve capacity and spectral efficiency.Source: B. D2D Clustering · p. 20

1) Channel Conditions based D2D clusters:

• Ad-hoc D2D uses short-range point-to-point links for cached common information, whereas clustered D2D uses point-to-multipoint links for dissemination and is described as offering higher capacity gain.The considered interference sources include primary cellular users and intra- and inter-cluster D2D nodes; the serving base station may use nearby D2D CSI for power optimization and beamforming.Source: 1) Channel Conditions based D2D clusters · p. 20, Figure 8 · p. 20

2) Social network based D2D clusters:

• Channel-only clustering may overestimate spectral gains because it assumes every node shares common information, motivating social-aware selection of cellular, ad-hoc, or clustered modes.Closeness centrality is proposed to weight users; the examples assign U7 to cellular, U1 and U6 to ad-hoc, and U3 to clustered dissemination, while adaptive clustering may use predicted dwell times and mobility patterns.Source: 2) Social network based D2D clusters · pp. 20–21, 3) Prediction based adaptive D2D… · p. 21

C. D2D CoMP

• D2D CoMP applies multicell cooperation to nearby devices for common-information exchange or content dissemination in ad-hoc or clustered modes.The centralized base station can discover nodes using localization and shortest-distance or reduced-path-loss criteria.Source: C. D2D CoMP · p. 21
• For ad-hoc D2D, the cBS can obtain CSI through reference signals and feedback, then design beamformers for cooperative transmission.A zero-forcing or MMSE precoder is given as a possible approach.Source: C. D2D CoMP · pp. 21–22
• Clustered-mode CSI acquisition schedules influential nodes in separate TDMA slots for reference signals measured by the requesting node.The cBS then designs and shares precoders; the greater number of distributed nodes makes acquisition more complex and adds signaling overhead.Source: C. D2D CoMP · p. 22

D. SARC in Cloud-RAN

• SC-RAN places some RRHs at the cBS for ubiquitous coverage and uses the remaining RRHs for data services.It retains C-RAN’s radio, baseband, and fronthaul components while separating control and data planes.Source: D. SARC in Cloud-RAN · p. 22
• Coexisting control and data BBUs can share signaling, channel conditions, and user data, supporting joint processing and centrally managed adaptive clustering.The architecture also allows more flexible cell sleeping, load balancing, mobility management, and interference management, with potential energy-efficiency gains and reduced OPEX.Source: D. SARC in Cloud-RAN · p. 22
• SC-RAN’s flexibility requires costly fronthaul/backhaul, especially when CoMP exchanges large amounts of information among cooperating base stations.Distributed caching at femtocells, relays, or base stations is proposed to relax backhaul capacity requirements and could be incorporated into partially centralized C-RAN.Source: D. SARC in Cloud-RAN · p. 22

V. CONCLUSION

• The article surveys performance measures for SARC, but notes that early-stage research leaves some measures with little supporting literature.It discusses energy efficiency, system capacity, interference management, and mobility handling.Source: §V · p. 22
• The survey reports potential capacity and energy-efficiency enhancements from separating control and data planes, while mobility flexibility comes at the cost of more complex signaling.It emphasizes sending CSI to the cBS and using proactive caching to support backhaul-relaxed CB and CoMP.Source: §V · p. 23
• SARC provides a centralized cBS with greater flexibility for CSI acquisition and beamforming, and for predicting nodes for D2D content sharing.The article suggests combining this with proactive caching and adaptive clustering for self-organized D2D communication.Source: §V · p. 23
• Realistic power-consumption models are still needed to investigate the full energy-efficiency gains of SARC.The article proposes evaluating traditional-architecture approaches with such models before considering more advanced SARC energy-management techniques.Source: §V · p. 23
• Dual connectivity using mm-Wave for data and lower frequencies for control is identified as a research direction requiring new channel models.The article presents mm-Wave spectrum and carrier aggregation as potential candidates for future cellular networks.Source: §V · p. 23
• In ultra-dense networks, underlay systems such as D2D can bring back intra-cell interference, motivating device-level CB or CoMP and proactive caching to address interference and backhaul limits.The article points to caching approaches in.Source: §V · p. 23
• Spot-beam coverage from mm-Wave cells in ultra-dense settings is presented as a source of capacity gains that D2D cooperation could further harness for content sharing.The passage frames this as a future-network prospect.Source: §V · p. 23
• SARC’s flexible mobility management brings more complex signaling and control-plane design, while seamless handovers, coverage probability, and user QoS remain research needs.The added complexity relates to the greater number of underlay-network tiers.Source: §V · p. 23