HANDBOOK ON GREEN INFORMATION AND COMMUNICATION SYSTEMS Chapter 6: Intercell Interference Coordination: Towards A Greener Cellular Network Duy T. Ngo, Duy H. N. Nguyen, and Tho Le-Ngoc McGill University Montreal, QC, Canada Elsevier 1 Introduction Intercell interference (ICI) is a critical issue in cellular communication systems. With universal frequency reuse, it is even more urgent to find effective solutions to this problem. Energy efficiency is crucial: Environmental effects of operating a huge number of cellular networks with high energy consumption Battery-powered user terminals have relatively short operating

time Towards a greener design, effective coordination of the intercell interference is key to Minimize the carbon footprint Maximize the overall network performance Elsevier 2 Introduction Cell coordination offers tremendous advantages over the traditional approaches that typically treat interference on a per-cell basis. This chapter reviews the state-of-the-art techniques that manage the intercell interference in multicell networks Elsevier

Homogeneous networks Coordinated multipoint transmission and reception (CoMP) Small-cell heterogeneous networks (femtocells) Energy-efficient interference coordination 3 Frequency Reuse and Interference Issue in Homogeneous Cellular Systems Broadcast nature of wireless medium results in the fundamental problem of interference. Fractional frequency reuse Reduced spectral efficiency Universal frequency reuse More spectrally efficient Only if intercell interference (ICI) is properly control. CDMA systems All users (UEs) share the same spectrum and are interfered. OFDMA systems

Joint subchannel assignment and power control is required to maximize system performance and reduce ICI. Elsevier 4 BS Coordination for Interference Management in Homogeneous Multicell Systems Conventionally, interference is usually controlled on a percell basis. The ICI is treated as background noise by each cell, and the base station (BS) of which has no intention to control the interference induced to other cells. Base station (BS) coordination is a more effective means to mitigate cochannel interference in multicell networks. Coordinated multipoint transmission and reception (CoMP) takes advantage of the inter-cell transmissions to enhance the overall system performance. Elsevier 5 Classification and Design Requirements for

CoMP Depending on the extent of coordination among cells, CoMP schemes can be classified into 3 categories: Joint Signal Processing (JP): multiple BSs are transmitting/receiving data signals to/from the UEs. Interference Coordination (IC): each UE transmits/receives data signals to/from its single serving BS. ICI is jointly controlled. Interference Aware (IA): ICI is not controlled, but is utilized to adjust the transmitting/receiving strategy at each BS. This scheme is a strategic noncooperative game (SNG). Different CoMP schemes impose different requirements on Data and signaling exchanges. Channel state information (CSI) knowledge needed at the coordinated BSs. Elsevier 6 Co-channel Deployment of Heterogeneous Small-cell Networks Small cells (i.e. femtocells) deployed at a home, connected to backhaul via

residential wireline links (e.g. DSL). Range of less than 50m and serve a dozen active users Elsevier 7 Key Benefits Higher capacity via larger area spectral efficiency Better coverage with lower power consumption Offload traffic for macrocell More cost-effective compared to cellpartitioning approach Cross-tier Interference in Femtocell Deployment Scenario A: A victim cell-edge macrocell user (MUE) is strongly interfered by the downlink transmission of a

nearby femtocell BS. Scenario B: Cross-tier interference can be severe and hard to control Elsevier 8 An MUE located far away from its serving macrocell BS transmits at high power in the uplink to compensate the path losses. This may jam the transmission of a nearby victim femtocell user (FUE). Challenges in Managing Interference for Femtocell Networks It is more challenging to mitigate inteference in femtocell than in traditional homogeneous settings. Unplanned deployment: Femtocells are deployed randomly without network planning that is normally

taken place. Femtocell BSs and users can be moved or switched on/off at any time. Access priority: Prioritized MUEs, the spectrum owner, need to be protected from cross-tier interference induced by lower-tier FUEs. Limited control/signaling: Residential network infrastructure only provide limited capacity for the exchange of control and signaling information. Delay can be a major issue. Elsevier 9 Interference Management in Femtocell Networks: Design Requirements Femtocell deployment: A paradigm shift from the traditional centralized macrocell approaches to a more uncoordinated and autonomous solution Available centralized solutions may not be applicable. Distributed interference management approaches are preferable in practical applications so that MUEs are robustly protected with their QoS requirements always maintained; and

FUEs effectively exploit residual network capacity to optimize their own performance. Elsevier 10 Interference Management Techniques in CDMAbased Homogeneous Cellular Networks (1) Power control is effective for CDMA-based systems SINR/Power balancing: can be implemented distributively, but diverges with infeasible SINR targets: imin Pi [t 1] Pi [t ] i Game-theoretical approach: users selfishly optimize their own performance, giving Nash equilibrium (NE), but not Pareto-efficient. max Pi 0 U i ( Pi , P i ) Game with pricing can substantially enhance the NE.

max Pi 0 U i ( Pi , P i ) Ci ( Pi , P i ) Elsevier 11 Interference Management Techniques in CDMAbased Homogeneous Cellular Networks (2) Using pricing scheme that is linearly proportional to Ci (i.e., Pi , P i ) i SINR, , NE is unique and Paretoefficient for single-cell settings. Observe: SINRs should not be fixed but adjusted to the extent that the system capacity can still support. A high SINR is translated into better throughput and reliability A low SINR implies lower data rates. Jointly optimize SINR and power to achieve Pareto optimality by Re-parametrization via the left Perron-Frobenius eigenvectors A locally computable ascent direction Elsevier 12

Interference Management Techniques in OFDMA-based Homogeneous Cellular Networks (1) Optimize over 2 dimensions Joint subchannel assignment and power allocation Typical design problem: max P, i wi ( m ,k ) log[1 m ,i ( m ,k ) ( P( k ) )] m k subject to : P( k ) Pmax Common approach: k Step 1: Given fixed power allocation P, find optimal subchannel assignment i* Step 2: Given fixed subchannel assignment i, find optimal power P* Go back to Step 1 and repeat until convergence.

Elsevier 13 Interference Management Techniques in OFDMA-based Homogeneous Cellular Networks (2) Game theoretical approach with virtual referee This referee mandatorily changes the game rules whenever needed, and helps improve the outcome of the game. Transmit power of UEs with unfavorable channel conditions are reduced. UEs generating significant interference to others may be prohibited from using certain subchannels. Low-complexity and heuristic approaches Affordable computational complexity Reduced feedback overhead Suitable for practical applications Elsevier 14

Coordinated Multipoint Transmission and Reception (CoMP) Consider a network with Q cells and K users. CoMP allows the data signals to a UE to be sent from multiple BSs. CoMP utilizes space division multiple access (SDMA) Each BS can send data signals to multiple connected UEs by means of precoding Beamformerw q for UE i at BS q. i Assuming each UE is assigned to a known subset of BSs. Elsevier 15 CoMP for Power Minimization (1) Interference Aware (IA) Each UE is assigned to only one BS BS adjusts its beamformers to ensure a set of target SINR at its connected UEs.

CoMP under IA scheme is a strategic noncooperative game. Players: BSs Admissible set of strategies: Constraints on the SINR at each UE. Utility function: Transmit power at the BSs min || w qi ||2 iK q The beam patterns are always unchanged, regardless the ICI power allocation game. Characterization of the NE: existence and uniqueness. Fully distributed implementation Elsevier 16 CoMP for Power Minimization (2) Joint Signal Processing (JP) and Interference Coordination (IC) Joint optimization to minimize transmit power across coordinated BS 2

min || w qi || q iK q Solution is Pareto-optimal. Convex optimization, easy to find the optimal solution. Multicell problem can be reformulated as a single cell problem well-known algorithms can be adopted. Drawbacks: Centralized implementation Signaling and synchronization between BSs Elsevier 17 CoMP for Power Minimization (3) Consider a new game Distributed implementation as in IA scheme Optimal solution as in IC scheme New utility function with pricing: min || w qi ||2 q j || WqH h q j ||2 iK q

jK q where q j : pricing factor charged on ICI caused by BS q to its unconnected UEs IA scheme with pricing Under the right pricing scheme, the new game approaches optimal performance offered by IC scheme. Elsevier 18 CoMP for Power Minimization (4) CoMP is more power-efficient than frequency reuse scheme. Elsevier 19 CoMP for Rate Maximization (1) Interference Aware: Each UE is assigned to only one BS BS adjusts its beamformers to maximize the data rate to its connected UEs.

CoMP under IA scheme is a strategic noncooperative game. Players: BSs Admissible set of strategies: Power constraint on the beamformers Utility function: Data rate at the BSs max log1 i iK q Nonconcave utility function difficult to analyze Apply zero-forcing (ZF) at each BS Simplify the game into a power iterative waterfilling game Easier to character of the NE: existence and uniqueness Elsevier 20 CoMP for Rate Maximization (2) Joint Signal Processing (JS) and Interference Coordination (IC) Joint optimization to maximize the data rate to all the UEs K max log1 i

i 1 Nonconvex optimization problem Difficult to find global optimum Approximation technique to find locally optimal solutions Solution approaches are usually centralized. IA scheme with pricing: new utility function with pricing max log1 i iK H 2 || W h || qj q qj jK q q

Under the right pricing, the network sum rate monotonically increases to a local maximum. Elsevier 21 CoMP for Rate Maximization (3) CoMP extracts higher sum-rate than frequency reuse scheme. Elsevier 22 Advanced Interference Coordination Techniques for CDMA-based Femtocells (1) Joint power and admission control for distributed interference management with dynamic pricing combined with admission control Net utility for MUE I to robustly protect the performance of all active MUEs: tot i U (i )

1 1exp[ bi ( i ci )] (m) i i a P Update of power for MUE i: Pi [t 1] max(i* , imin ) I i [t ] gi ,i Net utility for FUE j to balance the achieved throughput and the power expenditure: U Elsevier 23 tot

j (i ) W log(1 j ) a (f) j Pj Advanced Interference Coordination Techniques for CDMA-based Femtocells (2) For non-congested network, the proposed algorithm quickly converges to an equilibrium with the target SINRs achieved for all MUEs. For congested network, admission control can remove some FUEs, resulting in a noticeable growth in SINRs of the remaining FUEs. Removal of FUEs does not significantly affect the transmit powers and SINRs of MUEs. Elsevier 24 Advanced Interference Coordination Techniques

for CDMA-based Femtocells (3) Using convex optimization, distributed joint power and SINR allocation is devised such that All users attain their respective SINRs that are always optimal in Pareto sense, Every MUE i is protected with i . imin Every FUE j has its utility globally maximized. Key steps: Characterize Pareto-optimal boundary of the SINR feasible region Use load-spillage parametrization to realize every SINR point lying on such a boundary Determine a unique operating SINR point, based upon the specific network utility function of FUEs and the minimum SINR requirements of MUEs, Adapt transmit power according to Foschini-Miljanic's algorithm to attain such a design target. Elsevier 25 Advanced Interference Coordination Techniques for CDMA-based Femtocells (4) Proposed algorithm converges to global optima

for different utilities. Performance of the femtocell network optimized Minimum SINRs prescribed for MUEs always guaranteed MUE index 1 2 3 4 5 6 7 8 9

10 Target SINR 1.57 8 1.50 7 1.44 0 1.37 6 1.31 5 1.25 6 1.20

0 1.14 7 1.09 5 1.04 7 Achiev ed SINR 1.57 8 1.50 7 1.44 0 1.37

6 1.31 5 1.25 6 1.20 0 1.14 7 1.09 5 1.04 7 Elsevier 26 Advanced Interference Coordination Techniques

for OFDMA-based Femtocells (1) Joint allocation of resource block and transmit power Utility of each femtocell BS includes system capacity and other sources of interferences (i.e., femtocell to macrocell, macrocell to femtocell, and femtocell to femtocell). Formulated game belongs to the class of exact potential game, shown to always converge to a NE when a best response adaptive strategy is applied. Solution is an iterative process: Step 1: Optimal resource block allocation is determined given a transmit power policy. Step 2: Waterfilling allocation of power for femtocells is computed for a fixed resource block allocation. Go back to Step 1 and repeat until convergence. Elsevier 27 Advanced Interference Coordination Techniques for OFDMA-based Femtocells (2) Joint subchannel and transmit power allocation scheme Femto BSs are allowed to transmit on the same subchannel with MUEs as long as interference is limited to an acceptable level Maximizing capacity of cognitive radio network (e.g., femtocell)

ICI among different cognitive radio cells is controlled. Lagrangian dual method: Original design problem is decomposed into multiple subproblems in the dual domain Each problem is solved by an efficient algorithm. Duality gap approaches zero when the number of OFDMA subchannels is sufficiently large. Proposed solution outperforms the fixed subchannel allocation scheme. Elsevier 28 Tradeoff between Spectral and Energy Efficiency Spectral Efficiency (SE) SE R = log2 [1 P (W )][ b/s/Hz] W Energy Efficiency (EE)

EE Tradeoff relation: EE Elsevier 29 SE (2 1) SE R Wlog2 [1 P (W )] = [ b/J] P P With circuit power Pc R EE [ b/J] P Pc Energy-efficient Interference Management for

Multicarrier Multicell Networks Given interference powerI i( k ) on subchannel k, data rate of user i across all subchannels is (k ) g (k) P R i = log 1 (i,ik ) i Ii k EE of user i: Ri i Pi Pc Given the power allocation of all other users, P -i, each user i is required to solve the best-response problem: f n (P-i ) = arg max P i (Pi , P-i ) As i (Pi , P-i ) is strictly quasiconcave in Pi, there exists at

least one NE in this power control game. Under certain conditions, the NE is unique in frequencyselective channels i Elsevier 30 Energy-efficient Joint Power Control and BS Assignment in CDMA-based Multicell Networks Utility of user i received at its assigned BS ai: RL f (a i ,i ) u a i ,i (P) = M Pi Two-dimensional space: Transmit power Pi Base station ai Power control game with a linear pricing: max Pi ,ai uaci ,i ( P i , P i ) uai ,i ( P i , P i ) cai Pi Original problem is reduced to:

max Pi [max j j ,i ] ci Pi Improvement in EE with linear pricing is above 25% Elsevier 31 Chapter Summary Two conflicting goals in cellular network deployment: spectral v.s energy efficiency With universal frequency reuse, new communication paradigms are needed to proactively deal with intercell interference (ICI). Effective coordinating of ICI is the key to optimizing the two design goals towards a greener cellular network. For conventional homogeneous networks, CoMP schemes efficiently coordinate or even take advantage of the ICI. For heterogenous networks, advanced interference management mechanisms help mitigate cross-tier interference in mixed macrocell/femtocell deployment. Current advances in ICI coordination improve the energy efficiency of cellular networks while maintaining a good tradeoff with spectral efficiency goal. Elsevier

32 Some Potential Research Directions Design CoMP schemes that deal with quantization errors, fast-varying channels and CSI feedback delay Tradeoff between achieving optimal performance and incurring low computational complexity in CoMP Distributed implementation of robust CoMP schemes with only local CSI required Address energy-efficiency criterion in standardization of CoMP techniques Determine the optimal cell sizes and locations to deploy femtocell BSs, taking into account the energy expended for the backhaul and signaling overhead With cooperative relays, power-efficient resource allocation techniques (e.g., energy-efficient modulation, selective relaying) should be devised and adapted. Elsevier 33