KTH ROYAL INSTITUTE OF TECHNOLOGY Beam-searching and Transmission Scheduling in Millimeter Wave Communications Hossein Shokri-Ghadikolaei, Lazaros Gkatzikis, and Carlo Fischione Automatic Control Department and ACCESS KTH Royal Institute of Technology, Stockholm, Sweden and The Sixth Nordic Workshop on System and Network Optimization for Wireless (SNOW), Norway, 2015

Why millimeter wave frequencies? Ever growing demand for multi-Gbps data rates scarce spectrum resources in microwave bands However, as spectral efficiency is approaching its fundamental limits, we need to add more spectrum and increase deployment density without increasing the interference footprint. Enhancing spectral efficiency massive multiple input multiple output (MIMO) cognitive networks interference cancelation Enhancing spectral efficiency massive multiple input multiple output (MIMO) cognitive networks interference cancelation 2/27 explore centimeter/millimeter Wave (cmWave/mmWave)

Outline Introduction Problem formulation Numerical results Concluding remarks and future works

Characteristics of mmWave frequencies mmWave frequencies: 30–300 GHz 6–30 GHz are also often referred to as mmWave large bandwidth 4/27 high interaction with atmospheric constituents such as oxygen high path-loss (distance-dependent component of attenuation) typical coverage10 to 20 m in 60 GHz high attenuation through obstacles line of sight (LoS) communications short wavelength large number of antenna elements in the current size of radio chips high directivity gain T. Baykas, C. Sum, Z. Lan, J. Wang, M. Azizur Rahman, H. Harada, and S. Kato, "IEEE c: the first IEEE wireless standard for data rates over 1 Gb/s," IEEE Commun. Mag., v. 49, no. 7, pp , 2011.

Standardization activities WirelessHD consortium, wireless gigabit alliance (WiGig), ECMA 387 IEEE c (WPAN), IEEE ad (WLAN) 5/27 A coordinator schedules transmissions in a centralized manner channel access requests are registered with CSMA/CA protocol, and served with TDMA scheme Each time slot is assigned to a single transmitter-receiver pair  inefficient in mmWave networks with pencil-beam operation

Beamforming (1/2) Deafness: the main beams of a transmitter and the intended receiver are not aligned massive number of antennas (32/64 elements in existing WPAN products), so no digital beamforming! Analog beamforming only phase shifters and one RF chain, so only directivity gain finite size codebooks each covering a certain direction an exhaustive search over all possible directions select the combination of vectors that maximizes the signal-to-noise ratio no need for instantaneous CSI 6/27

Beamforming (2/2) Three-step alignment of the current standards a)a quasi-omnidirectional level search b)a coarse grained sector-level sweep c)a beam-level alignment phase 7/27 Alignment overhead: the time required to find the best beams It depends on the number of directions that have to be searched, which in turn depends on the selected transmission and reception beamwidths. J. Wang, et al. "Beam codebook based beamforming protocol for multi-Gbps millimeter-wave WPAN systems," IEEE J-SAC, v. 27, no. 8, pp , 2011.

Contributions Identifying alignment-throughput tradeoff in mmWave communications enabling concurrent transmissions in mmWave networks translating the proposed framework into protocols evaluating the performance gains arising from the proposed protocols 8/27 1.using extremely narrow beams (or excessively increasing the beamforming codebook size) is not beneficial due to the increased alignment overhead 2.very simple contention-based resource allocation may substantially outperform complicated contention-free resource allocation 1.using extremely narrow beams (or excessively increasing the beamforming codebook size) is not beneficial due to the increased alignment overhead 2.very simple contention-based resource allocation may substantially outperform complicated contention-free resource allocation

Introduction Problem formulation Numerical results Concluding remarks and future works

Alignment-throughput tradeoff Narrower beamwidths significant alignment overhead higher transmission rate due to higher directivity gains Larger beamwidths less alignment overhead reduced transmission rate 10/27

Problem formulation (1/3) A mmWave network with one coordinator and N links a path after the alignment procedure established sector-level alignment prior to the alignment phase per slot beam-level alignment 11/27 Alignment time: sector-level beamwidths of the transmitter and receiver of link i beam-level beamwidths of the transmitter and receiver of link i Pilot transmission time

Problem formulation (2/3) Discard the non-continuous ceiling function the alignment time cannot exceed total time slot duration T 12/27 ideal sectored antenna model:

Problem formulation (3/3) SINR at the receiver i 13/27 Sum-rate maximization Antenna beamwidths affect both τ i and SINR i transmission powers only affect SINR i the optimization problem is non- convex SINR i depends on the network topology

Single link scenario 14/27 The derivative has up to one root, so the throughout has up to one extremum. This extremum is a maximum (a simple gradient descent algorithm) This maximum is inside the feasible set, and not on the boundaries Adopting an extremely narrow or wide beam is not throughput optimal.

Multiple links scenario (1/4) 15/27 Adding spatial gain to the current standards by allowing concurrent transmissions proposing two topology-agnostic approaches to solve (6) decomposing a multiple links scenario into multiple single link scenarios substantial reduction of computational complexity a performance loss compared to (6) no power allocation, so a scheduling problem 1.Overestimation of interference conservative interference avoidance 2.Underestimation of interference ignore multiuser interference (noise-limited regime)

Multiple links scenario (2/4) 16/27 using OFDM transceivers (standard compliant) estimation of interference at sector-level, higher than actual interference

Multiple links scenario (3/4) 17/27 using OFDM transceivers (standard compliant) ignore multiuser interference (noise-limited network)

Multiple links scenario (4/4) 18/27 Interference over-estimatorInterference under-estimator Finding all independent sets NP-hard problem - Light computational complexity given independent sets *multiple executions of gradient descent algorithm Light computational complexity *multiple executions of gradient descent algorithm Underutilization of spatial resources *over-estimation of interference Some level of multiuser interference *mmWave systems are not necessarily noise-limited. Higher signaling overheadLower signaling overhead

Introduction Problem formulation Numerical results Concluding remarks and future works

Simulation setup 20/27

Contours of the throughput of a single link against transmission and reception beamwidths. Optimal hyperbola: Single link scenario (1/2) 21/27

Existence of the alignment-throughput tradeoff High beam-searching overhead with narrow beams (do not use pencil-beams!) Low directivity gain with wide beams Performance improvement with reduced pilot transmission overhead 22/27 Single link scenario (2/2)

Allocating only one channel per time slot (existing standards) is significantly inefficient. Inefficiency increases with the number of links. With 10 links, 525%, 401%, and 177% performance enhancement can be achieved by the Oracle, interference under-estimator, and over-estimator, respectively. MmWave WPANs are not noise-limited (gap between under-estimator and Oracle) 23/27 Single Link Activation: only the link of the highest SNR is activated Oracle: the solution of optimization problem (6) Multiple links scenario

Introduction Problem formulation Numerical results Concluding remarks and future works

Existing standards do not leverage full potential of mmWave communications We need to optimize the alignment-throughput Joint beamwidth selection and resource allocation: very high computational complexity exact network topology as an input Interference over-estimator: very high computational complexity substantially suboptimal Interference under-estimation: some level of interference increased optimality gap with the number of links Concluding remarks 25/27

Overhead of connection management (establishment, maintenance, recovery) with mobile users Frequent execution of connection recovery with pencil-beam operation Validity of noise-limited regime activating all links at the same time and frequency! Why do we need complicated contention-free resource allocations in mmWave networks? mmWave networks show a transitional behavior from a noise- to an interference-limited regime Future works 26/27

[1] H. Shokri-Ghadikolaei, L. Gkatzikis, and C. Fischione, “Beam-searching and transmission scheduling in millimeter wave communications,” Proc. IEEE ICC'15, [2]H. Shokri-Ghadikolaei, C. Fischione, G. Fodor, P. Popovski, and M. Zorzi, “Millimeter wave cellular networks: A MAC layer perspective”, submitted to IEEE Trans. Commun. (invited paper), Feb [3]H. Shokri-Ghadikolaei, C. Fischione, P. Popovski, and M. Zorzi, “Design aspects of short range millimeter wave networks: A MAC layer perspective”, KTH, Tech. Rep., [4]H. Shokri-Ghadikolaei and C. Fischione, “Transitional behavior of millimeter wave networks”, KTH, Tech. Rep., [5]T. Baykas, C. Sum, Z. Lan, J. Wang, M. Azizur Rahman, H. Harada, and S. Kato, "IEEE c: the first IEEE wireless standard for data rates over 1 Gb/s," IEEE Commun. Mag., v. 49, no. 7, pp , [6]T. BaykasJ. Qiao, X. Shen, J. Mark, Q. Shen, Y. He, and L. Lei, “Enabling device to device communications in millimeter-wave (5G) cellular networks,” IEEE Commun. Mag., vol. 53, no. 1, pp. 209–215, Jan References 27/27

KTH ROYAL INSTITUTE OF TECHNOLOGY Beam-searching and Transmission Scheduling in Millimeter Wave Communications Hossein Shokri-Ghadikolaei, Lazaros Gkatzikis, and Carlo Fischione Automatic Control Department and ACCESS KTH Royal Institute of Technology, Stockholm, Sweden and The Sixth Nordic Workshop on System and Network Optimization for Wireless (SNOW), Norway, 2015