LTE in Unlicensed Spectrum

LTE in Unlicensed Spectrum Prof. Geoffrey Ye Li School of ECE, Georgia Tech. Email: liye@ece.gatech.edu Website: http://users.ece.gatech.edu/liye/ Contributors: Q.-M. Chen, G.-D. Yu, and A. Maaref

Outline 2 / 53 1 Background and Challenges 2 Traffic Offloading or Resource Sharing? 3 Mobile Data Offloading 4 Energy Efficiency Optimization 5 Summary

Background 3 / 53 The ever-increasing data rate requirement for 5G Networks. Some proposed techniques: Massive MIMO, Small cell, Device-to-device communications. LTE in unlicensed spectrum (LTE-U) Improve user experience for existing unlicensed devices. Increase cellular operators capacity.

Challenges Harmonious coexistence among different systems. Fair and efficient spectrum sharing unlicensed spectrum. Ensuring QoS for LTE traffic. 4 / 53

Outline 5 / 53 1 Background and Challenges 2 Traffic Offloading or Resource Sharing? 3 Mobile Data Offloading 4 Energy Efficiency Optimization 5 Summary

Introduction: Traffic offloading 6 / 53 Using unlicensed bands to deliver cellular data traffic. How to guarantee QoS of cellular traffic? No guarantee on QoS due to DCF protocol in WiFi (unlicensed) band. Carefully selecting offloaded traffic to avoid saturation and excessive packet collisions.

7 / 53 Introduction: Resource sharing Transmitting cellular signals on the unlicensed spectrum. Advantages: Higher spectrum efficiency Better QoS Challenges: Effective resource sharing strategies for cellular and WiFi traffic.

Introduction: Existing works Mobile data offloading Inter-network: Offloading cellular traffic to WiFi networks [K. Lee, IEEE Trans. Netw., 2013]. Intra-network: Offloading macro base station traffic to femtocells [S. Yun, JSAC, 2012]. Resource sharing Joint radio resource management on licensed and unlicensed bands [A. R. Elsherif, JSAC, 2015]. Resource management for energy efficiency LTE-U networks [Q. Chen, JSAC, 2016]. 8 / 53

Introduction: Challenges Traffic offloading or resource sharing? Considering both? Strategy for LTE & WiFi networks. Q. Chen, G. Yu, H. Shan, A. Maaref, G. Y. Li, and A. Huang, Cellular meets WiFi: Traffic offloading or resource sharing, to appear in IEEE Trans. Wireless Commun./also in IEEE Globecom 2015. 9 / 53

Three Different Methods 10 / 53 Traffic offloading: SBS offloads some users to WiFi AP. Resource sharing: SBS occupies some time slots from WiFi AP. The hybrid method: SBS offloads some users to WiFi AP and occupies some time slots.

Single SBS & AP: System Model 11 / 53 SBS1 AP1 One time slot on unlicensed band WiFi AP SBS N: WiFi user number in AP, N S : User number in SBS, N A : User number in AP.

WiFi Throughput The saturation throughput of the WiFi network: R(n) = P tr P s E {P} (1 P tr ) T σ + P tr P s T s + P tr (1 P s ) T c, P tr : probability that at least one transmission in a slot time. P s : probability that a transmission is successful. T s : average time that the channel is sensed busy because of a successful transmission. T c : average time that the channel is sensed busy by each station during a collision. T σ : duration of an empty slot time. E {P}: average packet size. G. Bianchi, Performance analysis of IEEE 802.11 distributed coordination function," IEEE J. Sel. Areas Commun., vol. 18, no. 3, pp. 535-547, Mar. 2000. 12 / 53

Single SBS & AP: Traffic Offloading Traffic offloading only: L = 0 (LTE occupies no WiFi time slots) subject to max N C S N S N R ( N A + N ) (N A + N) R T C S : Average throughput of SBS on the licensed band. R T : The minimum per-user WiFi throughput. L: Occupied time slots. Results Average per-user throughput: C S a. N S min{n,n max } N S min {N, N max }: maximum offloaded user number. a N : the largest integer satisfying R(NA +N) N A +N R T. 13 / 53

Single SBS & AP: Resource Sharing 14 / 53 Results Resource sharing only: N = 0 (LTE offload no user to WiFi) subject to C S + C A L max L N S R ( N A) (1 L) N A R T C A : Average throughput of SBS on the unlicensed band. Average per-user throughput: CS +C A L. N S L = 1 RT N A : maximum time slots to LTE. R(N A )

Single SBS & AP: Hybrid method 15 / 53 Consider both traffic offloading and resource sharing C S + C A L max N,L N S N subject to R ( N A + N ) (1 L) (N A + N) R T Results Maximum average per-user throughput max 0 N N maxf (N) = CS + C A N S N C A R T (N A +N) R(N A +N)

Traffic Offloading vs Resource Sharing 16 / 53 Traffic offloading performs better than resource sharing only if the number of existing users in WiFi is small enough.

Hybrid Method vs Resource Sharing 17 / 53 When N A is large enough, offloading users to WiFi is no longer necessary and the hybrid method is identical to the resource sharing.

Multiple SBSs & APs: System Model SBS3 SU 31 AP1 AU 1 SBS1 SU 11 SU 12 SU 32 F 1 F 3 F 2 SU 13 AP3 SU 14 AU 3 AP2 SU 21 SU 26 AU 2 SU 24 AU 2 F 2 F 3 SU 22 SBS2 One time slot on unlicensed band SU 23 SU 25 WiFi AP SBS M: number of SBSs, K: number of WiFi APs, N k : WiFi user number in AP k. N S m: LTE user number in SBS m, N A k : LTE user number in AP k. 18 / 53

Multiple SBSs & APs: System Model 19 / 53 Assumptions SBSs and APs: located randomly according to the Poisson point process models. Goals The WiFi network supporting the IEEE 802.11n protocol. 1 Maximizing the LTE throughput while guaranteeing the throughput of each WiFi user. 2 Maximizing the minimum average per-user throughput of LTE SBSs to ensure fairness.

Multiple SBSs & APs: Problem Formulation Average per-user throughput among all SBSs: K Cm S + Cm A max min k=1 {N mk,l mk } m Nm S K N mk k=1 subject to ( R Nk A + M ) ( N mk 1 M ) L mk m=1 m=1 ( Nk A + M ) R T k, k, N mk m=1 N mk N max mk, m, k. L mk Minimum per-user WiFi throughput limitation. Maximum offloaded user number limitation. C S m: Average throughput of SBS m on the licensed band. C A m: Average throughput of SBS m on the unlicensed band. 20 / 53

Simulation Parameters 21 / 53

Single SBS & AP: Performance LTE throughput (Mbps) 8 7.5 7 6.5 HB; R T =8 Mbps TO; R T =8 Mbps RS; R T =8 Mbps HB; R T =10 Mbps TO; R T =10 Mbps RS; R T =10 Mbps WO 6 When WiFi user number is small: TO > RS. HB > {TO, RS}. 5.5 1 2 3 4 5 6 7 8 9 10 N A When WiFi user number is large: RS > TO. HB = RS. 22 / 53

Multiple SBSs & APs: Performance 23 / 53 Dash line: Upper bound of each minimum SBS. Fairness: Almost the same performance for each SBS.

Conclusions Offloading LTE traffic to WiFi network. Sharing unlicensed spectrum with LTE. WiFi performance is degraded but a minimum threshold is guaranteed. Any win-win approach? 24 / 53

Outline 25 / 53 1 Background and Challenges 2 Traffic Offloading or Resource Sharing? 3 Mobile Data Offloading 4 Energy Efficiency Optimization 5 Summary

Traditional Data Offloading 26 / 53 Traditional offloading: transferring cellular traffic to WiFi networks. The WiFi network: less spectral-efficiency due DCF & package collision. How about operating unlicensed spectrum by LTE?

Motivation: Novel Data Offloading Novel data offloading: Transferring WiFi users to LTE (opposite to traditional data offloading). Relinquishing more unlicensed spectrum to LTE. A win-win situation Better QoS of the transferred WiFi users. Better performance for the remaining WiFi users due to fewer packet collision. More unlicensed spectrum with efficient management in LTE. Q. Chen, G. Yu, A. Maaref, G. Y. Li, and A. Huang, Rethinking mobile data offloading for LTE in unlicensed spectrum, to appear in IEEE Trans. Wireless Commun./also in IEEE WCNC 2016. 27 / 53

Challenging Issues 28 / 53 1 How many and which WiFi users to be transferred to the LTE network? 2 How much unlicensed resources to be relinquished to the LTE-U network? 3 What if there are many WiFi APs?

Three Different User Transfer Schemes 29 / 53 Random transfer (RT): randomly selecting WiFi users to transfer. Distance-based transfer (DT): based on the distance between each WiFi user and SBS. CSI-based transfer (CT): based on CSI.

30 / 53 Single AP: System Model BS AP Unlicensed band in LTE Unlicensed band in WiFi Users in WiFi system Users in LTE system

Single AP: Problem Description 31 / 53 1 WiFi benefit (the improvement of per-user WiFi throughput): z w. 2 LTE benefit for the leftover unlicensed time slots z c. 3 Objective: balance the WiFi and LTE benefits. The win-win problem n: Transferred user number, max z c z w {n,ρ} ρ: Relinquished unlicensed spectrum.

Single AP: Insights Insights More WiFi benefit if more WiFi users are transferred. Less LTE benefit with more WiFi users transferred. The NBS strategy: balance between the two benefits. 32 / 53

Multi AP: System Model F2 AP LTE SBS F3 AP F1 AP Unlicensed time slots for LTE Unlicensed time slots for WiFi WiFi users transferred to LTE WiFi user LTE user K: number of WiFi AP. L k : distance between WiFi AP k and SBS. N k : WiFi user number in AP k. 33 / 53

Multi AP: Problem Description 34 / 53 1 WiFi benefit (the improvement of per-user WiFi throughput for AP k): z w k. 2 LTE benefit for the leftover unlicensed time slots z c. 3 Objective: balance the WiFi and LTE benefits. The win-win problem K max z c z w k, {n k,ρ k } k=1 n k : Transferred user number from AP k. ρ k : Relinquished unlicensed spectrum from AP k.

Multi AP: Insights 35 / 53 Fact 1 The WiFi APs near the SBS will transfer more users than those WiFi APs far away from the SBS. Fact 2 The optimal transferred users in the multi-ap case are in the range of [n c k, nw k ]. (nw k and nc k are the user number that maximize the WiFi and LTE benefits, respectively.)

WiFi benefit (Mbps) LTE benefit Single AP: Benefits for both WiFi and LTE 0.5 0.4 0.3 RT, Max DT, Max CT, Max RT, NBS DT, NBS CT, NBS 0.1 0.08 0.06 RT, Max DT, Max CT, Max RT, NBS DT, NBS CT, NBS 0.2 0.04 0.1 0.02 0 16 21 26 31 36 L (m) 0 16 21 26 31 36 L (m) The maximum WiFi (LTE) benefit is the upper bound for WiFi (LTE) benefit. The CT: the best performance. The RT: the worst performance. 36 / 53

Transferred user number 37 / 53 Single AP: Transferred User Number 40 35 30 25 20 15 10 n w, RT n NBS, RT n c, RT n w, DT n NBS, DT n c, DT n w, CT n NBS, CT n c, CT The maximum WiFi benefit: the upper bound. The maximum LTE benefit: the lower bound. 5 0 16 21 26 31 36 L (m) n c n NBS n w : confirm Fact 2.

Multi AP: WiFi Benefit and Fairness 0.5 0.4 Max WiFi NBS Benchmark WiFi benefit (Mbps) 0.3 0.2 0.1 0 WiFi AP1 WiFi AP2 WiFi AP3 WiFi AP4 (31m) (26m) (21m) (16m) Benchmark: Converting the K-dimensional searching into K 1-dimensional searching. WiFi AP closes to SBS achieves higher WiFi benefit. 38 / 53

Conclusions Transfer WiFi users and relinquish unlicensed resources to the LTE-U network (Subversively). Benefits both LTE and WiFi systems. The benefits: depend on the distance between APs and SBS, and the number of WiFi users in each AP. 39 / 53

Outline 40 / 53 1 Background and Challenges 2 Traffic Offloading or Resource Sharing? 3 Mobile Data Offloading 4 Energy Efficiency Optimization 5 Summary

Introduction Licensed-assisted access (LAA): Associating users with both licensed and unlicensed spectra under a unified LTE network infrastructure. EE in LAA systems: May degrade EE of the LTE system. How to allocate resource blocks (RBs) to improve the EE? How to jointly allocate licensed and unlicensed RBs to achieve EE fairness among SBSs? Q. Chen, G. Yu, R. Yin, A. Maaref, G. Y. Li, and A. Huang, Energy efficiency optimization in licensed-assisted access, to appear in IEEE J. Sel. Areas Commun. also in IEEE PIMRC 2015. 41 / 53

The System Model for LAA T T seconds seconds LTE ON LTE OFF LTE ON LTE OFF U One frame (10 ms)= M RBs M U kj RBs U M RBs LTE ON: for LTE. LTE OFF: for WiFi. 42 / 53

System Model AP 31 AP 11 AP 12 SBS1 Lic Unlic Lic Unlic SBS3 MBS Lic SBS2 Lic Unlic AP 23 AP 21 AP 22 LTE users Lic: Licensed band Unlic: Unlicensed band M L : RB number in licensed band. M U : RB number in unlicensed band. K: number of SBSs. A k : Number of WiFi access points in SBS k. N k : number of SBS users in SBS k. ρ kj : time slots on unlicensed band j occupied by SBS k. 43 / 53

EE Optimization Energy efficiency of each SBS k: N k α kn R L kn + N k A k β knj R U knj n=1 n=1j=1 η k = P c s + ϖs L N k α kn P L k + N k A k ϖu s β knj P U kj subject to K N k α kn M L, N k k=1 n=1 β knj ρ kj M U, k, j, n=1 A k n=1 α kn R L kn + β knj R U knj R min n, n, k. P L k j=1 : transmit power on the licensed band. α n=1j=1 limitation of total RBs on licensed band limitation of fair resource sharing on unlicensed band limitation of minimum data rate kn: RBs on licensed band. P U kj : transmit power on the unlicensed band. β knj: RBs on unlicensed band. 44 / 53

Properties of EE EE of each SBS increases with the licensed RBs. Don t use unlicensed bands if enough licensed ones. Improve EE by utilizing unlicensed bands only for small number of allocated licensed RBs. 45 / 53

Joint Licensed & Unlicensed Resource Allocation Maximize the EE for each SBS: Solutions η max = max {η 1, η 2,, η K }, {α,β} Based on Weighted Tchebycheff method: Several Pareto optimal solutions. Based on Nash Bargaining Solution: Fair EE RB allocation algorithm. 46 / 53

Pareto optimal solution set vs NBS solution 47 / 53 2, max max 1 2 NBS min 2 min 1 1 Results: The fair EE is also a Pareto optimal EE.

Simulation Results 48 / 53

Performance for Individual SBS 49 / 53 η (Mbit/J) 85 80 75 70 65 60 55 50 45 ρ=0.1 40 ρ=0.15 ρ=0.2 35 0 5 10 15 20 25 30 α EE increasing with licensed RBs (α). η (Mbit/J) 85 80 75 70 65 60 55 50 45 40 α=0 α=10 α=20 α=30 35 0.05 0.1 0.15 0.2 0.25 0.3 0.35 ρ More unlicensed RBs, and higher EE for small α. EE saturated for large α.

Performance for Multiple SBSs 50 / 53 η 4 (Mbit/J) 180 160 140 120 100 80 60 EE vs time slots η min 4 η max 4 η WT 4,φ 1 η WT 4,φ 2 η WT 4,φ 3 η NBS 4 α 4 60 50 40 30 20 Licensed RB vs time slots α min 4 α max 4 α WT 4,φ 1 α WT 4,φ 2 α WT 4,φ 3 α NBS 4 40 10 20 0.05 0.1 0.15 0.2 0.25 0.3 0.35 ρ 4 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 ρ 4 More licensed RBs (α), and higher EE for each SBS (η).

Conclusions Improving the EE of a SBS when the licensed RBs are not enough. Achieving EE balance and fairness among different SBS. 51 / 53

Outline 52 / 53 1 Background and Challenges 2 Traffic Offloading or Resource Sharing? 3 Mobile Data Offloading 4 Energy Efficiency Optimization 5 Summary

Summary 1 Performance comparison for traditional traffic offloading and resource sharing. Improve LTE performance, degrade WiFi performance. 2 Subversively, consider traffic offloading and resource sharing. Win-win strategy. 3 Improve EE in LAA systems. 53 / 53