EDGE-PRIORITIZED CHANNEL- AND TRAFFIC-AWARE UPLINK CARRIER AGGREGATION IN LTE-ADVANCED SYSTEMS AUTHORS R. SIVARAJ, A. PANDE, K. ZENG, K. GOVINDAN, P. MOHAPATRA PRESENTER R. SIVARAJ, Ph.D student in CS, UC DAVIS, CA, USA Email: rsivaraj AT ucdavis DOT edu http://spirit.cs.ucdavis.edu 1 AGENDA OF THE PRESENTATION INTRODUCTION ISSUES IN EXISTING LITERATURE PROBLEM STATEMENT KEY CONTRIBUTIONS WHY UPLINK? SYSTEM MODEL CARRIER AGGREGATION SCHEDULING PERFORMANCE EVALUATION DISCUSSION CONCLUSION 2 INTRODUCTION TO LONG TERM EVOLUTION 90 1.4 MHz CC 2600 MHz 5 MHz CC 2025 MHz 20 MHz CC 1900 MHz 15 MHz CC 1525 MHz 10 MHz CC 800 MHz 3 INTRODUCTION TO LTE-ADVANCED LTE RELEASE 10 (4G CELLULAR NETWORK – EVOLVED FROM OFDMA LTE) PROVISIONING NEXT-GEN TELECOMMUNICATION SERVICES PEAK DATA RATES: DOWNLINK (Low Mobility): 1 Gbps UPLINK (Low Mobility): 500 Mbps DOWNLINK (High Mobility): 100 Mbps PEAK BANDWIDTH: 100 MHz AGGREGATION OF UPTO 5 COMPONENT CARRIERS SCALABLE LTE BANDWIDTHS RANGING FROM 1.4 MHz to 20 MHz PHYSICAL RESOURCE BLOCKS (PRB) CONSTITUENT OF 12 SUB-CARRIERS (EACH 18 kHz) – MINIMUM RESOURCE ALLOCATION UNIT FOR ANY UE DOWNLINK: CONTIGUOUS/DIS-CONTIGUOUS SUB-CARRIERS FOR A SINGLE PRB (CHANNEL) UPLINK: CONTIGUOUS SUB-CARRIERS FOR A SINGLE PRB (RECOMMENDED) 4 CARRIER AGGREGATION SESSION ADMISSION CONTROL L3 – CC ASSIGNMENT L2 – PACKET SCHEDULING L2 – PACKET SCHEDULING L2 – PACKET SCHEDULING LINK ADAPTATION LINK ADAPTATION LINK ADAPTATION HARQ HARQ HARQ MAC MAC MAC CC CC CC 5 ISSUES IN EXISTING LITERATURE Am in here for a video conferencing Same number of resources for both ? Send an email to my boss applying for leave Now I wanna surf Facebook Video gaming with my friends for the next half-hour Ah !!!! Jus wanna surf Facebook I wanna stream a HD video A VoIP call to my manager Center/Close to center UEs contribute to less UL traffic Edge UEs contribute 6 UE GROUPING Grouping of UEs based on spatial correlation – Similar channel conditions and radio characteristics CQI CQI CQI CQI CQI CQI 7 KEY CONTRIBUTIONS OF THE PAPER PROBLEM STATEMENT: How to optimally provision next-generation telecommunication services in LTE-Advanced uplink ? APPROACH: Prioritization of cell-edge UE groups for channel- and Traffic-aware Carrier Aggregation Effectively accommodates log-normal shadowing, channel fading and propagation losses which adversely impact edge throughput Efficient representation of under-represented weak terminals Profile-based Proportional Fair Packet Scheduling Resolves contention of resources using inter- and intra-group scheduling on a time-domain and frequency-domain basis 8 MOTIVATION - WHY UPLINK ? HIGHER UPLINK TRAFFIC WITH THE EVOLUTION OF WEB 2.0 UPLINK COULD POSSIBLY EXCEED DOWNLINK IN 2020 (CISCO ESTIMATES) MOBILE TERMINALS – MORE POWER-LIMITED THAN THE eNodeB RADIO CHANNEL CHARACTERISTICS (LIKE PATH LOSS) INFLUENCE UE TRANSMISSION POWER FREQUENCY DIVERSITY AND AMC : NOT EFFECTIVELY-UTILIZED IN THE UPLINK MOTIVATION FOR SUBSCRIBERS TO CONTRIBUTE TO UPLINK TRAFFIC FROM THEIR HAND-HELD DEVICES 9 MOTIVATION - WHY CHANNELAWARE AND EDGE ? 41% INCREASE IN THROUGHPUT OVER CHANNEL-BLIND ASSIGNMENT 57% INCREASE IN EDGE UE THROUGHPUT PATH LOSS-BASED ASSIGNMENT FOR CHANNEL AWARENESS 10 SYSTEM MODEL QCI CALCULATE AMBR QCI NON-ADJACENT INTER-BAND CARRIER AGGREGATION (800 MHz, 1525 MHz, 1800 MHz, 2025 MHz, 2600 MHz) FOR ANY UE, ASSIGNABLE CC SET CONTAINS CCs WHOSE PATH LOSS IS LESS THAN A PRE-DEFINED THRESHOLD ASSIGN RESOURCES ONLY FROM ASSIGNABLE CCs TO UE GROUPS UEs FROM EACH GROUP SEND QCI TO eNB (TRAFFIC SUBSCRIPTION) DETERMINATION OF AMBR REQUIREMENTS FOR EACH GROUP EQUAL POWER ALLOCATION ON ALL FREQUENCY BANDS WITH SINR, CQI AND SPECTRAL EFFICIENCY COMPUTATIONS: 11 DISTRIBUTION OF ASSIGNABLE CCs TO UEs CC1 UE1 CC2 UE2 CC3 UE3 CC4 UE4 CC5 UE5 12 N CARRIER AGGREGATION – THE OBJECTIVE AGGREGATING THE CCs AND ASSIGNING THEIR PRBs TO THE GROUPS Theoretical Formulation: NP-Hard Generalized Assignment Problem: CC1 CC2 CCi βij G1 W1 G2 W2 Gj Wj … CCn ITEMS … … Gm Wm BINS pij SOLUTION : SUBSET OF ITEMS (AGGREGATED CARRIER U) TO BE ASSIGNED TO THE BINS FEASIBLE SOLUTION: SOLUTION WITH MAXIMUM PROFIT (ACHIEVED UPLINK THROUGHPUT) 13 HEURISTICS PRIORITIZING THE SPATIAL GROUPS : = SET OF ASSIGNABLE CCs FOR UE r IN GROUP Gi i:= SET OF ASSIGNABLE CCs FOR GROUP Gi PRIORITY METRIC := 800 G1 1525 G2 1800 Gj 2025 … 2600 Gm MGi LEAST PRIORITIZATION OF CELL-CENTER UEs IN RESOURCE ALLOCATION – COULD STILL GUARANTEE ALLOCATION OF GOOD CCs 14 PROOF OF CORRECTNESS – THE INTUITION ASSIGNABLE RESOURCES FOR GROUP Gi : A B C ASSIGNABLE RESOURCES FOR GROUP Gj : A B C D E ASSUME Gj GETS A HIGHER PRIORITY THAN Gi (by contradiction): A,B,C COULD BE ASSIGNED TO Gj (worst case traffic requirement) ASSIGNABLE RESOURCES FOR Gi – EXHAUSTED (SHOULD BE SCHEDULED IN THE NEXT TIME SLOT) IF Gi GETS A HIGHER PRIORITY THAN Gj A,B,C COULD BE ASSIGNED TO Gi (worst case traffic requirement) D,E COULD STILL BE ASSIGNED TO Gj HIGHER ADVERSE IMPACT FOR THE FORMER CASE – NOT A NEARLY OPTIMAL SOLUTION 15 PRB ASSIGNMENT GOAL To allocate the best set of contiguous channels to the UE groups To minimize resource contention and dependency on scheduling Already assigned resource/CC – considered for re-assignment to another group only while lack of choice for the other group FORMULATION: Sum of estimated bandwidths of UEs of group Gi yij := Fraction of the total number of PRBs in CC j allocated to Gi βij := Available bandwidth in CC j for group Gi 16 PRB ASSIGNMENT Traffic requirement for any group Gi : Estimate the SINR, CQI and MCS – Spectral Efficiency values for all the PRBs across each assignable CC for a given UE transmission power and path loss model Spectral Efficiency is given by: Channel allocation follows Maximum Throughput algorithm using the computed MCS levels 17 SCHEDULING TO RESOLVE CONTENTION AMONGST THE UE GROUPS AND INDIVIDUAL UEs PHASE I : TIME DOMAIN-BASED INTER-GROUP PFPS COMMON RESOURCES t2 G1 t1 PROFILE-BASED TD METRIC: SERVICE PRIORITIZATION G2 18 SCHEDULING INTRA-GROUP FREQUENCY-DOMAIN PFPS: UE1 UE3 UE with maximum FD metric : Total number of PRB combinations: UE2 19 SIMULATION DETAILS 1. 2. 3. NS3 LENA – LTE/EPC NETWORK SIMULATOR FULLY-IMPLEMENTED LTE UPLINK PHY AND MAC FUNCTIONALITIES FEATURES INCLUDE MODELING THE AMC, PATH LOSS MEASUREMENTS, CHANNEL-STATE INFORMATION FEEDBACK CELL SIZE 1 km NON-ADJACENT FREQUENCY BANDS = 10 (5 CCs CHOSEN FOR CA) FREQUENCY BANDS : 800, 1525, 1800, 2025, 2600 MHz CONSTANT POSITION MOBILITY MODEL FOR eNB, CONSTANT VELOCITY MOBILITY MODEL FOR Ues UNIFORM UE DISTRIBUTION ACROSS THE CELL MAXIMUM 10 UEs PER CELL, (MAX. 5 HIGH-END TRAFFIC APPLICATIONS PER UE), GBR TRAFFIC APPLICATIONS JAKES FADING MODEL, LOG NORMAL SHADOWING, 23 dBM UE TRANSMISSION POWER, 43 dBM eNB TRANSMISSION POWER, -120 dBM THRESHOLD PATH LOSS, -174 dBM/Hz NOISE SPECTRAL DENSITY 20 TRAFFIC DETAILS 21 RESULTS CC ASSIGNMENT (IMPR 33%, 15%) INTER- AND INTRA-GROUP PFPS (IMPR 15%, 21%) 22 RESULTS CC ASSIGNMENT IN UNIFORM SCENARIO CC ASSIGNMENT (EDGE UEs) (IMPR. 64%, 54%) 23 RESULTS INTER- AND INTRA-GROUP PFPS (EDGE UEs IMPR. 62%) UNIFORM CC ASSIGNMENT (EDGE UEs IMPR. 10%) 24 RESULTS CDF OF LTE-A UPLINK THROUGHPUT (IMPR. 20%, SD 14%) ACHIEVED GBR (91.7%, 87.4%) 25 DISCUSSIONLUSIONS IMPACT ON POWER OPTIMIZATION : ESTIMATED TRANSMISSION POWER FOR UE ON ANY CC: LOW VALUES OF M AND PL FROM OUR PROPOSED MECHANISMS – MINIMIZES POWER CONSUMPTION REST OF THE PARAMETERS ARE CC-SPECIFIC OPEN-LOOP AND CLOSED-LOOP VALUES NOT APPLICABLE TO HIGH-SPEED MOBILE UEs – NO DISTINCT CELLCENTER AND CELL-EDGE UEs, IRRELEVANT FEEDBACK REQUIRES COORDINATION AMIDST MULTIPLE, NEIGHBORING eNBs 26 CONCLUSIONS CHANNEL- AND TRAFFIC-AWARENESS IN RESOURCE ALLOCATION FOR LTE-ADVANCED SYSTEMS FOCUS ON UPLINK CARRIER AGGREGATION – CC ASSIGNMENT AND PFPS MOTIVATION FOR GROUPING OF UEs CC ASSIGNMENT – PROBLEM THEORETICALLY-MODELLED AS NP-HARD GENERALIZED ASSIGNMENT PROBLEM PROPOSAL OF EDGE-PRIORITIZED CC ASSIGNMENT PROPOSAL OF TIME-DOMAIN INTER-GROUP AND FREQUENCY-DOMAIN INTRA-GROUP PFPS PERFORMANCE EVALUATION – 33% IMPR IN CC ASSIGNMENT AND 15% IMPROVEMENT IN SCHEDULING MECHANISMS DOWNLINK LTE-ADVANCED MULTICAST, HIGH-SPEED SCENARIO – ENVISIONED FOR FUTURE WORK 27 QUERIES?? 28 THANK YOU 29