ABSF Offsetting and Optimal Resource Partitioning for eicic in LTE-Advanced: Proposal and Analysis using a Nash Bargaining Approach

Transcription

1 ABSF Offsetting and Optimal Resource Partitioning for eicic in LTE-Advanced: Proposal and Analysis using a Nash Bargaining Approach Mahmoud Ibrahiem Kamel Research Assistant (4G++ project) Siemens Certified Mobile Technology Instructor mkamel@4gpp-project.net Prof. Dr. Khaled M. F. Elsayed Department of Electronics and Communications Engineering Faculty of Engineering, Cairo University khaled@ieee.org 7/16/2013 1

2 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/2013 2

3 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/2013 3

4 HetNet, eicic, and ABSF An Introduction 7/16/2013 4

5 HetNet Environments X2 X2 7/16/2013 5

6 eicic Framework Power Control Time Domain Frequency Domain eicic 7/16/2013 6

7 eicic Framework Lightly- Loaded Subframe Symbol Muting ABSF Almost Blank Subframe Time Domain Consecutive Subframe Blanking 7/16/2013 7

8 ABSF Technique Almost Blank SubFrame Figure from [3GPP TSG RAN WG1, R ,] Comparison of Time-domain eicic Solutions 7/16/2013 8

9 ABSF Technique Almost Blank SubFrame Figure from [3GPP TSG RAN WG1, R ,] Performance of eicic with Control Channel Coverage Limitation 7/16/2013 9

10 ABSF Patterns (FDD mode) 1/8 Pattern 2/8 Pattern 3/8 Pattern 3/20 Pattern 7/16/

11 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/

12 Our Proposed ABSF Framework A Background 7/16/

13 The Proposed ABSF Framework VMUEs Tracking Prioritized Scheduling Aggressor HeNBs Identification ABSF Framework SINR Tracking during ABSF (Kalman Filter) HetNet Nodes Coordination 7/16/

14 The Proposed ABSF Framework The framework is first introduced in the paper: Mahmoud Kamel and Khaled Elsayed, "Performance Evaluation of a Coordinated Time- Domain eicic Framework based on ABSF in Heterogeneous LTE- Advanced Networks," Proceedings of IEEE Globecom 2012, Anaheim, CA, USA. The framework provides: A tracking algorithm to identify the victim macro UE s An algorithm to identify the aggressor Home enbs A coordination scheme between the HetNet nodes An algorithm based on Kalman filter to track the SINR of the victim MUEs during the ABSF A modification to the scheduling scheme to prioritize scheduling the resources to victim MUEs during the ABSF 7/16/

15 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/

16 ABSF Offsetting 7/16/

17 ABSF Offsetting Concept ABSF Offsetting A novel approach to reduce the femto-cells throughput degradation due to ABSF. The ABSF blanking normally starts at subframe 0 The ABSF offsetting means that ABSF blanking may start at an offset, i.e. at a subframe other than subframe 0. Offset 0, 1, or 2 means the blanking starts at subframe 0, 1 or 2 respectively 7/16/

18 ABSF Offsetting Concept 7/16/

19 ABSF Offsetting Procedure ABSF Pattern Selection ABSF Pattern Reduction ABSF Offset Assignment 7/16/

20 ABSF Offsetting A Motivating Example 7/16/

21 ABSF Offsetting A Motivating Example HIA-1 HIA-1 HIA-2 HIA-2 HIA-3 HIA-3 HIA-4 HIA-4 (a) No Offsetting (b) ABSF Offsetting (Round Robin) 7/16/

22 ABSF Offsetting A Motivating Example The question now is whether this improvement in the aggressor HeNBs throughput has an impact on the throughput of the macro-cell. The delightfully surprising answer is NO. The ABSF offsetting magically achieves the same effective blanking rate of 3/8 for the macro cell victim users which are not aware or do not even care how the offsetting is operated in HeNB s 7/16/

23 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/

24 The Optimal ABSF Pattern Selection A Nash Bargaining Approach 7/16/

25 Nash Bargaining The Problem Nash 1953: One states as axioms several properties that would seem natural for the solution to have and then one discovers that axioms actually determine the solution uniquely. Player 1 Player 2 7/16/

26 Nash Bargaining - Axioms PAR The solution is Parito Optimal SYM The solution is Symmetric TRA The solution is Invariant under Linear Transforms IIA The solution is Independent of Irrelevant Alternatives 7/16/

27 Nash Bargaining The Solution 2 Players Solution NNN S, d = max(v1 d1)(v2 d2) v1,v2 s. t v1, v2 S N Players Solution NNN S, d = max v1,v2,,vv N (vv dd) i=1 s. t v1, v2,, vv S 7/16/

28 Nash Bargaining Proportional Fairness NNN S, d = max v1,v2,,vv N (vv dd) i=1 Consider the special case when : dd = 0 i It s proven that : NNN = max v1,v2,,vv N vv i=1 PPP = max v1,v2,,vv N log vv i=1 7/16/

29 Pattern Selection Problem Formulation φ S, d = max U MMMM = U S,U d max U S,U d U n d n. U v d v N n U n = 1 α. G N n i=1 φ S, d = max U S N n log U n + log U v N v. r i U v = α. G(N v) j=1 N v. r j N n max U S N n log 1 α + log i=1 G N n N n. r i + N v N v log α + log G N v j=1 N v. r j 7/16/

30 ABSF Pattern Selection Results Modeling the problem as a Nash bargaining problem, we can find that the value of α that maximizes the aggregate macro-cell utility is: α = N v N n + N v The ABSF pattern is selected via I FFF = mmm (mmm α. 8, 3, 1) I TTT = mmm (mmm α. 10, 2, 1) 7/16/

31 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/

32 The Reduced ABSF Patterns and Offset Selection 7/16/

33 Reduced ABSF Patterns HIA-1 HIA-1 HIA-2 HIA-2 HIA-3 HIA-3 HIA-4 HIA-4 (a) No Offsetting (b) ABSF Offsetting (Round Robin) 7/16/

34 Pattern Reduction Problem Formulation H max N h log β h + log G N h U S h=1 N h k=1 N h. r k β h = 1 β 1 β 2 β h 1 β i = 0 N i β i N h 1 β 1 β 2 β h 1 = 0 i 7/16/

35 Reduced ABSF Patterns and Offsetting Results Modeling the problem as a second stage of Nash bargaining, we can find that the value of β 1, β 2,, β h that maximizes the aggregate victim MUEs utility U v is: β i = N i N v The ABSF pattern is selected via I i = mmm β i. I FFF, 1 i I i = mmm β i. I TTT, 1 i 7/16/

36 Reduced ABSF Patterns and Offsets HIA-1 HIA-2 HIA-3 HIA-4 (c) Balanced ABSF Offsetting 7/16/

37 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/

38 Performance Evaluation 7/16/

39 System Scenario Parameter Value System HetNet in LTE-Advanced, deployment cantered at 2000 MHz, with a carrier bandwidth = 5 MHz Macro-cell Cell with a radius of 500 m, and ~0.65 km 2 cell area, the enb is centred in the cell area with omni antenna 40 CSG femto-cells in 40 buildings uniformly distributed in the Femto-cells cell area with a density of ~60 femto-cell/km 2. Each femto-cell is serving one HUE and fully loaded, and therefor utilize all subcarriers. Frequency Reuse 3 Duplex Technique Scheduling Scheme Macro-cell Subscribers Femto-cell Subscribers FDD Modified Proportional Fair 12 MUEs uniformly distributed in the cell area with an application has a varying rate from 100 kbps to 650 kbps for each MUE. 6 of the MUEs are VMUEs and they are positioned such that each 2 VMUEs are located in the same HIA 1 stationary HUE for each femto-cell, with an application of a rate 17.6 Mbps. 7/16/

40 Throughput of the Macro-cell 7/16/

41 Aggregate Throughput of the VMUEs 7/16/

42 Aggregate Throughput of the Femto-Cells 7/16/

43 Fairness Index of the Macro-cell UEs 7/16/

44 Agenda HetNet, eicic, and ABSF - An Introduction Our Proposed ABSF Framework - Background ABSF Offsetting The Optimal ABSF Pattern Selection A Nash Bargaining Approach The Reduced ABSF Patterns and Offset Selection Performance Evaluation Conclusions and Future Trends 7/16/

45 Conclusion A novel algorithm called ABSF offsetting is proposed In ABSF offsetting the VMUEs are grouped in HIAs, and the aggressor HeNBs in each HIA are assigned a reduced blanking rate with an offset. The offset is the subframe to start blanking at. The performance evaluation results show that the ABSF offsetting preserves the performance of the macro-cell as in the equivalent ABSF pattern, however it significantly improves the performance of ABSF-mode triggered HeNBs. 7/16/

46 Future Trends Formulation of Silence-Whisper scheme as a mix between ABSF and power control schemes. Applying the proposed schemes in a manner that is fully SON (Self Organizing Network) compliant 7/16/

47 Thanks? 7/16/