SON in 4G Mobile Networks

Size: px

Start display at page:

Download "SON in 4G Mobile Networks"

Edgar Warren
5 years ago
Views:

1 SON in 4G Mobile Networks Self-Optimization Techniques for Intelligent Base Stations Bell Labs Stuttgart Ulrich Barth 9. Fachtagung des ITG-FA 5.2, Oktober 2010

Self- organizing Radio Access Networks Motivation Current situation for radio access network management Deployment and maintenance become more and more complex and cost extensive Trend to smaller

2 Self- organizing Radio Access Networks Motivation Current situation for radio access network management Deployment and maintenance become more and more complex and cost extensive Trend to smaller cells, multi-band operation, heterogeneous mobile networks High manual intervention for configuration, capacity upgrade or in failure cases required High effort required for optimisation of system performance Deep system expertise required High effort necessary for measurement campaigns (drive tests) Different tools for planning, configuration, measurement/kpi acquisition and optimisation involved increasing effort for network management and optimisation new concepts for simplified network operation required All Rights Reserved Alcatel-Lucent 2008, XXXXX

3 Self- organizing Radio Access Networks Requirements Management of radio access networks has to be self-organized in future Automated configuration, optimization and fault management: towards real plug-and-play self-configuration continuous up to autonomous self-optimization fast self-healing mechanisms Paradigm change: to put network optimization know how into intelligent self-x algorithms to focus network management on high level monitoring and performance tuning High performance self-x algorithms required: fast convergence stable operation tuneable according to operator requirements managing mutual dependencies between self-x use cases All Rights Reserved Alcatel-Lucent 2008, XXXXX

4 Self-X Architecture Vision of fully distributed self-management NEM less network management Fully autonomous, distributed RAN optimisation Self-x functions in UE and enb Network Management NM OSS Itf-N X2-Itf measurements, UE location info alarms, status reports, KPIs distributed self-x algorithms Network management in NM OSS focussed on network planning alarm and performance monitoring high level performance tuning self-x enb high level network performance tuning LTE RAN enb self-x RAN selfoptimization OSS: Operation Support System NEM: Network Element Manager performance monitoring KPIs alarms self-x enb All Rights Reserved Alcatel-Lucent 2008, XXXXX

6 Configuration Parameters for Handover in LTE Filtered RSRP [db] Source Cell Target Cell source cell UE target cell Radio Link Failure (RLF) threshold LTE handover performance has large impact on system performance Configuration parameters Filtered RSRP values Handover Margin, i.e. hysteresis between source and target Time to trigger (TTT) Cell Individual Offset (CIO), add on handover margin Target Hyst(dB) High handover success rate TTT(ms) Handover event from UE to enb X2 delay Handover command from enb to UE Handover failure due to RLF Time A3 HO event HO command Normalized HO Rate Normalized HO Rate Vs Residual BLER for ; TTT=0 to 200 ms; 20ms step BLER [%] BLER: block error rate of HO command RLF Normalized HO rate: without slow/fast fading SON algo: find best trade-off between minimum BLER and minimum HO rate (ping pong) All Rights Reserved Alcatel-Lucent

7 SON definition: too late Handover cell A UE cell B RLF, as UE still associated to cell A Filtered RSRP [db] Source Cell A Target Cell B Filtered RSRP [db] Source Cell A Target Cell B Radio Link Failure (RLF) threshold Hyst(dB) TTT(ms) Handover failure due to RLF no uplink communication: no Handover event from UE to enb A Time Radio Link Failure (RLF) threshold Hyst(dB) TTT(ms) Handover event from UE to enb X2 delay Handover command from enb to UE Handover failure due to RLF Time Alternative 1: a failure occurs in the source cell before the HO was initiated Alternative 2: a failure occurs in the source cell during the HO procedure All Rights Reserved Alcatel-Lucent

8 Messages and algorithm for too late handover UE enb A enb B UE connected to enb A Measurement Report (enb B) RRC Connection Reconfiguration HO Request HO Request Acknowledge RLF RRC Connection Reestablishment Request including PCI#A RRC Connection Reestablishment Update Reestablishment Statistics (PCI#A) RRCConnectionReestablishmentComplete RLF INDICATION UE Context Release Update Reestablishment Statistics (PCI#A) SON Algorithm - sufficient measurements for decision of HO problem? - analysis of measurements and other data (own enb and other enb) - modify HO parameters - for all UE speed classes - mobility state dependent - start new measurement cycle All Rights Reserved Alcatel-Lucent

9 SON based HO optimization Characteristics of simulation scenario Frankfurt Available ray-tracing data Input data characteristics: 16 real world antenna locations in the city of Frankfurt Tri-sectorized configuration (i.e. 48 cells) individual antenna type information, heights and beam directions topographic map of region (4000m 4000m) Output data characteristics: Integer type pathloss data per 10m 10m grid point per antenna Range: -88dB to -214dB All Rights Reserved Alcatel-Lucent

SON based HO optimization Algorithm test in simulation Event based simulation of HO events, one algorithm instance per simulated cell Interference level: 50% load mobile speed: 3, 30, 120 km/h

10 SON based HO optimization Algorithm test in simulation Event based simulation of HO events, one algorithm instance per simulated cell Interference level: 50% load mobile speed: 3, 30, 120 km/h 16 mobiles active per speed moving along given roads at crossroads, mobiles choose randomly next road same initial mobility parameters at simulation start All Rights Reserved Alcatel-Lucent

11 SON based HO optimization HO success rate window controller Simulation results (Simulated network operation: 3 days = s) HO Success Rate [%] Gained Key Performance Indicators for TRX #31 100, , , , , , ,0 HO Success Rate 12 Mean Resting Time 30,0 9 20,0 6 10,0 3 0, Simulated Time [s] Mean Mobile Resting Time [s] A quick convergence of the cell global parameters can be observed for frequently visited cells (i.e. on many HO events) Choosing initial HO parameters above optimal settings causes a HO success rate below given limit The algorithm instance tunes the parameters towards earlier decision for HO, achieving an improvement of the cell total HO success rate Mobility Parameters for TRX #31 Hysteresis [db] 2,50 2,25 2,00 1,75 1,50 1,25 1,00 0,75 0,50 0,25 Hysteresis Time-To-Trigger 2,0 1,8 1,6 1,4 1,2 1,0 0,8 0,6 0,4 0,2 Time-To-Trigger [s] 0,00 0, Simulated Time [s] All Rights Reserved Alcatel-Lucent

13 Antenna Tilt Optimization Optimization goals: sector coverage sector capacity based on downlink performance metric bits/sec/hz Approach: Distributed optimization: optimization of single cells together with closest neighbors targeting global optimum by adjusting the antenna tilt for each sector individually distributed approach, co-operating enb and neighbours All Rights Reserved Alcatel-Lucent

correlated shadowing is invisible in best server plots due to the equal attenuation of all sectors

14 Antenna Tilt Optimization Best Serving Sectors, Displaced Site Locations after optimization sites are displaced but playground borders are kept fixed, with wrap around slow fading with area correlated shadowing is invisible in best server plots due to the equal attenuation of all sectors at a certain point, independent of the direction of the signal All Rights Reserved Alcatel-Lucent

15 Antenna Tilt Optimization Metrics Cell wide optimization approach target: coverage target: capacity M: performance metric G: Geometry B: bandwidth p(g): probability T tf : throughput per MCS W: weighting factor coverage metric: M 1 = B T tf ( G5 percentile ) capacity metric: G M 2 = B p( G) T G max min tf ( G) dg next charts: concrete simulation studies: operator tunable weighting parameter W = 0.91 weighted coverage/capacity metric: M = W B Ttf ( G5 percentile) + (1 W ) B p( G) Ttf ( G) dg G G max min (1) All Rights Reserved Alcatel-Lucent

16 Antenna Tilt Optimization Performance Gains of Sector Average Performance and Sector Edge Perf. Optimized Performance (red) [%] Reference Performance 15 (blue) [100%] Trade-off given by Utility Metric (green): - weight factor W between coverage and capacity defines the slope of this line Results: significant gains with respect to equal tilts +44% capacity performance +7% coverage performance +44% +7% gain of utility +21% of All Rights Reserved Alcatel-Lucent

17 Antenna Tilt Optimization Convergence Speed and Gains of Performance Metric with optimized tilts Results: promising, playground wide improvement non-oscillating fast convergence two reference curves for all sectors with 14 and 15 degrees equal downtilt one cyle through all sectors equals 57 steps All Rights Reserved Alcatel-Lucent

19 Mutual interactions of SON optimization algorithms Radio System enb Radio System KPI/ measurement 1 SON use case a target function metric a optimization algorithm a Parameter I Parameter II KPI/ measurement 2 SON use case b KPI/ measurement 3 target function metric b optimization algorithm b Parameter III Mutual impact on optimization target: One metric is influenced by parameters of different SON algorithms Interference Coordination <-> Load balancing Coupling by same control parameter: One parameter is modified by different SON algorithms Handover <-> Load balancing different coupling mechanisms, different coupling strength solution required to manage SON use case interworking! All Rights Reserved Alcatel-Lucent

Fault compensation algorithm based on handover margins in LTE networks

de-la-bandera et al. EURASIP Journal on Wireless Communications and Networking (2016) 2016:246 DOI 10.1186/s13638-016-0742-x RESEARCH Fault compensation algorithm based on handover margins in LTE networks