Slide 2
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Overview
MotivationSimulation of operator use cases - examples⇒ System level simulation⇒ Cell outage compensation⇒ Home NB parameter optimization⇒ Interference coordination
Knowledge based reconfigurationDynamic spectrum management by RLConclusions
Slide 3
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Motivation
• Introduction and deployment of new wireless services and systemsshould be accelerated.
• Usability of future wireless access solutions should be improved (“plug&play”)
• High cost pressure requires improvement of operational efficiency
• Complexity and heterogeneity of radio access networks is dramatically increasing
Self-organization• Self-planning• Self-configuration• Self-optimization • Self-healing • Self-maintenance• …
LTE
UMTS
GSM
WLAN
SON / Self-x functionalities are mandatory for future radio networks!
Slide 4
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
System level simulation: Common evaluation framework
10 different simulation environments in E3-WP3 categorization of partners’ simulators according to functionalities studiedCommon assumptions and KPI’s for simulation
Class ANetwork & System
Level Simulator
Class BSystem Level
Simulator
Class CSystem Level
Simulator
Categorisation Protocol & Dynamic system behaviour
Dynamic system behaviour, Snap-shot based
Snap-shot/Quasi-static based
Examples of functionalities addressed
JRRM, self-x JRRM, DSM, self-x Self-x
Simulated time scale Up to several hours order of minutes Order of secondsBasic features Modelling of
protocols, dynamic user behaviour
Semi-dynamic change of user behaviour
Quasi-static positions of users
Cell/System configuration Multi-cell, multi-RATs Multi-cell, multi-RATs or single RAT
Multi-cell, single RAT
Main outputs • Network throughput
• Service outage
• System throughput• User TP (DL, UL)• Call/packet
blocking/ dropping rate
• Spectrum utilization
• System throughput
• User distribution
• SIR distribution
Slide 5
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
⇒ evaluation of self-x algorithms based on system-level simulations for a LTE mobile network.
⇒ Self-x use cases implemented: handover optimization, load balancing, cell outage compensation, and radio parameter optimization for home base stations.
Simulator example: LTE system simulator for self-x
-1 -0.5 0 0.5 1
-1
-0.5
0
0.5
1
12 3
45 6
78 9
1011 12
1314 1516
17 18
1920 21
2223 24
2526 27
2829 30
3132 33
3435 36
3738 39
4041 4243
44 454647 48
4950 51
5253 54 55
56 57
x in kmy
in k
m
Slide 6
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Cell outage compensation
Best possible compensation of coverage loss due to BS failure by COC mechanism Trade-off between⇒ Reassignments of lost UEs
to the network ⇒ Additional interference
introduced by compensating cells with changed parameters (e.g. increasedTX Power)
Cell outage situation with a random
set of UEs
Slide 7
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Idea of the creation of a compensation networkCompensation network consists of all cells which could participate on COC ⇒ Needs additional information out of daily network operation:
• Neighbour Cell List (NCL) of all cells that appear in a NCL of a cell
• E-UTRAN Cell Identifier (ECI) of all cells appearing in foreign NCLs
• UE IDs of all UEs assigned to foreign cells
Communication via X2 Interface (decentralized solution)⇒ Every cell takes the Neighbour Cell
List (NCL) of all neighbours, detected by its own NCL
⇒ Knowledge of all E-UTRAN CelI Identifier (ECI)s which are useful to compensate cell outage in principal
⇒ Every cell takes constantly the IDs (e.g IMSI) of all UEs that are assigned to their direct neighbours which appear in the NCL
⇒ Knowledge of all UE IDs of affected UEs, if cell outage occurs
⇒ Knowledge of all UEs which are reassigned tocells in the neighbourhood
Cell outage compensation
Simplified example for the information search mechanism
to create a compensation network
Slide 8
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Radio parameter optimization of Home Base Stations (HBS) in a co-channel situation⇒ Same RF carrier for HBS and
Macro Base Station (MBS) Trade-off between⇒ Additional resources provided
by each HBS⇒ Additional interference for MBS
and existing HBS introduced by each new HBS
Key benefits⇒ Improved indoor coverage ⇒ Additional capacity provided by
HBS⇒ Cost-efficient connections to the
core network via DSL
Home NB parameter optimization
Co-channel situation for the downlink direction
Useful link DL
Outdoor-Indoor Channel
Indoor Channel Building wall HeNB
Indoor-Outdoor-Indoor Channel
1st floor
2nd floorHeNB
HeNB
HeNB
Interferer DL
Indoor Channel
Indoor Channel
Slide 9
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Example for the distributions of mobile stations (MS) and home base stations (HBS)
Each BS – MS link can be characterized by a channel model.Multiple channel models based on WINNER II channel models are implemented.
The following rules are assumed for the selection of the channel model:⇒ If two HBS have a distance less than 20m, both HBS are within the same building (indoor channel). Otherwise the
femtocells are located in different buildings (indoor-outdoor-indoor channel).⇒ If an indoor MS@MBS have a distance to a HBS less than 6m, the MS is located in the same building like the HBS
(indoor channel). Otherwise the MS is located in another building (indoor-outdoor-indoor channel).It is possible that an indoor MS@MBS is located in a building with multiple HBS.
⇒ An outdoor MS@MBS is always outside of any building.⇒ An indoor MS@HBS is always inside the femtocell.
Home NB parameter optimization
-1.5 -1 -0.5 0 0.5 1 1.5-1.5
-1
-0.5
0
0.5
1
1.5
x in km
y in
km
0.35 0.4 0.45 0.50.22
0.24
0.26
0.28
0.3
0.32
0.34
0.36
0.38
0.4
0.42
x in km
y in
km
MBS
Indoor MS@MBS
HBS
MS@HBS
Slide 10
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Inter-cell Interference Co-ordination (ICIC) based on real network scenario
ICIC
ICIC: fixed reuse of resourcesfor cell-edge users
SON-ICIC: adaptationof cell-edge resourcesto traffic
Gain in SNIR
Slide 11
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Self-x selection of algorithms and first evaluation results
Knowledge based reconfiguration
0
2
4
6
8
10
12
14
16
18
20
Video Streaming Brow sing
Services
Sess
ions
per
cent
age
(%)
Have we addressed thecontext before?
Dynamic Sub-carrier Allocation (DSA) algorithm indicative results
Matching probability
MeanOptimization
Delay
512 Kbps
384 Kbps
128 Kbps
64 Kbps
Higher QoSlevels
assignment
Slide 12
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Dynamic spectrum management (DSM) to achieve an efficient utilisation of the scarce and valuable spectrum resources:⇒ Maximise spectrum reuse amongst users, cells, radio access
networks (RAN’s) and systems⇒ Ensuring that mutual interference between them remains at
acceptable levels at the same timeOptimization methodologies covered in E3:⇒ Machine learning,⇒ Genetic algorithms,⇒ Simulated annealing,⇒ Heuristics, etc.
DSM
Slide 13
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Multicell OFDMAFixed spectrumassignment strategies comparedwith DSA algorithmsHeuristic Algorithmsand Reinforcement Learning (RL)Algorithms are proposedDSA algorithms improve spectral efficiency, user’s QoSand spatial spectrum usage.First static results show that RL is suitable for DSA
18 27 36 45 54 63 72 810
50
100(a) Average Dissatisfaction
%
18 27 36 45 54 63 72 81
3.6
3.8
4
(b) Average Spectral Efficiency
Number of users in scenario
bits
/s/H
z/ce
ll
FRF1FRF3DSA-heurRL-DSA
FRF1FRF3DSA-heurRL-DSA
Dynamic spectrum mangement - example
Slide 14
E3
June 15, 2009 09062009_Simulation_of_self-x_algorithms.ppt
Increasing processing power and flexibility leads to enhanced radio resource management concepts
Optimized operability and optimized usage of resources of radio access technologies is in focus of future systems
Requirements and concepts for self-x were developed and are under investigation by means of simulation of algorithms
Derivation of recommendations for advanced RRM including performance analysis of different algorithms
Simulation of further operator use cases - HO parameter optimization & load balancing - : “Rule-based algorithms for self-x functionalities in radio access networks”, Rosenberger, M. et al., ICT Mobile Summit Santander
Conclusions and outlook