classification of cells based on mobile network context...
TRANSCRIPT
1 © Nokia Solutions and Networks 2015
Classification of Cells Based on Mobile Network Context Information Simon Lohmüller
University of Augsburg, Nokia
Sören Hahn, Thomas Kürner
Technical University of Braunschweig
Dario Götz, Andreas Eisenblätter
atesio GmbH
Lars Christoph Schmelz
Nokia
2 © Nokia Solutions and Networks 2015
Motivation SON Function Configuration
SON Function behaviour (impact on KPI values) can be influenced through
SON Function Configuration Parameters (SCPs)
by adjusting
SCP Values (SCVs)
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
SON Function
Network Configuration Parameters
Measure-ments / KPIs
SCP SCP SCP SCP
3 © Nokia Solutions and Networks 2015
Motivation SON Management
I want the network to….
Manual SON Function Configuration
I want the network to….
Technical Objectives
Overcome Manual Gap
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
4 © Nokia Solutions and Networks 2015
Basics SON Objective Manager
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
SON Objective Manager
SCV Set A SCV Set B
SON Function B
Operator Domain
Objective Model
Context Model
Manufacturer Domain
Manufacturer A Manufacturer B
SON Function Model A
SON Function Model B
SON Function A
5 © Nokia Solutions and Networks 2015
Problem Description KPI Target Definition in the Mobile Network
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
• Different KPI targets for different areas in the network
• KPI targets may change over time KPI targets C
KPI targets A KPI targets B
6 © Nokia Solutions and Networks 2015
Problem Description Goal for SON Objective Manager
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
Goal
• Find suitable SCV Sets… • for the SON Functions implemented at each cell • for every condition the cell may be in
Problem: Impossible to select suitable SCV Sets for each individual cell manually
SON Objective Manager Mapping
Conditions CellsX
SCV Sets+
7 © Nokia Solutions and Networks 2015
Problem Description Context – Context Space
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
Problem: n-dimensional context space
with possibly infinite context attributes
Context
• All possible context combinations that may exist
• One dimension for each context parameter
Context Space
• Abstract description of a cell‘s properties and capabilities as well as the environment and situation it operates in
• Cell Type ∈ {Pico, Micro, Macro} • Cell Technology ∈ {LTE-1800,
LTE-2600, UMTS-2100, GSM-900} • … C
ell
Ty
pe
Mic
roM
acro
Pic
o
Available Technology
8 © Nokia Solutions and Networks 2015
Concept Introduction of Context Attributes
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
First Reduction
• Introduction of context attributes • SCV Set selection based on description of cell‘s context
Assumption: Cells in the same context (i.e., operating in the same situation and environment) can be handled in a similar way
SON Objective Manager Mapping
ContextsConditions CellsX
SCV Sets+
9 © Nokia Solutions and Networks 2015
Concept Introduction of Objectives
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
Objectives
• Depend on the cell‘s context • Formulated by the network operator
Problem: Impossible to define objectives for each individual cell context manually
Assumption: Cells in equal context have equal objectives
SON Objective Manager Mapping
Contexts
Objectives
Conditions CellsX
SCV Sets+
10 © Nokia Solutions and Networks 2015
• all possible context combinations that may exist
• one dimension for each context parameter
Concept Context Space – Context Classes
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
• Combination of context attributes • Each cell class represents certain cells in the
network
Context Space Context Classes
Problem: n-dimensional context space with possibly infinite context attributes
Solution: Partitioning of context into context classes
Ce
ll T
yp
e
Mic
roM
acro
Pic
o
Available TechnologyAvailable Technology
Ce
ll T
yp
e
Mic
roM
acro
Pic
o
A
C
B
11 © Nokia Solutions and Networks 2015
Concept Reduction to Cell Classes
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
Classes
• Reduce the amount of objectives one objective per cell class • Reduce the complexity of the context space partitioning into cell classes
SON Objective Manager Mapping
Contexts
ObjectivesClasses
Conditions CellsX
SCV Sets+
12 © Nokia Solutions and Networks 2015
Concept SON Function Model Mapping
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
SON Function Model (SFM)
• Predicts the expected network behaviour in terms of KPIs for a specific SCV Set
Assumption: Behaviour depends on cell context and the environment context dependent effects in the SFM
SON Function Model Mapping
BehavioursClasses SCV SetsX
Contexts SCV SetsX
13 © Nokia Solutions and Networks 2015
Concept Combined Transformation Process
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
SON Objective Manager
• Combines both mapping processes in order to reduce complexity
• Determines the appropriate objective for a cell under a given condition based on cell class definition
• Behaviour prediction in the SFM enables selection of SCV Sets that are in line with the given objectives
SON Function Model Mapping
BehavioursClasses SCV SetsX
SON Objective Manager Mapping
Contexts
ObjectivesClasses
Conditions CellsX
SCV Sets+
14 © Nokia Solutions and Networks 2015
Implementation Context Attribute Identification Techniques
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
Expert Knowledge
• Basic set of context attributes can be provided manually by the operator
Problems • Hard to classify thousands of cells in the network • Cell‘s context may change over time
Automation
• Determine context attributes of a cell with regards to the type of land it covers • E.g., urban vs. rural, high-speed mobility vs. normal mobility
• Use so-called „land use maps“ (or „clutter maps“) and „pixel maps“ Example
• Large parts of cell‘s footprint consists of the land use classes „low-density area“
and „forest“ Cell will be classified as „rural“
15 © Nokia Solutions and Networks 2015
Implementation Detection of Faults in the Assignment
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
• Introducing an automated mechanism raises questions about • How can results be verified? • How may faults be detected?
Problem
• Fault detection by analysing the similarity of the behaviour of cells belonging to the same context class • Statistical outlier detection • Classification methods
Solution
16 © Nokia Solutions and Networks 2015
Conclusion and Future Work
S. Hahn, D. Götz, S. Lohmüller, L.C. Schmelz, A. Eisenblätter, T. Kürner
• A mechanism to classify cells based on network context information has been
introduced complexity in the management of the network can be significantly reduced
• Applications for Context and Classes in the management of a SON have been introduced
• Methods to classify cells and detect incorrectly classified cells have been explained
Conclusion
• Apply self-learning techniques (e.g., to deal with wrong cell class assignment) • Ultimate goal: Facilitate the adjustment of cells and the SON Function running on
that cell individually so that they best fulfil given operator objectives
Future Work