LAC (Location JArea Code) Optimisation

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 28

LAC

optimisation
Ned (Mohaned) HASSOUN
Master in Intelligent Systems

Supervised by Sharon Wood (Uni. of Sussex)


and Michael Ratford (Motorola)
Introduction

Problem Definition
Assimilation of this problem with an existing
one
Definition of Graph partitioning problem
2 important ways of Partitioning
Presentation of Multilevel approach
Graph partitioning algorithms
Results
Basic Components: Cell
A cell provides radio signal to a mobile station in a
certain area
Different sizes: up to 50 km

Theoritical view Real-world situation


Network Hierarchy
Hierarchical structure: Public Land Mobile Network (PLMN) to a cell.
PLMN controls many Mobile Switching
Centres (MSC)
Mobile Switching Centre:
computer that places the calls, and PLMN
takes and receives data from the
subscriber or from PSTN MSC
MSC controls 1 to many Location
Area (LA) Location Location
Area Area
Location Area
Cells are grouped into location Area.
Identified by Internationally unique
code composed of several elements.
Among them, Location Area Code
assigned by the operator
Each cell stores the appropriate
identifier.
The assignment is done by operator.
Hand-over
Handover between cells:
When a mobile station moves from one cell to another, the signal has
to be passed to the station of the new cell. This is known as handover.

Mobiles Location updated in the MSC register to be able to forward any


call
Handover between MSC:
Most general form of handover
Call forwarded to the new MSC
MSCB

Location Area 1
Location Area 2
MSCA
Two Major Problems
Ping-pong handover: mobile station crosses very frequently the MSC
boundaries.
Consequence: For incoming calls, the last reported MSC may no longer be valid,
the call will fail.

Location Area 1 Location Area 2

Paging problem:
Too many mobiles crossing boundaries provokes overload of the connections
between MSCs.
Objectives
Upgrade the expensive material (Not really a good idea)
OR
Rearrange cells into LAs, such that:
It minimises the number of mobiles near the MSC boundaries,
i.e. minimise handovers.
Constraint: MSCs can deal with limited number of mobiles. Thus,
LAs have limited number of cells.
Maximise this limitation to reduce number of LAs.

Data available:
- Concentration of mobiles in a cell
- Estimation of handovers between cells

e.g. Rearrange in England

Bad solution better solution


Problem
Identification / Solution
Problem type:
NP-complete [Garey et al. 76]: untractable.
e.g., 100 cells into 2 groups, 2100 1032 solutions.
Combinatorial optimisation problem: all or none type, no step size.
Not a continuous optimisation problem: No notion of gradient.

Similar problems:
Geometric clustering: cluster groups according to position in space.
Not appropriate, geographical info not available
Quadratic Assignment Problem: assign finite number of cells to
locations
Very similar, except number of cells in LA not fixed
Graph Partitioning Problem: Define problem as a graph model
Most appropriate
K-way graph partitioning
problem
Given a weighted graph G = (V, E)
A cell i is represented by a vertex vi and its density of mobiles a
vertex weight wv(i)
Handover between two cells represented by an edge
3 2
4
3

4 5
2
2 4
3 1
3

Objectives: Partition G into k disjoint subsets S1, S2, , Sk with following


constraints:
1. Handover minimisation: Minimise the edge cut size, i.e. total
weight of edges whose incident vertices belong to different
subsets.
2. Partition into k equal size subsets
Comparison
Comparison of objectives:
Weighted graph problem LAC optimisation
Minimise the edge cut size Minimise the number of
between partition handovers between LAC
groups
Partition into k equal sized Subset are constrained with a
subsets maximum total weights
Maximise the size of a LAC to
reduce the number of groups
It is quite similar. created
Objectives 2 and 3 are valid if subset size is maximised and k is
reduced at maximum
2 Partitioning Approaches
K-way partitioning:
Partition at once in k subsets
Problem: very difficult

Recursive Bisection:
Most commonly used technique for partitioning
It first divides the graph into 2 equal sized subsets.
And recursively divide the 2 subsets, until size criterion is satisfied.

S1
S2
S1 S2
G
S3
Inconvenient: S4
Lack of global information
Difficult to maintain good balancing condition
Advantage:
It simplifies the problem
Very popular!
Multilevel Paradigm
Introduced by Barnard & Simon (1994).
3 phases: S S
Coarsening phase: 1
2

Coarsen down to a few hundred vertices.


The graph G0 is transformed into sequence
of smaller graphs G1, G2,, Gm S
S
Partitioning phase: 3
4

Partitioning heuristic used to compute k-


way partition of the graph Gm = (Vm, Em) Coarsening phase Uncoarsening phase
Uncoarsening phase:
The partition Sm of Gm is projected back S
through intermediate partitions 1

Use of a Refinement Method

S
Advantages: 3
S

Robust and Simple 4

Speed up the search process Partitioning phase

Improve the solution quality


Coarsening phase
A set of vertices of Gi is combined to form vertex v of Gi+1.
Approach: collapsing adjacent vertices.
Aim: Finding a matching of a graph, i.e. a set of edges, no two of
which incident on the same vertex.

Coarsening step with matching method


Coarsening phase (cont.)
Edge Contraction:
2 Vertices form a vertex with their weight added
Edge between them removed
Other edges preserved
1
1
0.3 2 1.5
1.7 0.3
0.8 3 4.9
2 1.6 0.8 3
3.2 3 2 5

Various ways to match vertices:


Random matching: vertices are visited in random order.
Heavy edge matching: select edge with highest weight [Karypis
95]
Light Edge matching: select edge with lowest weight. Important
for certain partitioning algorithms like Kernighan-Lin [Bui-Jones
93]
Partitioning phase
No single method is always best (Hendrickson & Leland, 1994)
However, 3 criteria to define best method:
Ability to respond to different input scale
Speed: not a real constraint
Solution quality (minimum cut-size with balanced subsets)

Various types of algorithms:


Combinatorial methods
Stochastic Methods
Gain-based Methods

S1
S2

S3
S4
Combinatorial Methods
Rely on graph connectivity
Group together vertices that are highly connected

Methods:
Spectral algorithm: constructs geometric representation
Greedy Graph growing algorithm:
Start from empty subset
Choose a random vertex and add it to subset
Select a neighbouring vertex with highest edge weight
Until subsets maximum size is reached

Starting vertex
Added vertices
Remaining vertices

Cut-size
Stochastic Methods
Methods with decisions based on uncertainty
3 methods:
Simulated Annealing
Randomly chooses solution and Wanders into search space, uphill or
downhill
Evaluation:
Offers good results, but not better than any other.
Long runtime.
Genetic Algorithm
Evolutionary-based technique: Combine good solutions
Evaluation: powerful, but limited when graph size is important
Ant Colony Optimisation:
Very recent
Based on emergent behaviour of ants when foraging
Drawback: Complex implementation
Move-based Methods
Iterative in nature

Kernighan-Lin/Fidduccia-Mattheyses (KL/FM)
Start from initial solution
Series of passes
Move bordering vertex to neighbouring subset according to Gain
Gain: benefit of reducing the cut-size
Choose maximum gain, but allow negative gain to move out of local minima
To avoid infinite loop, vertex allowed to move once (locking mechanism)
At the end, recall the best solution found

Advantage: Simple, flexible at handling various graph sizes and


objectives
Inconvenient: Short-sighted. But variations exist (e.g. relaxed locking
mechanism)
Uncoarsening phase
Partition Pm of the coarser graph is projected back to the
original graph,
S1 S2
S1
S2


S3 S4
S3 S4

An optimal solution Pi+1 is not anymore optimal after


projection Pi , but not far from it.
Need to use refinement algorithm:
KL/FM Refinement are efficient (Karypis 97)
Start from good solution
Few swaps converge to good solution
Implemented Solution

Multilevel paradigm:
Coarsening phase:
Random/Heavy-edge/Light-Edge matchings

Partitioning Phase
Greedy Graph Growing Algorithm (GGG)

Uncoarsening Phase
KL/FM algorithm

Testbed:
Data: Graph with 1657 vertices, 24707 edges
Over 100 runs per experiment
Result: Matching
Partition into 10 subsets
Heavy Edge Matching (HEM) provides best result

1. Random
2. HEM
3. LEM
Result: Partitioning
3 variants of Greedy Graph Growing (GGG) tested
For partitions into 2 to 26 subsets

No exponential increase: GGG is relatively constant


Very reliable
Refinement: KL/FM
Test improvement of cut-size during uncoarsening phase for a
partition into 10 subsets.

High improvement after each


uncoarsening step

Zoom in to a pass:
Able to avoid local minima
Refinement: KL/FM (cont.)
Weight Evolution:
System of balancing relaxation:
Allows subsets to have weight overpassing balancing condition
After each uncoarsening step, relaxation tightened
Result:
Explore more solutions
Be flexible: depending on average vertex weight
At the end, all subset weights are under threshold, and maximised
Comparison with libraries
Comparison with CHACO (Hendrickson 1998) and METIS
(Karypis 1997) libraries
Result:
Algorithm outperforms both libraries in cutsize
Weight balancing are nearly similar
But in term of speed, our software is much slower
Conclusion
Problem formulation and study:
NP-complete, Combinatorial
From a real-world problem to a Graph partitioning problem
Description: Cells as vertices, handovers as edges
Objectives: Minimisation of the edge cut size and subset balancing.
Multilevel paradigm as a framework for research
Efficacy and Simplicity of KL/FM and GGG over other
techniques
Improvements of solution compared to other available
libraries
Conclusion
Visualisation in geographical coordinates
For partitioning into 2 subsets and 10 subsets
WITHOUT knowledge of coordinate position of the cells
THANK YOU

Any Question?

You might also like