System Level Simulations for

Cellular Networks Using MATLAB

Sriram N. Kizhakkemadam, Swapnil Vinod Khachane, Sai Chaitanya Mantripragada
Samsung R&D Institute Bangalore

1
Cellular Systems

 Cellular Network: A wireless communication network that ideally provides ubiquitous

voice & data service
 Deployment:
 Base stations (BS) of varied transmit power levels are installed on a terrain
 Receivers of users (UE) decode signals from “attached” Base Stations
 Challenge: Terrain and propagation effects significantly affect performance
 Solution: Schemes that mitigate challenge have to be evaluated using Link & System Level
Simulations before deployment
2
Fig. courtesy of : 4GAmericas
System Level Simulation Process
Deployment
System configuration

Move UE’s: Pre-Defined

Mobility Pattern

Link Establishment
(BS-UE Association)

Scheduler
(PF / RR)

Channel Generation
(LoS/NLoS Channel)

Receive Processing Measurement & Processing

Feedback Report Handover Execution Basic
modules

Throughput
Modules

Metrics Mobility
(Throughput & Handover) Modules

3
Deployment: BS & UE Placement

Uniform Dropping
Discard region

10

11 9

12 2 8

3 1
[2] [1]
13 [3] [0] 7
[4] [5]
4 6

14 5 18

15 17

16

• Place BS at the center of a hexagon

• Randomly drop UE’s in rectangular area
• Discard UE’s dropped in the discard region
4
Link Establishment: Single Tier Case (1/5)
10

10 11 9

11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17 10

15 17 10 16 11 9

10 16 11 9 12 2 8

11 9 12 2 8 3 1

12 2 8 3 1 13 7
[2] [1]

3 1 13 [3] [0] 7 4 6
[4] [5]

13 7 4 6 14 5 18

4 6 14 5 18 15 17

14 5 18 15 17 10 16

15 17 10 16 11 9

16 11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17

15 17 16

16

• Find Nearest Cell to a UE using wrap around model

• Wrap around model ensures correct mapping of distance from UE to all Base Stations
5
Link Establishment: Single Tier Case (2/5)
10

10 11 9

11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17 10

15 17 10 16 11 9

10 16 11 9 12 2 8

11 9 12 2 8 3 1

12 2 8 3 1 13 7
[2] [1]

3 1 13 [3] [0] 7 4 6
[4] [5]

13 7 4 6 14 5 18

4 6 14 5 18 15 17

14 5 18 15 17 10 16

15 17 10 16 11 9

16 11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17

15 17 16

16

• Find the set of neighboring 19cells for each UE according to the wrap around
model
6
Link Establishment: Single Tier Case (3/5)
10

10 11 9

11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18
Path loss calculation
including shadow17effect
14 5 18 15 10

15 17 10 16 11 9

10 16 11 9 12 2 8

11 9 12 2 8 3 1

12 2 8 3 1 13 7
[2] [1]

3 1 13 [3] [0] 7 4 6
[4] [5]

13 7 4 6 14 5 18

4 6 14 5 18 15 17

14 5 18 15 17 10 16

15 17 10 16 11 9

16 11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17

15 17 16

16

Calculate path loss and shadow from neighboring 19 cells 3 sector

7
Link Establishment: Single Tier Case (4/5)
10

10 11 9

11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17 10

15 17 10 16 11 9

10 16 11 9 12 2 8

11 9 12 2 8 3 1

12 2 8 3 1 13 7
Serving Cell/Sector
[2] [1]

3 1 13 [3] [0] 7 4 6
[4] [5]

13 7 4 6 14 5 18

4 6 14 5 18 15 17

14 5 18 15 17 10 16

15 17 10 16 11 9

16 11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17

15 17 16

16

• Find serving cell and sector : Lowest path loss including shadow effect
8
Link Establishment: Single Tier Case (5/5)
10

10 11 9

11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

14 5 18
Interferers
4 6

14 5 18 15 17 10

15 17 10 16 11 9

10 16 11 9 12 2 8

11 9 12 2 8 3 1

12 2 8 3 1 13 7
Serving Cell/Sector
[2] [1]

3 1 13 [3] [0] 7 4 6
[4] [5]

13 7 4 6 14 5 18

4 6 14 5 18 15 17

14 5 18 15 17 10 16

15 17 10 16 11 9

16 11 9 12 2 8

12 2 8 3 1

3 1 13 7

13 7 4 6

4 6 14 5 18

14 5 18 15 17

15 17 16

16

• Identify interferers from 19 cells

9
Design of Parameters for Deployment
• Inter BS/Pico Distance and Number of Picos/Sector, UE’s/Sector
BS & Pico • Transmit Power Levels and Antenna patterns
Deployment

•Large Scale Fading: Line of Sight (LOS), Non-Line of Sight (NLOS)

•Shadow Fading
Fading •Small Scale Fading Models

•SINR Profile
Coverage •Throughput Profile
Maps

•Handover Parameters
•UE Mobility Pattern
Mobility

•Full Buffer
•Partial Buffer: Arrival Rates • Iterative
Data Traffic • Time Consuming

•Choice of Scheduler Algorithm and Granularity of Allocation

Resource •Transmission Modes: Single, Multiple, Coordinated Transmission
Allocation

•Throughput
•Mobility
Metrics

10
Design of Parameters Using Symbolic Math Toolbox

PPP model accurately

models the large scale SINR
for several practical
deployment scenarios

Fig. Deployment of Base Stations and Users Fig. Deployment of Base Stations by a major
according to a Poisson Point Process Model (PPP) cellular network provider in 40 X 40 Km area.

Integral of Hypergeometric Function

Evaluated Using Symbolic Math Toolbox

Obtain Density of BS for

required Throughput

Figs. obtained from: “A Tractable Approach to Coverage and Rate in Cellular Networks,” J. G. Andrews, F. Baccelli, and R. K. Ganti
Ergodic Rate as given in, “Modeling, Analysis and Design for Carrier Aggregation in Heterogeneous Cellular Networks,” X. Lin, J.G. Andrews and A. Ghosh 11
Design of Deployment Parameters Using Symbolic Math Toolbox
•Inter BS/Pico Distance and Number of Picos/Sector, UE’s/Sector
BS & Pico •Transmit Power Levels and Antenna patterns
Deployment

•Large Scale Fading: Line of Sight (LOS), Non-Line of Sight (NLOS)

•Shadow Fading
Fading •Small Scale Fading Models
Parameters Designed Using PPP
Model and Symbolic Math Toolbox
•SINR Profile
Coverage •Throughput Profile
Maps

•Handover Parameters
•UE Mobility Pattern
Mobility

•Full Buffer
•Partial Buffer: Arrival Rates
Data Traffic

•Choice of Scheduler Algorithm and Granularity of Allocation

Resource •Transmission Modes: Single, Multiple, Coordinated Transmission • Fewer Iterations
Allocation

•Throughput
•Mobility
Metrics

12
Scheduler
• Preliminaries:
• Each time-frequency resource is called Resource Element (RE) Slot 1 Slot 2
• A group of RE’s form an Allocation Unit (AU)
• Each spatial layer has N AU’s 1 2
• Assumptions:
• K Users in set U={U1, U2, …UK} 3 4
• L possible Transmission Modes (one at a time)
5 6
• Role of Scheduler:
• Allocate M*N resources among K Users in time-frequency : :
• Let w be a metric that indicates the weight considering
proportional fairness criterion, say N-1 N
• Then, the resource allocation is according to:

 
N L
iˆ, ˆj , aˆ , lˆ  arg max     wia, ,jl xia, ,jl
a 1 l 1 iU jU
NP-Hard Problem
s.t.  x
a 1...U l 1... L
a ,l
i, j  NL, i, j  1,2, , K Non-convex
x _ opt (iˆ, ˆj , aˆ , lˆ)  1

r  Instantaneous rate to MS j and k

  j, k
r ja, k,l
a ,l
w  log1  
i, j

 T  1R j , k 

R j, k  Average rate to MS j and MS k over time window T

13
Scheduler: Numerical Solution Using MATLAB

• Use MATLAB’s Compiler Runtime Engine

• Solve using MATLAB’s Optimization Toolbox:

iˆ, ˆj, aˆ, lˆ   arg max    

N L
wia, ,jl xia, ,jl
a 1 l 1 iU jU

s.t.  x
a 1...U l 1...L
a ,l
i, j  NL, i, j  1,2, , K

x _ opt (iˆ, ˆj , aˆ , lˆ)  1

• Use Binary Integer Programming to Solve the problem

• Link the script for optimization to C++ SLS using dynamic linked libraries

• Harnesses the power of MATLAB’s Optimization Toolbox

• Save implementation time for developing complex optimization routines

14
Scheduler: Numerical Solution Using MATLAB

 MATLAB’s Optimization & Global Optimization Toolbox:

• Wide array of solvers helps to choose different optimization techniques

• CVX Toolbox increases options to solve the problems

• Ease of representation

• Certain scheduling algorithms have multiple local minima

• Solution: Use functions from Global Optimization Toolbox

 Advantages:

• Evaluation of System performance:

• Optimal Algorithm with MATLAB Toolboxes provides limits of performance

• No artificial limit on performance due to sub-optimal algorithms

• Give insights to standards on limits of performance, paving way for improved system design

• 50% decrease in simulation time

15
 Summary
• Faster Turn-Around Time Using Symbolic Math Toolbox for design of
deployment and mobility parameters

• Harness numerical optimization package of MATLAB to solve complex

scheduler optimization and interface it to C++ based SLS using Matlab
Compiler Runtime

16