A Novel Multiobjective Framework for Cell Switch-Off in Dense - - PowerPoint PPT Presentation

▶
a novel multiobjective framework for cell switch off in
SMART_READER_LITE
LIVE PREVIEW

A Novel Multiobjective Framework for Cell Switch-Off in Dense - - PowerPoint PPT Presentation

Feb 25, 2023 126 likes •512 views

A Novel Multiobjective Framework for Cell Switch-Off in Dense Cellular Networks alez G 1 , Halim Yanikomeroglu 2 , David Gonz a-Lozano 1 , Silvia Ruiz Boqu e 1 Mario Garc 1 Departament of Signal Theory and Communications Universitat

slide-1
SLIDE 1

A Novel Multiobjective Framework for Cell Switch-Off in Dense Cellular Networks

David Gonz´ alez G1, Halim Yanikomeroglu2, Mario Garc´ ıa-Lozano1, Silvia Ruiz Boqu´ e1

1 Departament of Signal Theory and Communications

Universitat Polit` ecnica de Catalunya, Spain

2 Departament of System and Computer Enginnering

Carleton University, Canada

IEEE International Conference on Communications (ICC) 2014 Mobile and Wireless Networking Symposium Sydney, Australia: 10-14 June 2014

slide-2
SLIDE 2

Outline

slide-3
SLIDE 3

Introduction Framework Description Closing Remarks

Motivation

1 Energy Efficiency (EE) in cellular networks.

Cellular industry is growing exponentially. Hyper dense small cells deployment → boost energy consumption!

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 1 / 20

slide-4
SLIDE 4

Introduction Framework Description Closing Remarks

Motivation

1 Energy Efficiency (EE) in cellular networks.

Cellular industry is growing exponentially. Hyper dense small cells deployment → boost energy consumption!

2 Why switch-off base stations?

Energy consumption models Solution approach → Multiobjective Optimization

(aggregate capacity ↔ active cells).

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 1 / 20

slide-5
SLIDE 5

Introduction Framework Description Closing Remarks

CSO: Problem statement & practical insights

Main intuition

Switch off lightly loaded base stations to save energy.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 2 / 20

slide-6
SLIDE 6

Introduction Framework Description Closing Remarks

CSO: Problem statement & practical insights

Main intuition

Switch off lightly loaded base stations to save energy.

The challenge

The CSO problem consists in determining the largest set of cells that can be switched off without compromising the QoS.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 2 / 20

slide-7
SLIDE 7

Introduction Framework Description Closing Remarks

CSO: Problem statement & practical insights

Main intuition

Switch off lightly loaded base stations to save energy.

The challenge

The CSO problem consists in determining the largest set of cells that can be switched off without compromising the QoS. Theoretical aspects: Deployments density. Traffic behavior. Network capacity. ICIC.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 2 / 20

slide-8
SLIDE 8

Introduction Framework Description Closing Remarks

CSO: Problem statement & practical insights

Main intuition

Switch off lightly loaded base stations to save energy.

The challenge

The CSO problem consists in determining the largest set of cells that can be switched off without compromising the QoS. Theoretical aspects: Deployments density. Traffic behavior. Network capacity. ICIC. Practical aspects: Coverage. Switch on/off transitions. Architecture. Others.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 2 / 20

slide-9
SLIDE 9

Introduction Framework Description Closing Remarks

CSO: Related work

Ref. Context C1 C2 C3 C4 C5 C6 [283] CSO Heuristic CE × × P Full buffer [284] CSO Analytical CE P × × Full buffer [285] Planning: how to deploy cell for minimizing energy consumption Analytical NA P × NA NA [286] CSO Heuristic CE × × × Poisson [287] Cell size adaptation Heuristic CE × × × Full buffer [288] An interesting RRM strategy for energy savings Heuristic CE × × × Poisson [289] CSO Heuristic Both × ×

  • Full buffer

[277] CSO Analytical CE × × × Poisson [290] CSO Analytical CE × P × Full buffer [291] CSO Heuristic CE × P P Full buffer [292] CSO Heuristic CE P P × Full buffer [293] CSO Heuristic CE ×

  • P

Realistic [294] CSO Heuristic SD × × P Realistic [295] Cell size adaptation Heuristic Both × × × Realistic [296] CSO Heuristic CE × × × Full buffer [297] CSO Heuristic CE × × × Full buffer [298] CSO Heuristic CE × × × Several models [299] Impact of power reduction on coverage and capacity Analytical NA

  • NA

Full buffer

P: Partially CE: Centralized SD: Semidistributed DI: Distributed

Heuristic is the preferred approach.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 3 / 20

slide-10
SLIDE 10

Introduction Framework Description Closing Remarks

CSO: Related work

Ref. Context C1 C2 C3 C4 C5 C6 [283] CSO Heuristic CE × × P Full buffer [284] CSO Analytical CE P × × Full buffer [285] Planning: how to deploy cell for minimizing energy consumption Analytical NA P × NA NA [286] CSO Heuristic CE × × × Poisson [287] Cell size adaptation Heuristic CE × × × Full buffer [288] An interesting RRM strategy for energy savings Heuristic CE × × × Poisson [289] CSO Heuristic Both × ×

  • Full buffer

[277] CSO Analytical CE × × × Poisson [290] CSO Analytical CE × P × Full buffer [291] CSO Heuristic CE × P P Full buffer [292] CSO Heuristic CE P P × Full buffer [293] CSO Heuristic CE ×

  • P

Realistic [294] CSO Heuristic SD × × P Realistic [295] Cell size adaptation Heuristic Both × × × Realistic [296] CSO Heuristic CE × × × Full buffer [297] CSO Heuristic CE × × × Full buffer [298] CSO Heuristic CE × × × Several models [299] Impact of power reduction on coverage and capacity Analytical NA

  • NA

Full buffer

P: Partially CE: Centralized SD: Semidistributed DI: Distributed

Most of solutions require real-time centralized operation.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 3 / 20

slide-11
SLIDE 11

Introduction Framework Description Closing Remarks

CSO: Related work

Ref. Context C1 C2 C3 C4 C5 C6 [283] CSO Heuristic CE × × P Full buffer [284] CSO Analytical CE P × × Full buffer [285] Planning: how to deploy cell for minimizing energy consumption Analytical NA P × NA NA [286] CSO Heuristic CE × × × Poisson [287] Cell size adaptation Heuristic CE × × × Full buffer [288] An interesting RRM strategy for energy savings Heuristic CE × × × Poisson [289] CSO Heuristic Both × ×

  • Full buffer

[277] CSO Analytical CE × × × Poisson [290] CSO Analytical CE × P × Full buffer [291] CSO Heuristic CE × P P Full buffer [292] CSO Heuristic CE P P × Full buffer [293] CSO Heuristic CE ×

  • P

Realistic [294] CSO Heuristic SD × × P Realistic [295] Cell size adaptation Heuristic Both × × × Realistic [296] CSO Heuristic CE × × × Full buffer [297] CSO Heuristic CE × × × Full buffer [298] CSO Heuristic CE × × × Several models [299] Impact of power reduction on coverage and capacity Analytical NA

  • NA

Full buffer

P: Partially CE: Centralized SD: Semidistributed DI: Distributed

Coverage analysis is oftem missed.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 3 / 20

slide-12
SLIDE 12

Introduction Framework Description Closing Remarks

Multiobjective Optimization: Essentials

1 Target: problems with conflicting criteria.

F = { fi(x) : Rn → R, i = 1, 2, · · · , m }

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 4 / 20

slide-13
SLIDE 13

Introduction Framework Description Closing Remarks

Multiobjective Optimization: Essentials

1 Target: problems with conflicting criteria.

F = { fi(x) : Rn → R, i = 1, 2, · · · , m }

2 Structure:

Design variables: x = [ x1, x2, · · · , xn ], (x ∈ X). Feasible set: X = X1 × X2 × · · · × Xn, (domains). Objetive space: f : X → Rm, f(x) = [ f1(x), f2(x), · · · , fn(x) ]. Constraints.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 4 / 20

slide-14
SLIDE 14

Introduction Framework Description Closing Remarks

Multiobjective Optimization: Essentials

1 Target: problems with conflicting criteria.

F = { fi(x) : Rn → R, i = 1, 2, · · · , m }

2 Structure:

Design variables: x = [ x1, x2, · · · , xn ], (x ∈ X). Feasible set: X = X1 × X2 × · · · × Xn, (domains). Objetive space: f : X → Rm, f(x) = [ f1(x), f2(x), · · · , fn(x) ]. Constraints.

3 Optimality: Pareto efficiency (x⋆ and the set X ⋆).

x1 ≻ x2, ⇐ ⇒ fi(x1) ≤ fi(x2) ∧ ∃ j | fj(x1) < fj(x2) x⋆ ∈ X ⋆ ⇐ ⇒ ∄ x ∈ X | x ≻ x⋆.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 4 / 20

slide-15
SLIDE 15

Introduction Framework Description Closing Remarks

Multiobjective Optimization: Essentials

Figure: A representation of the Pareto Front.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 4 / 20

slide-16
SLIDE 16

Introduction Framework Description Closing Remarks

Multiobjective Optimization: Essentials

(a) The hypervolume indicator (υ). (b) The nonuniformity index (̺). Figure: Quality measures in multiobjective optimization. υ(X ⋆, xref) = Λ

x∈X

ˆ x | x ≺ ˆ x ≺ xref

  • ̺ = df + dl + N−1

i=1 |di − ¯

d| df + dl + ¯ d (N − 1)

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 4 / 20

slide-17
SLIDE 17

Introduction Framework Description Closing Remarks

System Model

Downlink of an OFDMA cellular network → L cells. Target → Average ICI conditions (full reuse). Any topology: → Network geometry (G ∈ RA×L). Flexible analysis: → Operator-defined QoS policies.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 5 / 20

slide-18
SLIDE 18

Introduction Framework Description Closing Remarks

System Model: Terminology

1 Network Operation Point (NOP)

x ∈ {0, 1}L (design variable). |X| = (2L − 1) (search space).

2 Network Energy Level (NEL)

Xj = {x ∈ X | x · 1 = j} (a set of NOPs).

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 6 / 20

slide-19
SLIDE 19

Introduction Framework Description Closing Remarks

System Model: Formulation

1 Cell selection

RRS = G · diag (pRS ⊙ x )

2 SINR figures ← actual ICI.

Ψ = [(S ⊙ G) · (pD ⊙ x )] ⊘

  • [(Sc ⊙ G) · (pD ⊙ x )] ⊕ σ2

3 Coverage aspects

Minimum received power: RRS (a, l⋆) ≥ PRx

min.

Minimum SINR: Ψ (a) ≥ ψmin. H (a) = u (Ψ (a) − ψmin) · u

  • RRS (a, l⋆) − PRx

min

  • · log2(1 + Ψ (a))
  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 7 / 20

slide-20
SLIDE 20

Introduction Framework Description Closing Remarks

System Model: Objective functions

1 Number of active cells (f1)

f1 = x · 1

2 Weighted Network Capacity (f2)

f2 = (B · A) ·

  • (H ⊙ Γ )T · S
  • ⊙ n
  • · 1
  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 8 / 20

slide-21
SLIDE 21

Introduction Framework Description Closing Remarks

System Model: Multiobjective problem

minimize [ f1 (x) , −f2 (x) ] subject to: . (vT · 1) A ≤ κCOV . x ∈ {0, 1}L, x = 0 Solution tool: Multiobjective Evolutionary Algorithms (MOEAs).

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 9 / 20

slide-22
SLIDE 22

Introduction Framework Description Closing Remarks

MOEAs: An approach to CSO

f1 = ⇒ discontinuities f2 = ⇒ non-convex, local optima

The Non-dominated Sorting Genetic Algorithm II (NSGA-II)

[Deb et al.] @ IEEE Trans. on Evolutionary Computation. A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 10 / 20

slide-23
SLIDE 23

Introduction Framework Description Closing Remarks

CSO: Conceptual solution design

Figure: Conceptual design of the proposed framework.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 11 / 20

slide-24
SLIDE 24

Introduction Framework Description Closing Remarks

CSO: Conceptual solution design

Figure: Conceptual design of the proposed framework.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 11 / 20

slide-25
SLIDE 25

Introduction Framework Description Closing Remarks

Results: Test case and evaluation setting

Figure: Small dense deployment. Figure: Traffic distribution (Γ).

Minimum SINR (ψmin)

  • 7.0 dB

Minimum received power (PRx

min)

  • 123 dBm

Outage threshold (κCOV) 2.0 %

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 12 / 20

slide-26
SLIDE 26

Introduction Framework Description Closing Remarks

Results: Estimation of the Pareto Front

(a) Estimated Pareto Front. (b) Convergence pattern (NSGA-II). Figure: Estimation of nondominated solutions: the set X ⋆

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 13 / 20

slide-27
SLIDE 27

Introduction Framework Description Closing Remarks

Results: Evaluation of NOPs and QoS

System level simulations setting:

1 Scheduling → Best first (rT = 250 kbps). 2 QoS → Satisfaction ≥ 97.5%. 3 QoS checking interval → 1 s. 4 System bandwidth → 5.4 MHz. 5 Users distribution → ( Γ, λ )

Inter-arrival time (exponential, 1/λ = 0.075 s). Session time (exponential, 1/µ = 60 s).

6 Two methodologies:

NOP selection based on NEL sorting (binary search). Single NEL performance evaluation.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 14 / 20

slide-28
SLIDE 28

Introduction Framework Description Closing Remarks

Results: Evaluation of NOPs and QoS

(a) NEL selection CDF. (b) Performance. Figure: NOP selection based on NEL sorting: solutions in X ⋆

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 15 / 20

slide-29
SLIDE 29

Introduction Framework Description Closing Remarks

Results: Evaluation of NOPs and QoS

(a) Users rate. (b) QoS. Figure: Single NEL performance evaluation: NEL=18.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 16 / 20

slide-30
SLIDE 30

Introduction Framework Description Closing Remarks

Results: Evaluation of NOPs and QoS

Benchmarks:

1 Cell Zooming algorithm.

[Z. Niu et al.] @ IEEE Communications Magazine. Cell Zooming for Cost-Efficient Green Cellular Networks. 2010.

2 Improved Cell Zooming algorithm.

[F. Alaca et al.] @ IEEE Globecom 2012. A Genetic Algorithm based Cell Switch-Off Scheme for Energy Saving in Dense Cell Deployments.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 17 / 20

slide-31
SLIDE 31

Introduction Framework Description Closing Remarks

Results: Evaluation of NOPs and QoS

(a) NEL sel. vs. system load (b) QoS vs. system load Figure: Performance comparison among several CSO schemes

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 17 / 20

slide-32
SLIDE 32

Introduction Framework Description Closing Remarks

Results: Evaluation of NOPs and QoS

Table: Performance indicators: feasibility

Scheme NEL QoS Handovers Transitions x18 ∈ X ⋆ 18.00 97.81 0.00 0.00 NOP sel. (NEL sorting) 18.47 98.74 7.14 0.31 Cell zooming 41.45 99.05 57.56 3.47 Improved cell zooming 25.36 99.87 80.14 7.47

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 18 / 20

slide-33
SLIDE 33

Introduction Framework Description Closing Remarks

Results: Evaluation of NOPs and QoS

(a) NEL vs. target rate (b) QoS vs. target rate (c) NEL vs. QoS Figure: Impact of operational parameters: tradeoffs

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 19 / 20

slide-34
SLIDE 34

Introduction Framework Description Closing Remarks

Conclusions

1

Novelty MO framework for CSO

Network/traffic-specific characterization → Pareto eff. NOPs.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 20 / 20

slide-35
SLIDE 35

Introduction Framework Description Closing Remarks

Conclusions

1

Novelty MO framework for CSO

Network/traffic-specific characterization → Pareto eff. NOPs.

2 Feasibility

Low real-time complexity. Less transitions/handovers.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 20 / 20

slide-36
SLIDE 36

Introduction Framework Description Closing Remarks

Conclusions

1

Novelty MO framework for CSO

Network/traffic-specific characterization → Pareto eff. NOPs.

2 Feasibility

Low real-time complexity. Less transitions/handovers.

3 Merit → excellent performance.

  • H. Yanikomeroglu @ IEEE ICC ’14

Multiobjective Cell Switch-Off ◮ 20 / 20

slide-37
SLIDE 37

Thanks !!

[ david.gonzalez.g@ieee.org ] [ Halim.Yanikomeroglu@sce.carleton.ca ]