Download presentation
Presentation is loading. Please wait.
Published byFlora Sharp Modified over 8 years ago
1
Optimal Placement of Femto Base Stations in Enterprise Femtocell Networks Adviser: Frank, Yeong - Sung Lin Present by Li Wen Fang
2
Agenda Introduction Related Work Proposed Work LPP Model and Solution Experimental Setup in MATLAB and Performance Results Conclusions and Future Work 2
3
Agenda 3 Introduction Related Work Proposed Work LPP Model and Solution Experimental Setup in MATLAB and Performance Results Conclusions and Future Work
4
Introduction 4 A huge demand for power and data rates Outdoor environments : Long Term Evolution But indoor users still suffer from lower data rates owing to poor coverage and path loss due to obstacles. According to Cisco and Huawei [1], 70% of traffic in cellular networks comes from indoor User Equipments (UEs).
5
Introduction 5 Femto is a low power base station connected via backhaul to LTE Evolved Packet Core (EPC) system. The presence of a large number of Femtos can lead to problems like high inter-cell interference and frequent handovers. The optimal placement is very crucial for attaining desirable SNR (Signal-to-Noise Ratio) for the indoor UEs.
6
Introduction 6 Two major parameters that determine optimal Femto location are – distance between Femto and farthest point inside the building – minimum SNR needed by each UE
7
Agenda 7 Introduction Related Work Proposed Work LPP Model and Solution Experimental Setup in MATLAB and Performance Results Conclusions and Future Work
8
Related Work 8 3GPP introduced Femtos into cellular networks There exist works on optimal relay node placement in tunnels [5], sensor placement in terrain regions and optimal placement of Wi-Fi APs. Algorithm for optimal Femto placement based on distance between first Femto and macro BS is given in [6], [7], but authors did not consider Femto to Femto interference inside the building.
9
Related Work 9 Femtos are placed optimally inside the building to guarantee good SINR by considering interference between Macro and Femto but this solution does not guarantee good signal for each point inside the building[8]. Though authors of [9], [10] claim to provided an algorithm for the Femto placement inside building, it is of limited practical importance as it ignores walls inside building.
10
Related Work 10 Building upon the above work, we constrained optimal Femto placement problem by considering path loss across walls inside the building, thereby making it a realistic enterprise building scenario.
11
Agenda 11 Introduction Related Work Proposed Work LPP Model and Solution Experimental Setup in MATLAB and Performance Results Conclusions and Future Work
12
Proposed Work 12 We ensure that our model guarantees good SNR for indoor UEs on par with outdoor scenarios. In order to achieve this objective, the underlying idealization for enterprise building and assumptions are described below.
13
13 A.Building Dimensions B.System Model and Assumptions C.Formulation of Mathematical Model D.Femto placement constraints E.Linearization of Eqn (1) F.Linearization of Eqn (8) Proposed Work
14
A.Building Dimensions 14
15
A.Building Dimensions 15 Fig. 1. (a) Top view of a floor in the building.
16
A.Building Dimensions 16 Fig. 12. Numbering of rooms and calculation of number of walls between Femto and sub-region.
17
A.Building Dimensions 17
18
A.Building Dimensions 18 Fig. 1. (b) Top view of sub-regions in the building. The imaginary dotted sub-regions are formed for ease of calculations.
19
A.Building Dimensions 19 Each sub-region is referenced using indices i, j and k along x, y and z axes, respectively. Fig. 13. Numbering of sub-regions in Floor 1 and Floor 2.
20
B.System Model and Assumptions 20
21
C.Formulation of Mathematical Model 21
22
C.Formulation of Mathematical Model 22
23
C.Formulation of Mathematical Model 23 SNR in dB scale is given by Considering attenuation factors for SNR, the total attenuation is given by
24
C.Formulation of Mathematical Model 24 SNR in dB scale considering wall losses is given by respectively.
25
C.Formulation of Mathematical Model 25
26
C.Formulation of Mathematical Model 26 Then (2) becomes from (8) SNR received in any sub-region should be greater than minimum threshold SNR.
27
C.Formulation of Mathematical Model 27
28
C.Formulation of Mathematical Model 28 Our objective is to minimize These equations are first converted to convex equations and then to linear equations.
29
D.Femto placement constraints 29
30
D.Femto placement constraints 30
31
D.Femto placement constraints 31 Greater than the left wall and less than right wall Greater than the lower wall and less than upper wall
32
D.Femto placement constraints 32
33
E.Linearization of Eqn (1) 33
34
E.Linearization of Eqn (1) 34
35
E.Linearization of Eqn (1) 35 Equality can be converted into certain inequalities without loss of generality. Hence Eqn (1) becomes
36
E.Linearization of Eqn (1) 36 Let Written as
37
E.Linearization of Eqn (1) 37 Let Then, Eqn (19) becomes
38
E.Linearization of Eqn (1) 38 Eqns (24), (25) and (26) are convex constraints which are converted into linear constraints by applying PLAP
39
F.Linearization of Eqn (8) 39 Let It can be written as Eqn (32) is a convex constraint and is linearized by PLAP. Hence, we obain
40
F.Linearization of Eqn (8) 40 Let Then Eqn (8) can be written as
41
F.Linearization of Eqn (8) 41
42
Agenda 42 Introduction Related Work Proposed Work LPP Model and Solution Experimental Setup in MATLAB and Performance Results Conclusions and Future Work
43
LPP Model and Solution 43
44
LPP Model and Solution 44 The MILP algorithm which is used for solving the problem is actually an implementation of a branch- and-bound search with modern algorithmic features such as cuts and heuristics. The algorithm basically performs a bisection search on M and repeats the Branch and Cut procedure until problem becomes feasible.
45
45
46
LPP Model and Solution 46
47
Agenda 47 Introduction Related Work Proposed Work LPP Model and Solution Experimental Setup in MATLAB and Performance Results Conclusions and Future Work
48
Experimental Setup in MATLAB and Performance Results 48 We have simulated in a two-storied building of dimensions (120m × 80m × 12m), with six rooms of dimensions (40m × 40m × 6m) in each floor. Further each room is divided into 4 virtual sub- regions of dimensions (20m × 20m × 6m) as illustrated in Fig. 3. UEs are distributed randomly in each floor.
49
Experimental Setup in MATLAB and Performance Results 49 Fig. 3. Building Layout.
50
Experimental Setup in MATLAB and Performance Results 50 Fig. 10. UE Occupant Probabilities inside the building.
51
Experimental Setup in MATLAB and Performance Results 51 We use 4 Femtos for covering the entire building (F = 4), guaranteeing a minimum of −5dB SNR (γ min = −5) with indoor path loss constant of α = 3.5. Optimal Femto co-ordinates for occupant probability
52
A. Performance evaluation 52 To show the optimality of our Femto placement, we compared capacity received by each of UEs by placing Femtos – Randomly – Center of the building – Optimally
53
A. Performance evaluation 53 Fig. 11. Variation of SNR inside the building.
54
B. Simulation Results 54 We cross validated performance of our optimal Femto placement model using NS-3 simulator. We extensively tested our placement algorithm even by considering a more practical setup with interference from neighbouring Femtos.
55
B. Simulation Results 55 Fig. 4. Optimal Placement of Femtos in Floor 1. Fig. 5. Center Placement of Femtos in Floor 1. Fig. 6. Random Placement of Femtos in Floor1.
56
B. Simulation Results 56 Fig. 7. Optimal Placement of Femtos in Floor 2 Fig. 8. Center Placement of Femtos in Floor 2 Fig. 9. Random Placement of Femtos in Floor2
57
B. Simulation Results 57 As most of the users in our model are in close vicinity to Femto they are guaranteed to receive better SINR compared to random/center placement. Further we constrained our model to guarantee minimum SINR to all users, including those in the least occupant probability region. – Better Fairness
58
Agenda 58 Introduction Related Work Proposed Work LPP Model and Solution Experimental Setup in MATLAB and Performance Results Conclusions and Future Work
59
59 We have provided a model for optimal placement of Femtos based on occupant probabilities inside an enterprise building scenario considering realistic constraints. In future, we intend to provide an algorithm for optimal Femto placement for various environments considering more complex scenarios.
60
60 Thanks for listening !
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.