1. Guomei Zhang, Man Chu, Jie Li Personal Ubiquitous Computing 2016
Interference coordination based on random fractional spectrum reuse in femtocells toward Internet of Things
Guomei Zhang, Man Chu, Jie Li
Personal Ubiquitous Computing 2016

2. Outline Motivation Problem Scope and Challenges Objective
Problem Formulation
Solution
Experimental Results
Md Abdul Alim

3. Indoor femtocell is the solution
I. Motivation
Demand for mobile data will increase 1000x by 2020
Ensuring coverage and QoS is tricky
Uninterrupted service in high-rise buildings for IoT devices: a challenge
Popularity of OSNs makes many people integrate OSNs into their daily lives
OSNs is a great platform for information sharing and facilitate information sharing
Indoor femtocell is the solution
Md Abdul Alim

4. I. Motivation (contd.)
Many IoT applications occur indoors: smart homes, smart offices, smart industrial plants and smart museum
Femtocells:
spatial reuse and spectral efficiency
Macrocell offload
Short distance: less power reqd.
Popularity of OSNs makes many people integrate OSNs into their daily lives
OSNs is a great platform for information sharing and facilitate information sharing
Md Abdul Alim

5. Challenges More Femtocells bring more challenges:
Interference management
QoS provisioning
Information security guarantee
Current literature focuses:
Cross tier interference mitigation
With increased IoT devices
Co tier interference between femtocells is getting prominent
Md Abdul Alim

6. Contributions
Assumption: Macrocell and femtocells use separate frequency bands
All femtocells in a multi-story building is divided in two groups
Each group: half orthogonal frequency
A femtocell in one group reuse the resources of another (ROFS)
Each group: full orthogonal frequency
A femtocell in one group retreats randomly over one half of frequency resources (RRFS)
Optimization problem is formed to find out the appropriate probability value for maximizing throughput
Md Abdul Alim

7. System Model
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.
Md Abdul Alim

8. Notations Femtocell scenario: multi-story building
N floors
One stripe in each floor
M apartments in a single stripe
One FAP (femto access point) in each apartment
Total NF=MxN FAPs
Considered downlink scenario
Interference induced from neighboring FAP including dashed cell upstairs and downstairs (dashed line)
Dominant interference links involves penetration of of one internal wall or floor
Centralized FAP controller connected to all FAPs using gateway
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.
Md Abdul Alim

9. Channel Model 15 RBs, total BW 3MHz
Considers indoor channel comprising of
Path loss
Shadowing fading (also called slow fading)
Fast fading
Channel gain on i-th RB between user k served by FAP n and FAP m (in dB)
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.
Md Abdul Alim

10. Channel Model (contd) The first term represents the propagation loss
First term: distance dependent free-space path loss, second term: indoor attenuation
d<k,n>m : distance between k in femtocell n and FAP m
a: # penetrated floors
q: # of penetrated walls
Liw: per wall penetration loss
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.
Md Abdul Alim

11. Channel Model (contd)
The second factor in channel gain represents log-normal shadowing fading with zero mean and a standard deviation of σξ
The last term corresponds to Rayleigh fading power in dB
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.
Md Abdul Alim

12. Problem Objective
RB allocation through Interference coordination (IC) method for
Reusing partial spectrum randomly among femtocells
Probability associated : p
The objective is to increase the system utility sum rate by optimizing the value of p
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.
Md Abdul Alim

13. ROFS (randomly occupying fractional spectrum) scheme
Total resources are equally partitioned
Group 1 : R1
Group 2: R2
femtocells in one group randomly (with probability p) occupy the resources allocated to another group: dynamic
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.

14. RB allocation: ROFS (contd)
rk,i is the instantaneous data rate at i-th RB
rk is the average data rate of k
The rate has to be adjusted for stochastic interference
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.

15. ROFS (contd)
Case 1 (type1): which femtocell in F2 will utilize the RBs of R1 , the data rate becomes average over different cases
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.

16. Case 2 (type2): The data rate becomes average over different cases
ROFS (contd)
Case 2 (type2): The data rate becomes average over different cases
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.

17. Algorithm for RB allocation
Choose the user with the highest average data rate gain
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.

18. How to choose p?
We’ll come to that after introducing RRFS (randomly releasing fractional spectrum)
Full frequency assigned to both groups
To mitigate interference, a dynamic resource allocation method is utilized
Femtocells randomly release half resources with prob p
We are not focusing on the details about how to identify the truth and the false. Instead, we are based on existing techniques, which are referred to as monitors in our work.
Md Abdul Alim

19. RRFS Similar to what we did for ROFS, we get the SINR Md Abdul Alim

20. Optimization problem
The optimization of p in both ROFS and RRFS schemes is same, so we focus on ROFS
Aggregate utility function for RB set R1
Where
Md Abdul Alim

21. Problem formulation (Contd)
Aggregate utility function for RB set R2
Where
Md Abdul Alim

22. Problem formulation (Contd)
Finally, the optimization problem for the probability p is given by
This rate function is logarithmic and concave
The second-order derivative can’t proven to be less than zero
Which makes the overall function neither concave nor convex
Exhaustive search is an option
This work proposes a suboptimal approach based on particle swarm optimization PSO
Md Abdul Alim

23. PSO algorithm Complexity:
Average throughput calculation, C1=K(N1+N2)(N2 2N2+N1 2N1
Suboptimal p by PSO, C2=K[N22N1+(N2-1)2N2+N12N2+(N1-1)2N1]ST
Md Abdul Alim

24. Experimental Evaluation
Datasets
Md Abdul Alim

25. Experimental Results
Md Abdul Alim

26. Experimental Results
Md Abdul Alim

27. Experimental Results