Cluster Based Femtocell Efficiency Evaluation Omar Arafat, Mark A. Gregory RMIT University Melbourne, Australia {omar.arafat, mark.gregory}@rmit.edu.au Abstract: Femtocell networks are a cost-effective way to deliver the bandwidth needed to support advanced mobile data services in heavily congested urban and indoor locations. Locally-networked femtocell arrays are used to enhance mobile network coverage and capacity in crowded urban and indoor locations. Cluster based femtocell deployments in large indoor environments provide a flexible approach to improve localized coverage and performance. In this paper, the performance of cluster based femtocell deployments along with a capacity based cognitive resource allocation scheme is analysed under different channel configurations in a hybrid access control network. The results of a performance analysis of the cluster based femtocell configurations in a priority based users’ network are presented.
Keywords — femtocell; interference; clustering; resource allocation; interference avoidance I. INTRODUCTION Rapid Internet usage growth has led to higher demand for reliable data transmission and increased network capacity. The femtocell Access Point (FAP) was introduced into Long Term Evolution (LTE) and LTEAdvanced networks to alleviate some of the issues being found with increased use of mobile phones in urban areas and especially within buildings. FAPs provide highquality voice services and high-speed mobile data services at a fraction of the cost of macro nodes, however the femtocell coverage area is limited and there remains a need to provide backhaul. In today’s networks, FAPs can be deployed in large indoor environments to provide consistent coverage and to provide additional coverage in areas where large numbers of customers congregate. Femtocell deployments can improve spectrum utilization by employing co-channel operation. In a dense FAP coverage area, interference control has to be performed in a very autonomous way following a specific infrastructure topology and traffic distribution. Interference control provides a flexible and important mechanism to adjust transmission characteristics to respond to random changes in the network conditions. Spectrum management is a key issue that increases the complexity of cell location planning for ad-hoc deployments. The aim of this paper is to study the performance of cluster based FAP deployments in indoor locations such as enterprise deployments, shopping malls, or other large public buildings (airport, bank, campus) utilizing different operating channel configurations. Opportunistic channel scheduling is performed at the FAP considering the Quality of Service (QoS) requirements of the users according to their access levels (subscribers and nonsubscribers) and the network’s throughput [1].
Afaz Uddin Ahmed University of Malaya Kuala Lumpur, Malaysia
[email protected] Open access, closed access and hybrid access are the three existing access-control methods that decide users’ connectivity to the FAP. Open access is suitable for interference reduction and cost optimization. However, utilizing FAP resources for an unknown number of nonsubscribers may degrade the QoS of the FAP subscribers. FAPs can be optimized for indoor coverage to provide improved service for subscribers. Closed access control allows only a particular group of users to get access to the FAP, thus avoiding unwanted traffic congestion and possible interference from non-subscribers. In this case, the QoS is guaranteed at the expense of decreasing spectral efficiency. Hybrid access control tunes the resource ratio according to the number of femtocell subscribers and non-subscribers. Limited resources are available to users who are within the femtocell coverage range and the “closed subscriber group” (CSG) have higher resource access privileges. Hybrid access control provides a compromise between the impact on the performance of subscribers and level of access granted to non-subscribers allowing them to possess a limited amount of features [2-5].
Fig.1: Cluster deployment of FAP
Open access leads to a lower QoS when the number of non-subscribers is too high [3] and in CGS mode, the cell capacity will not be sufficient to satisfy the QoS requirements of the non-subscribers. So preferably, hybrid access is used for cluster based femtocell deployments. Performance can be reduced due to femto-to-femto interference and for high femtocell concentrations there are different approaches found in the literature for controlling power [6-8] and allocating sub-channels [914]. Hybrid approaches are considered in [15-17] and an
energy efficient cognitive radio approach is provided in [18]. Other select cognitive radio based approaches are provided in [19-22]. A virtual clustering algorithm for cognitive femtocell networks was adopted in [23] that exploits the FAPs relative position to form logical clusters. In this paper, the performance of cluster format FAPs in wide indoor areas with different channel deployment configurations is analyzed with two user types, subscriber and nonsubscriber. A cognitive co-channel deployment is also proposed for a hybrid access control mechanism with capacity based cell selection. The cluster based deployment of FAPs will allocate spectrum intelligently to users according to their priority and taking into account network congestion. The rest of paper is organized as follows. The system model is described in Section II and the deployment configurations are provided in Section III. The simulation model and results are provided in Section IV and Section V contains the conclusion. II. SYSTEM MODEL The level of interference in a dense FAP network depends upon the spectrum utilization strategy. In this paper three deployment and spectrum utilization configurations are considered and simulated using the same network model. An indoor area of 100m x 100m is studied with OFDMA used in the downlink. FAPs are deployed so that no coverage holes exist. The simulation includes femto-tier interference with no macrocell interference. The users are distributed using a homogeneous Poisson Point Distribution (PPP) and users are 1 m apart from each other and no closer than 1 m to the FAPs. Subscribers and non-subscribers are enlisted in the CSG and all of the FAPs possessing the same CGS are connected with X2 interfaces and share updates regarding user parameters which are assembled in the uplink. FAPs are connected to the core network via Home eNode Base Station Gateways (HeNB-GW) using the S1 interface. They are also connected to a Hetnet Management System (HMS) using the TR-069 protocol, which allows the FAPs to update and adapt to the radio environment with semistatic configurations (e.g: adaptive power level, maximum throughput level). Each user is able to assess the link gain from its neighboring FAPs with the help of transmitter specific training sequences. Distinct feedback channels are assumed that enable user devices to send link gain estimates to nearby FAPs. Only the downlink communication scenario is considered in this paper. An example of the simulation is given in Figure 2. Path-loss Model: Indoor path-loss scenarios for line-of-sight (LOS), WINNER II channel models were analysed [24]. The simulation was carried out using a large indoor area without considering the indoor obstacle and walls losses. L LOS ( dB ) = 18 .7 log 10 d m + 46 .8
+ 20 log 10
[
f c GHz 5 .0
]
Fig. 2: Sample of the simulation environment
A.
Adaptive Power Model: The downlink transmission range of the FAPs was found using the pilot signal power. Tuning the transmit power of the channels was used to maintain QoS. The speed of the power control is determined by the channel coherence time of a communication link. In cluster based FAP network, the FAP channel condition does not have wide variation as the cell radius is smaller and the users are mainly nomadic with relatively slow motion. For fixed transmission power, the FAP sets the transmit power level to a pre-defined value, however, an adaptive power configuration is used to maintain QoS. In this adaptive power control approach, the transmit power level is adjusted based on the end user interference level. Easy plug and play features of FAPs are critical for indoor deployment since many parameters depend on the local radio condition. Therefore auto-configuration capability to adjust the transmit power level was adopted and the transmit power of each FAP is auto-configured by the HMS to a value that is above the power received from the neighboring interfering FAPs within the coverage area. The adaptive transmit power from FAPs can be expressed as:
PFAP
where
where fc , d m , w and L w are frequency in GHz, distance between FAP and user, number of penetrated walls and penetration loss for each wall, respectively.
L FAP
(2)
is the path-loss at the line of sight of the
targeted user. The transmit power is considered nonsensitive to errors and indoor obstacles. Only the downlink power scenario is taken into account and assumed that each user connects to the FAP with the best SINR. The maximum uplink power calculation can also be expressed considering the neighboring FAP users served on the same channel. B.
(1)
F Fi i ∑ ∑ = min PFAP − L FAP + L FAP , Pmax =1,i ≠ Fi i =1,i ≠ Fi i1 44 424444 3 4neighbouri ng FAPs
SINR Model: The use of OFDMA in LTE networks provides a robust defense against multi-path interference and provides frequency selectivity [25]. In an OFDMA LTE network, FAPs have the advantage of allowing the allocation of orthogonal frequency/time resources to users. During the simulation, the available bandwidth is divided into resource blocks (RB) and sub-carriers are allotted to
each of the RB. Randomly deployed FAPs operate using the same bandwidth. The RB assumption is adopted from the 3GPP-LTE approach provided in [26]. For k sub-carriers, the signal-to-interference-plusnoise-ratio (SINR) expression for the user is: −α f PF , k G F , f , k D F
SINR f , k =
σ
2
+
− α fi ∑ Pi , k G i , f , k D i i∈ F i
(3)
2
where σ , PF ,k , G F , f ,k , D F , Pi ,k , Gi , f ,k and Di are the additive white Gaussian noise power, transmission power from serving FAP to target user, random channel gain for the serving FAP, distance from serving FAP to target user, transmit power of interfering FAPs, random channel gain from interfering FAP to target user and distance from interfering FAP to target user, respectively. α f is the path-loss exponent of the link from FAP to the user, respectively. Fi is the set of interfering FAP. In Figure 3, the signal strength of the FAPs is illustrated using the approach described.
Fig. 3: Received signal strength of the FAPs over the functioning area
III.
Channel Deployment configuration
A. Dedicated channel deployment: In a dedicated channel deployment, the performance of the network is limited by bandwidth selection. Separate spectrum is assigned to each FAP and this reduces the cotier interference, however, decreasing spectrum availability makes it infeasible to apply this approach in a congested region. The number of channels allocated to a FAP is inversely related to the ratio of FAPs deployed in the cluster based scenario. The number of users supported in dedicated channel deployment can be expressed as:
C ( dc ) =
N ch ∆f max ( N FAP )
log 2 (1 + α SINR )
(4)
where N ch , N FAP , and ∆ f are the number of available channels in the assigned bandwidth, the number of FAPs in the functioning area and sub-carrier spacing, respectively. B.
Co-Channel deployment:
In a dedicated channel deployment scenario valuable spectrum is inefficiently utilized so the practical solution is co-channel deployment. In a cluster based scenario cochannel deployment is not inevitable and it is a risky configuration to deploy in a dense user environment because even with the capacity enhancement and more efficient spectrum utilization, it is more suitable for a heterogeneous network. The cell selection process is also easier for users as they do not have to search for cells in different frequency bands. The capacity of FAPs in cochannel channel deployment can be expressed as:
[
]
C ( dc ) = max ( N ch ) ∆ f log 2 (1 + α SINR )
(5)
C. Cognitive co-channel deployment: Considering the tradeoff between the co-channel interference and spectrum usage efficiency, a cognitive cochannel deployment is adopted. In this spectrum management based resource allocation with users’ priority level, a minimum level of resource is assigned to the CGS. The strategy aims to develop a fully distributed scalable and autonomous channel allocation scheme with minimal information sharing between FAPs in the cluster. Initially FAP ensures the minimum uninterrupted service to the owners and utilizes the rest to the guest users. Depending on the user congestion in the network, HMS selects a minimum capacity for the guest users. However, for higher user congestion, FAP serves more guest users with less resource. The network balances the total traffic with maximum resource efficiency in the functioning area. The selection of guest users for each FAPs is based on the level of subjected to co-tier interference due to strong FAP interferences. As most of the time, it is more likely to be an uneven distribution of users; the allocated ratio of resources for guest users in each FAP is independent from its neighboring FAPs. In some cases, the CGS users get such a geographical position where they get very low capacity due to path loss and other radio losses. Ensuring minimum level of service to that user will be costly for the system and the minimum level of service will be quite high leaving very little resources for the non-subscribers to share. In such cases, average capacity can be taken to ensure the minimum level of service. In this study, the individual CGS users’ performance is considered since the femtocells are operating in a large open space. The first channel allocation is very crucial in this study, since the capacity improvement and loss experienced by the users will depend heavily on the optimality of this process. When the FAP considers allocating a new RB to a user, all the neighboring FAP that have the same resource in use will suffer from interference. The FAP recognizes the interference signature from the network and intelligently blocks the infected channel at that FAP-end which is responsible for the highest interference to its neighboring FAP users. By introducing the environment perception and interference recognition techniques, the complicated FAP interference can be managed in an effective way. After a time interval, the blocked channel is assigned to a random user. If the gain is less than the loss, it will be deferred; otherwise it will go ahead with the resource allocation by scheduling in orthogonal resources. The user channel count in this configuration can be expressed as:
[
]
th th (6) D ( ccd ) = max ( N ch ) − ρ N th th where Dccd and ρ N are the number of channels per FAP
in a cognitive co-channel deployment and the probability of blocking N sub-channels in a particular time interval in that FAP, respectively. The interference signatures of the channels are easily obtainable through the channel information exchange between adjacent cells. The number of CSG and guest users under each FAP coverage are
np
and nq . p and q are the number of total
CSG and guest users in the targeted area. For users under a FAP’s coverage: (7)
C p ∝ Cq
where C p and Cq are CSG throughput and guest users throughput under the same FAP’s service, respectively. Assuming a hybrid access resource distribution constant k as below: k =
Cq
, { x ∈ k : 0 < x < 1}
(8)
Cp
The cells assign spectrum as RB. For a given time interval the transmission power to all the RBs are determined by the HMS. Now the throughput of a particular guest user can be expressed as: Cq =
δq N p + Nq
, δ q = D (thccd ) ∆f log 2 (1 + α SINR f , r )
(9)
where Cq , N RB , C RB , ∆f and α are the throughput of guest users, throughput of random users under FAP service, sub-carrier spacing and the constant for target Bit Error Rate (BER), respectively. The users provide continuous SINR feedback in the uplink for the assigned RB to its FAP. For the maximum throughput and resource efficiency, the ratio can be stated as: δq th ∑k = (10) th th max C p ( n p + nq ) th
f −1
th
f −1
Where C p > C q , n p = P − ∑ n p , nq = Q − ∑ nq and f i=1 i =1 is the number of active FAPs in the functioning area. Each FAP will allocate k resources to non-subscribers under its coverage so that a satisfactory level of service for the subscribers is maintained. For both tiers of users, if the assigned bandwidth is not high enough, users might experience poor service even if it gets better coverage. Therefore, any users who get better service from neighboring FAP, will not switch station. Hybrid access control mechanism in cluster based FAP deployments acts like an open access control mechanism with different levels of service. In a typical network, cell selection and handover events are performed based on received signal strength of the neighboring FAPs. The users prefer the FAP with better signal strength to trigger the hand-off mechanism. However, available bandwidth and interference are the two critical parameters
that affect the capacity of FAPs. Even though, the link quality is better, the FAP sometimes offers limited capacity to the users because of being overloaded. Neighboring FAPs with comparably worse signal strength may offer better capacity. In this cognitive co-channel deployment, capacity based cell selection is used over the link-quality based cell selection. The capacity maximization cell selection method is ∧ f = arg max C f , q f
{ }
(11)
where C f ,q denotes the capacity of the non-subscriber q with cell-F. IV.
SIMULATION ENVIRONMENT AND RESULTS Table 1 shows the system parameters that were used for the network model. For the duration of the snapshot, all users are assumed to be static so that the effects due to Doppler spread are neglected. Each FAP has a minimum coverage range of about 20 m from the access point and they are deployed with no coverage holes. Mobile users are associated initially with nearby FAPs in the coverage area using SINR. The simulations are event-based and follow the 3GPP standards. The results are an average of 100 independent simulations. Table 1 System Parameters System Parameters Experimental Area Number of FAP Number of active owners Number of active guest Range of FAP FAP Antenna height User equipment height Frequency Bandwidth Sub-carrier spacing FAP owners minimum required throughput FAP Transmit Power (APC / Fixed) Distribution time interval FAP owner arrival intensity Guest user arrival intensity White noise power density Modulation Scheme Number of Resource block Sub-carrier per resource block Resource block size BER CGS minimum capacity (cognitive)
Value/Range 100 m x 100 m 5 15 60 20 m 1m 1m 2 GHz 20 MHz 15 KHz 20 Mbps ≥ 6 dBm/ 10 dBm 100 0.5 1 -174 dBm/Hz 64-QAM 50 12 180 Khz 10-6 20 Mbps
The standard length of the cyclic prefix in LTE is 4.69 µs. Each sub-carrier can carry maximum data rate of 15 Ksps (kilo-symbols per second) and 64-QAM represents 6 bits per symbol meaning that 10 MHz can provide a raw symbol rate of 9 Msps or 54 Mbps. This enables the system to compartmentalize the data across subcarriers. The simulation was performed using two distinct power levels.
more consistent as FAPs intelligently allocates power in each channel. HMS considers the SINR of the users registered under each FAP and evaluates the transmit power level. Table 2 provides performance percentages. Table 2 Compassion of the throughout with and without APC under different deployment configuration Dedicated CoCognitive APC Channel Channel Co-channel Guest User (without 0% 8% 16% APC) Guest User (with APC) 7% 13% 19%
Fig. 4: Comparison of the deployment configuration without adaptive power control
Fig. 6: SINR performance on different deployment configuration
Fig. 5: Comparison of the deployment configuration using adaptive power control
The six plots in Figures 4 and 5 illustrate the performance of dedicated channel, co-channel, and cognitive co-channel in the cluster based deployment of 8 FAPs. The configuration performance is evaluated in terms of guest users (non-subscribers) and overall users. Spectrum reuse increases the gain for the co-channel deployments to overcome co-tier interference. The introduction of the cognitive radio features in the cochannel deployment ensures the best throughput level for guest users. The subscribers are served with a priority tag to provide a minimum performance level. Analogously, the dedicated channel configuration fails to provide a suitable service to all users due to its limited spectral efficiency. Co-channel maintains a higher throughput with channel gain seemingly higher than the induced cochannel interference. Capacity based cell selection in the cognitive co-channel configuration constantly maintains a higher throughput giving significantly improved performance. At the same time its intelligent allocation of resources provides higher average throughput for the guest users whilst maintaining nominally standard quality coverage for the CSG users. Adaptive Power Control (APC) minimizes inference in a dense network. Figure 5 plots the channel deployment performance using APC. Predictably the throughput increases in all configurations and the capacity found was
Figure 6 plots the SINR performance of the deployment configurations and predictably, the dedicated channel is outperformed. The principle observation however, is the performance of cognitive co-channel compare to the co-channel configuration. Selection of FAPs based on higher capacity leads to a connection that is subjected to poor link gain and higher interference. Even though having a lower signal level, the increasing number of assigned channels from the serving FAP enhances the throughput level, thus ensuring more efficient usage of the limited available resources. V. CONCLUSION This paper evaluated the performance of dedicated and co-channel deployment in cluster based FAP scenarios with the help of channel capacity formulation and simulation. A femtocell architecture is proposed, that aims at cognitive resource allocation in co-channel deployment featuring capacity based cell selection. Based on the observation, the QoS deviation of the users within the cluster is balanced ensuring the maximum spectral efficiency of each FAP. This can motivate the use of cognitive co-channel deployments in cluster based FAPs for large indoor environments. It will also overcome the demand for additional macrocells for indoor coverage and allow the existing macrocell to offload user traffic through the FAPs backhaul. In future, the research will continue on spectrum sharing for cognitive femtocell deployments.
REFERENCES [1]
A. U. Ahmed, M. T. Islam, and M. Ismail, "A Review on Femtocell and its Diverse Interference Mitigation Techniques in Heterogeneous Network," Wireless Personal Communications, pp. 1-22, 2014.
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
P. Xia, V. Chandrasekhar, and J. G. Andrews, "Open vs. closed access femtocells in the uplink," IEEE Transactions on Wireless Communications, 2010, 9: 3798-3809. M. Zuair, "Development of an access mechanism for Femtocell Networks," Journal of Theoretical and Applied Information Technology, 2013, 51. A. U. Ahmed, M. T. Islam, R. Azim, M. Ismail, and M. F. Mansor, "Microstrip Antenna Design for Femtocell Coverage Optimization," International Journal of Antennas and Propagation, vol. 2014, p. 8, 2014. I. Demirdogen, I. Guvenç, and H. Arslan, "A simulation study of performance trade-offs in open access femtocell networks," IEEE 21st International Symposium on Personal, Indoor and Mobile Radio Communications Workshops (PIMRC Workshops), 2010:151-156. F. Bernardo, R. Agustí, J. Cordero, and C. Crespo, "Selfoptimization of spectrum assignment and transmission power in OFDMA femtocells," in Telecommunications (AICT), 2010 Sixth Advanced International Conference on, 2010: 404-409. J.-H. Yun and K. G. Shin, "Adaptive interference management of OFDMA femtocells for co-channel deployment," IEEE Journal on Selected Areas in Communications, 2011, 29: 12251241. H. Saad, A. Mohamed, and T. ElBatt, "A Cooperative Q-learning Approach for Real-time Power Allocation in Femtocell Networks," arXiv preprint arXiv:1303.2789, 2013. J. Lu, W. Zheng, T. Su, and X. Wen, "Interference Mitigation Spectrum Allocation for Energy Efficient OFDMA Femtocell Networks." International Conference on Computer Science and Network Technology (ICCSNT), 2012: 252 – 256. K. D. Nguyen, H. N. Nguyen, and H. Morino, "Performance study of channel allocation schemes for beyond 4G cognitive femtocell-cellular mobile networks," IEEE Eleventh International Symposium on Autonomous Decentralized Systems (ISADS), 2013: 1-6. L. Li, C. Xu, and M. Tao, "Resource Allocation in Open Access OFDMA Femtocell Networks." IEEE Wireless Communication Letters, 2012, 1(6): 625-628 N. Diaa El-Din, E. A. Sourour, K. G. Seddik, and I. A. Ghaleb, "Coordinated partial co-channel deployment in two-layer networks," International Conference on Computing, Networking and Communications (ICNC, 2013: 1162-1167. H. Zhang, X. Chu, W. Zheng, and X. Wen, "Interference-aware resource allocation in co-channel deployment of OFDMA femtocells," IEEE International Conference on Communications (ICC), 2012: 4663-4667.
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
M. C. Ertürk, I. Güvenç, S. Mukherjee, and H. Arslan, "Fair and QoS-oriented resource management in heterogeneous networks," EURASIP Journal on Wireless Communications and Networking, 2013: 1-14. A. U. Ahmed, M. T. Islam, M. Ismail, and M. Ghanbarisabagh, "Dynamic Resource Allocation in Hybrid Access Femtocell Network," The Scientific World Journal, vol. 2014, p. 7, 2014. X. Zhao, T. Lv, Y. Ma, and Y. Lu, "Joint Cognitive Access and Power Control in Two-Tie Femtocell Networks," Trustworthy Computing and Services, 2013: 279-286. N. Sung, J. P. M. Torregoza, W. Hwang, S. Lee, and H. Yoon, "A joint power control and converge scheme in a cognitivefemtocell architecture for wireless networks for throughput maximization," IEEE International Conference on Industrial Informatics (INDIN), 2010: 1025-1030. Q. Kuang, J. Speidel, and H. Droste, "Joint base-station association, channel assignment, beamforming and power control in heterogeneous networks," IEEE Conference on Vehicular Technology, 2012: 1-5. R. Xie, F. R. Yu, and H. Ji, "Energy-efficient spectrum sharing and power allocation in cognitive radio femtocell networks," INFOCOM, 2012 Proceedings IEEE, 2012: 1665-1673. S. Al-Rubaye, A. Al-Dulaimi, and J. Cosmas, "Cognitive femtocell," Vehicular Technology Magazine, IEEE, 2011, 6: 4451. R. Urgaonkar and M. J. Neely, "Opportunistic cooperation in cognitive femtocell networks," IEEE Journal on Selected Areas in Communications, 2012, 30: 607-616. J. P. M. Torregoza, R. Enkhbat, and W.-J. Hwang, "Joint power control, base station assignment, and channel assignment in cognitive femtocell networks," EURASIP Journal on wireless communications and networking, 2010: 6 F. Tariq, L. S. Dooley, and A. S. Poulton, "Virtual clustering for resource management in cognitive femtocell networks," in Ultra Modern Telecommunications and Control Systems and Workshops (ICUMT), 2011 3rd International Congress on, 2011: 1-7. P. Kyosti, J. Meinila, L. Hentila, X. Zhao, T. Jamsa, C. Schneider, et al., "WINNER II channel models," European Commission, Deliverable IST-WINNER D, 2007, 1. E. S. Hassan, X. Zhu, S. E. El‐Khamy, M. I. Dessouky, S. A. El‐ Dolil, and F. E. A. El‐Samie, "Performance evaluation of OFDM and single‐carrier systems using frequency domain equalization and phase modulation," International Journal of Communication Systems, 2011, 24, 1-13. 3GPP Release 8 (http://www.3gpp.org/Release-8)
