WLAN hot spots to increase UMTS capacity

Transcription

1 WLAN hot spots to increase UMTS capacity Alessandro Bazzi WiLab, IEIIT-BO/CNR, DEIS-University of Bologna, V.le Risorgimento 2, 4136 Bologna, Italy. Abstract The seamless integration of heterogeneous wireless access networks is currently one of the issues of major interest in the field of telecommunication systems. In an early future technologies like UMTS, WLAN and WiMAX will give broadband access to users that perform their voice/video calls or data sessions without knowledge of the specific technology their terminals are using, following the paradigm of always best connected. In this scenario, UMTS and WLAN integration has a primary role, due to the coverage of cellular systems and the low cost of the latter technology. This work, in particular, focuses on the gain in terms of capacity that the deployment of WLAN access points allows in an area fully covered by an UMTS network, assuming that the radio resources are not managed by complex multi radio resource management algorithms. For this scope a real scenario is accurately simulated and the end users perception is evaluated taking care of all aspects related to each level of the protocol stack, with particular attention to radio propagation. I. INTRODUCTION The integration and interworking of different technologies, including among others UMTS, WLAN, WiMAX, and Bluetooth is the next step in the evolution of wireless systems. An increasing number of projects and standardization bodies are working on new architectural solutions, signalling protocols and multi radio resource management algorithms aimed at giving seamless connection to users through the most suited available technology, depending on the context. Among all the technologies, particular interest has been given to the integration of cellular systems (UMTS in particular) and WLANs; cellular systems guarantee a high level of coverage on the territory, while WLANs allow high bandwidth at low costs. UMTS and WLAN interworking has been largely studied for the last ten years, trying to solve a wide variety of challenging problems: How to gain this integration in terms of architectural solutions and signalling among infrastructure nodes in order to allow vertical handovers (i.e., handovers form one technology to the other), as discussed for example in [1]; which multi radio resource management (MRRM) algorithms maximize the network performance, as focused in [2]; how to solve problems that protocols at higher layers like TCP encounter during vertical handovers, as discussed in [3]. Differently from previous works on this topic, here we aim at investigating and quantifying the impact of WLANs hot spots deployed in a real urban scenario fully covered by an UMTS network; our study is related to a short term future implementation where the heterogeneous resources are not managed by a centralized control system able to implement advanced MRRM algorithms. Although not allowing to optimize the system performance, the absence of a controller defining to which technology should each terminal connect makes this scenario feasible in a shorter term due to the separate management of the two UMTS and WLAN networks. At the application level, this work will mainly focus on the voice service, since it is the most challenging for WLANs and still the most used by cellular phones owners. Numerical results will be provided by means of simulations performed through the dynamic simulation platform SHINE [4], which accurately models all the aspects of a real network: the territory map, the generation of session data packets, a realistic modeling of the signal propagation also taking into account time and frequency correlation of the fading due to multipath propagation. Particular attention will be given to the WLAN hot spots call admission control (CAC) algorithm, which is very challenging, especially when dynamic link adaptation on correlated variable channel and the hidden terminal problem are not neglected. The rest of the paper is organized as follows: In section II the reference scenario and the methodology of investigation are discussed; in particular, a description of the scenario will precede details on WLAN and UMTS implementation, and, finally on the considered figures of merit. In section III the adopted WLAN CAC is deeply discussed. In section IV, numerical results are illustrated before drawing the conclusions in section V. II. SCENARIO AND ASSUMPTIONS In this paper, the 3rd generation frequency division duplex (FDD) UMTS technology is assumed for cellular communications and the IEEE82.11e at MAC layer with the IEEE82.11a at physical layer are considered for WLANs. The level of integration of the two access networks is supposed to allow vertical handovers also for voice sessions without excessive service degradation or interruption; details on architectural solutions and signalling issues, although of primary importance, are not here discussed for the seek of conciseness, and the interested reader may refer, for example, to [1], [5] or [6]. All terminals are supposed dual mode (i.e. they support both technologies). A. Scenario The reference scenario is shown in Fig. 1: it represents a portion of the medium sized Italian city of Bologna, consisting in a rectangular area of the city center sized 1.8 Km (longitude) x 1.6 Km (latitude) with 35 UMTS cells covered by 15 Nodes- B(1, 2 or 3 cells per Node-B are assumed). An approximated

2 Fig. 1. Reference scenario (1.8 km x 1.6 km): WLAN APs with approximate coverage area and UMTS Nodes-B with antenna directions and approximate cells. area of coverage is depicted in Fig. 1 for each cell with random colors. Colored arrows represent UMTS antenna directions and maximum available power at the antenna connection. Five WLAN hot spots are considered in the scenario, represented in Fig. 1 with circles approximately corresponding to APs coverage area. Please note that deploying WLAN APs near to Nodes-B allows the sharing of site resources, with a reduction of operator s costs; for this reason three APs are located near to a Node-B and only two APs are located at the edge of UMTS cells. Users are randomly generated in the area, yet not uniformly: a higher density is considered where cells are smaller, thus reproducing the real operation conditions. Users randomly move at a variable speed modelled as a Gaussian random variable truncated between km/h and 3 km/h with 1.5 km/h in average and 1 km/h standard deviation, performing sessions with a duration that follows an exponential distribution of 9 seconds in average. As long as voice calls are considered, the AMR codec [7] at 12.2 kb/s is adopted; a 12.2 kb/s bearer is assumed when served by the UMTS system, with an activity factor of.5 in interference calculations. A voice to VoIP conversion is supposed in the case of transmission over WLANs (like in [1] and [6]); in this case the on-off model described in [8] is considered for voice detection. Silence suppression procedures are reproduced as detailed in [9], with full voice and SID (silence insert descriptor) packets generation; 4 bytes of RTP, UDP and IP overhead are added to packets. B. UMTS The FDD version of UMTS is hereafter considered, which is the most diffused 3G cellular technology, with a channelization bandwidth of 5 MHz in the 2 GHz band. An UMTS planning with a single frequency is hereafter assumed; this implementation is possible thanks to the WCDMA access technique supported by a 15 Hz fast power control and to soft/softer handovers. The Hokumura-Hata model for an urban macro cell environment [1] is considered for the UMTS path loss calculation (L UMTS [db] = log 1 (D), with the distance D expressed in kilometers); a spatially correlated log-normal shadowing with 8 db variance and a timely correlated fast fading due to multipath adopting Pedestrian A model [11] are added. Following the specifications, the power level is adjusted per each terminal in both directions 15 times per second and the signal to noise ratio is calculated taking into account all inter-cell and intra-cell interference. Intra-cell interference in downlink is reduced to 3% due to the orthogonality of the codes. Three fingers rake receivers and realistic antenna radiation diagrams (omnidirectional at the mobile terminals, 12 o at the nodes-b with 15 db of maximum gain) are adopted. The total transmitting power ranges from 9 to 6 dbw for mobile terminals while it is limited at the Nodes-B to a value between 1 and 13 dbw, depending on the considered cell (see Fig. 1); the power level used in a cell for each voice connection ranges between 63 and 16 db with respect to the total available power. Noise figures are set to 7 db at the mobile terminal and 3 db at the base station. Soft and softer handover are implemented, with a maximum active set of 3 cells. Three downlink common channels (common pilot channel CPICH, primary common control physical channel P-CCPCH timely multiplexed with synchronization channel SCH, and paging indicator channel PICH) are transmitted in all cells with a spreading factor (SF) equal to 256 and a power level equal to 23 db with respect to the total available power. The contribute to interference due to all other common channels is assumed negligible. The 12.2 kb/s voice traffic is transmitted through dedicated channels with a SF equal to 128 in downlink and 64 in uplink. Per each transport block (1ms in this case), the average BER is evaluated. If its value exceeds a threshold of 2%, then the transport block is said in outage. An user will be satisfied of the performed call if the outage periods have not been more than 5% of the total duration, in both downlink and uplink directions. C. WLAN In WLAN hot spots, the IEEE82.11e standard is considered, which introduces the concept of quality of service in the frame delivery procedure at the MAC layer of WLANs systems. The idea at the basis of the IEEE82.11e MAC is to manage traffic flows with different throughput/delay requirements by means of different queues, which are given different probabilities to gain the access to the channel: the more the requirements of the traffic assigned to a queue are stringent, the higher is the probability that the queue gains access to the channel. Only the contention-based access mode (enhanced distributed coordinated function EDCF), is considered here, since the polling based access mode (hybrid coordination function controlled channel access HCCA) is not always provided in actual implementations. Adopted

3 TABLE I IEEE 82.11E MAC LAYER PARAMETERS SETTING AIFS Min Max Short Retry Long Retry TxOp CW CW Limit Limit [ms] VoIP Data IEEE82.11e parameters for voice and data traffic classes are listed in Tab.I. Since IEEE82.11e specifications only define the MAC layer strategies, IEEE82.11a is considered at the physical layer. IEEE82.11a, with a channelization bandwidth of 2 MHz in the 5 GHz band, provides 8 combinations of convolutional channel coding and modulation, hereafter denoted as modes. The nominal data rate ranges from 6 Mb/s (code rate 1/2, BPSK modulation) to 54 Mb/s (code rate 3/4, 64-QAM modulation). The control messages (including request to send RTS, clear to send CTS, and acknowledgments ACK) shall always be transmitted adopting one of the three modes with coding rate 1/2; in our implementation, BPSK (code rate 1/2) is adopted for control messages if BPSK or QPSK (at any code rate) is used for data packets, QPSK for control messages if 16-QAM is used for data packets, and 16-QAM for control messages if 64-QAM is used for data packets. All physical layer aspects are considered in details, including time and frequency correlated channel models; since an urban area is considered, the ETSI channel C [12] and a path loss decaying with the third power of the distance are assumed (L WLAN [db] = log 1 (d), with the distance d expressed in meters). All terminals and APs transmit using an omnidirectional antenna with a power level set to 1 dbw; a noise figure of 1 db is assumed. Both two-ways and four-ways handshake methods are implemented and the hidden terminals and capture effect phenomena are not neglected. The MD-AARF (mode dependent adaptive auto rate fall back) [13] link adaptation algorithm is adopted, which does not require power measurements and is designed for scenarios with a large number of terminals. The quality of voice connections is assessed through the average packet loss. As motivated in [6], if more than 5% of packets are not correctly received with a wireless link delay lower than 3 ms in both directions, then the user is considered unsatisfied of the service. Voice packets waiting in the MAC layer queue for more than 3 ms are automatically discarded. D. Figures of Merit for the End User Quality of Service Assessment In order to evaluate the quality perceived by the end users the following merit figures are defined: call setup success rate (CSSR): it describes the ratio between the calls successfully connected to the system and the overall number of call attempts; CSSR = N attempt N block N attempt, (1) where N attempt is the number of call attempts and N block is the number of blocked calls (i.e. that will not perform their session); drop call rate (DCR): it is the ratio between the abnormal releases (i.e. dropped calls against the will of the user, due for example to poor radio conditions or congestions) and the overall number of call releases (that is both the normal and the abnormal releases); DCR = N drop N release, (2) where N drop is the number of dropped calls and N release is the number of call releases; outage rate (OutR): it is the ratio between the number of calls which although normally released perceived an unacceptable quality of service, and the overall number of call releases; OutR = N outage, (3) N release where N outage is the number of calls normally released but with an insufficient perceived quality; satisfaction rate (SatR): it expresses the degree of users that were satisfied of the service since they were not blocked, nor dropped, nor they perceived an insufficient quality of service; SatR = N attempt N block N drop N outage. (4) N attempt III. WLAN CALL ADMISSION CONTROL In order to enable the use of a WLAN network for voice service with acceptable quality, a call admission control (CAC) algorithm must be implemented at the APs, as highlighted for example in [6] and [14]. Due to the CSMA/CA mechanism, in fact, a new connection that starts in an almost saturated WLAN network not only gains an insufficient quality of service, but also degrades the quality of all active sessions. The critical aspect in the design of the CAC algorithm is that the maximum number of voice users to be admitted in a WLAN hot spot depends on many factors, including the configuration of the system parameters, the scenario and the position of the users, and the traffic generated by the active sessions. Since most of these factors vary in time, a dynamic algorithm is considered in this work, which follows the proposals of [15] and [16]. The algorithm is based on the monitoring by the AP of the channel occupation, through the assessment of the channel occupation rate parameter, C O [15]. C O is defined as the ratio between the amount of time the medium is busy, T B, and the related observation time ΔT : C O = TB ΔT. The evaluation of the channel occupation rate is particularly simple to be implemented in existing APs, owing to their carrier sensing capability: the AP simply adds the busy medium sensed time T SB to its transmission time T AP (T B = T AP + T SB ). In order to improve the accuracy of

4 the estimation of T B, also the mandatory idle times SIFS and AIFS are considered in the assessment of T AP and T SB. Hereafter we adopt the strategy denoted in [16] as two thresholds CAC that foresees the block of incoming users when the system is in a congestion state; the congestion state is started when C O exceeds a given congestion threshold C T and it ends when C O goes below a decongestion threshold D T, with C T >D T. In order to verify the effectiveness of the adopted CAC algorithm, simulations were carried out in the presence of users performing voice calls (voice users) and best effort data downloads (data users). Since no minimum quality requirements are assumed for data users, no admission control is considered for that class of service. Consequently, the C O assessment and the threshold comparisons are performed for voice users neglecting data traffic; please note that the channel is considered as busy not only when voice packets are transferred, but also when collisions or channel errors do not allow to understand the class of service of the packet. Differently from previous works, the provided algorithm is here tested taking into account the hidden terminals problem, that occurs when there are users transmitting to the same destination yet not able to sense to each others signal. The hidden terminal problem causes a severe degradation of the quality of service and some restrictions are needed in order to maintain it sufficient for voice users. In particular, the following two restrictions will be considered: 1) all packets are transmitted adopting the four-ways handshake, even if they are small (i.e., also in the case of voice packets and TCP acknowledgments); please note that the transmission time of an RTS packet (2 bytes MAC payload) is significantly lower than that of a full VoIP packet (73 bytes), a SID packet (47 bytes) or a TCP ACK packet (47 bytes); hereafter, this assumption will be denoted as always four-ways handshake; 2) the APs reject those terminals whose power level is not enough to exceed sensitivity of 3 db; this assumption implies that the hot spot radius is forcedly reduced from about 9 meters to about 7 meters; this assumption will be denoted as reduced coverage. In Figg. 2, 3, and 4, the effects of CAC on voice and data users performance is shown, comparing the adoption of either of the two described assumptions and of both of them to the case where they are not considered (full radius and two-ways handshake for packets smaller than 1 bytes are adopted in this case). Simulations are reported considering a single hot spot with voice and data users randomly placed; sessions randomly start in time and their duration follow an exponential distribution with average 9 seconds. All curves are depicted varying the congestion threshold C T ; D T is always equal to C T minus.1 (e.g., C T =.5 corresponds to D T =.4). The observation time ΔT is set to 25 ms. Figg. 2 and 3 show the average number of admitted voice users and their satisfaction rate, respectively. As it can be noted, the average number of admitted voice users linearly increase with C T, thus validating the implemented CAC Average number of active voice users Reference 5 Always four ways handshake Reduced coverage Both assumptions C T Fig. 2. Voice calls admitted in a single WLAN hot spot, varying C T. Voice users satisfaction (among admitted) Reference Always four ways handshake 1 Reduced coverage Both assumptions C T Fig. 3. Percentage of satisfied voice users among those admitted in a single WLAN hot spot, varying C T. Average TCP level throughput for data users [Mbps] Reference Always four ways handshake Reduced coverage Both assumptions C T Fig. 4. Average TCP level throughput of data users in a single WLAN hot spot, varying C T. algorithm. Moreover, focusing on Fig. 3, the need for both the introduced restrictions is highlighted; due to the hidden terminal problem, less than 8% of users among those admit-

5 ted are satisfied of the service if only one of the described assumptions is adopted and less than 6% if they are both not considered. Please note that results are given starting from a very reduced number of average active voice calls (also with 5 voice calls in average, referring to Fig. 2): this means that it is a problem related to the CSMA/CA mechanism rather than a consequence of the adopted CAC algorithm. Fig. 4 shows the average throughput perceived by the data users, which, as expected, decreases linearly as a function of C T. As a consequence of the constant density, 18 data users are active in average when the reduced coverage assumption is considered, while 3 data users are active in average in the other cases. Comparing the curves related to a full coverage, it can be noted that although the adoption of RTS/CTS in all transmissions increases the transmissions duration, it still allows almost the same data throughput performance of the other case; this is due to the significant reduction of the collision probability. On the contrary, when observing the reduced coverage cases, the reduction in collisions probability has a lower impact than the increase in transmissions duration, and the adoption of RTS/CTS for all packets reduces the average TCP throughput perceived by data users. Please remark that the always four-ways handshake assumption is still necessary in order to serve voice users. Following the discussed results, the described CAC algorithm with both restrictions will be hereafter adopted at all APs, with C T =.6, D T =.5, and ΔT = 25 ms. IV. WLAN AND UMTS INTERWORKING PERFORMANCE Once described the envisioned scenario and defined the parameters adopted for UMTS and WLAN configurations, here we investigate the effects of APs deployment, focusing on the voice service. As already stated, no centralized multi radio resource management is considered: each terminal by itself will choose to which technology to connect and when a vertical handover is needed. Reminding that all terminals are supposed dual mode, they all act following one of the two following opposite strategies: 1) terminals preferably connect to UMTS (UMTS first); 2) terminals preferably connect to WLAN (WLAN first). The case where only UMTS is available (UMTS only) isalso investigated for comparison. In order to limit the number of vertical handovers in ongoing sessions, terminals do not modify the access technology during a call, except if necessary. For example, a terminal with an active session through UMTS does not try a vertical handover as a consequence of the approaching to a WLAN AP adopting neither strategy. Obviously, in both UMTS first and WLAN first strategies the terminals try to switch to the other technology if blocked by the CAC algorithm of the preferred one or if an abnormal release occurs. The latter event may happen, for example, when a terminal performing a session through WLAN exits the AP coverage area; in this case it tries a vertical handover to UMTS to continue its session. SatR UMTS only.82 UMTS first WLAN first N av Fig. 5. Ratio of satisfied voice users in the reference scenario among those that require the service (SatR), varying the average number of potential voice calls. DCR / OutR UMTS only UMTS first WLAN first OutR DCR N av Fig. 6. Ratio of dropped voice calls and ratio of voice calls perceiving an insufficient quality of service in the reference scenario among those that are admitted (DCR and OuR, respectively), varying the average number of potential voice calls. The CAC algorithm adopted in the UMTS system is based on the codes and power availability, while the one described in section III is implemented at the WLAN access points. In Figg. 5 and 6, SatR, DCR and OutR (see section II-D) are shown, varying the number of average potential voice users N av. N av is obtained multiplying the number of voice call attempts by the average duration of each voice call (i.e., 9 s in the investigated scenario) and divided by the duration of the simulation; this value corresponds to the average number of voice calls that would be active in the scenario if none of them was blocked and no abnormal releases occurred. Observing Figg. 5 and 6, it can be noted that no advantage can be obtained by APs deployment if the terminals prefer to connect to UMTS. This is due to the fact that the probability that an user blocked or abnormally released in the UMTS system is in the coverage of a WLAN AP is very low. Let us highlight, in fact, that the total area covered by WLAN APs is small: all hot spot coverage regions are almost circular with a

6 Average active connections UMTS only UMTS first WLAN first Cell 14 Cell 15 Cell 16 AP 1 AP 2 Fig. 7. Average connections in some of the investigated cells and APs in the reference scenario, with N av 18. radius of approximately 7 meters, meaning an area of.154 km 2. It follows that the area covered by WLANs is less than 3% of the total considered region ( x1.6 =.267). On the contrary, if the dual mode terminals prefer to connect to WLANs, than a significant improvement can be obtained. With the WLAN first strategy most of users in the WLAN APs proximity will not connect to UMTS, thus not reducing the available resources and not generating interference to other connections in the cellular system. From Fig. 5 it can be noted that in this case the capacity is increased of more than 6 users in average with respect to the other two cases. To be noted that results are heavily impacted by the position of the WLAN APs: A deployment in a heavier loaded area means a higher probability that users are under WLAN coverage. Although an appropriate deployment is considered here, different scenarios may allow even higher advantages. In order to go deeper inside what is happening in the heterogeneous wireless network, the average occupancy for three selected cells and two APs is shown in Fig. 7 when N av 18. The occupancy is given in terms of average number of active connections; let us remind that, since during soft/softer handovers terminals are connected to more than one cell at the same time, one connection does not always correspond to one user in the UMTS system. Looking at the histograms, nearly 15 users served by each of the two WLAN APs in the WLAN first case allow a reduction in the UMTS cells of almost 1 average active connections. No significant variations, on the contrary, can be noted when the UMTS first strategy is considered; in this case, the average number of users served by WLAN APs is very low. V. CONCLUSIONS In this paper, the deployment of WLAN access points in an area with full UMTS coverage was investigated. The focus was on the advantage that can be obtained without advanced centralized multi radio resource management in an urban scenario, with particular attention to the voice service, which is still of primary interest for cellular systems and very challenging for WLANs technologies. It was shown that even considering a deployment covering less than 3% of the total area, an increase in the number of users that can be served with acceptable quality is possible. It was also highlighted that this result requires smart WLAN call admission control and only if the terminals equipped with both technologies take advantage of the WLAN access whenever available. ACKNOWLEDGMENT This work was carried out in the framework of the PEGA- SUS Project, funded by MSE. The author would like to thank Prof. O. Andrisano for motivating and supporting this research activity and G. Pasolini for helpful discussions. REFERENCES [1] A. K. Salkintzis, N. Passas, and D. Skyrianoglou, On the support of voice call continuity across UMTS and wireless LANs, Wireless Communications and Mobile Computing, vol. 8, pp , 28. [2] O. Yilmaz, A. Furuskar, J. Pettersson, and A. Simonsson, Access selection in WCDMA and WLAN multi-access networks, in IEEE Vehicular Technology Conference, 25. (VTC 25-Spring). Stockholm, Sweden: IEEE, May - June 25, pp vol. 4. [3] Y. Gou, D. Pearce, and P. Mitchell, A receiver-based vertical handover mechanism for TCP congestion control, IEEE Transactions on Wireless Communications, vol. 5, pp , October 26. [4] A. Bazzi, C. Gambetti, and G. Pasolini, SHINE: Simulation platform for heterogeneous interworking networks, in IEEE International Conference on Communications, 26 (ICC 26). Istanbul, Turkey: IEEE, June 26, pp [5] R. Ben Ali and S. Pierre, On the impact of soft vertical handoff on optimal voice admission control in PCF-based WLANs loosely coupled to 3G networks, IEEE Transactions on Wireless Communications, vol. 8, pp , March 29. [6] O. Andrisano, A. Bazzi, M. Diolaiti, C. Gambetti, and G. Pasolini, UMTS and WLAN integration: architectural solution and performance, in IEEE 17th International Symposium on Personal, Indoor and Mobile Radio Communications 25. (PIMRC 5), Berlin, Germany, September 25, pp vol. 3. [7] 3GPP TS 26.71, Mandatory speech codec speech processing functions; AMR speech codec; general description. [8] H. Stern, S. Mahmoud, and K. Wong, A comprehensive model for voice activity in conversational speech-development and application to performance analysis of new-generation wireless communication systems, Wireless Networks, vol. 2, no. 4, pp , December [9] 3GPP TS 26.92, Mandatory speech codec speech processing functions; adaptive multi-rate (AMR) speech codec; comfort noise aspects. [1] H. Holma and A. Toskala, WCDMA for UMTS, 3rd Ed. Wiley, 25. [11] ITU-R Rec. M.134-1, Requirements for the radio interfaces(s) for future public land mobile telecommunication systems (FPLMTS), January [12] J. Medbo, H. Andersson, P. Schramm, H. Asplund, and J.-E. Berg, Channel models for HIPERLAN/2 in different indoor scenarios, ETSI, Sophia Antipolis, France, Tech. Rep. Technical Report EP BRAN 3ERI85B, March [13] A. Bazzi, M. Diolaiti, and G. Pasolini, Link adaptation algorithms over IEEE82.11 WLANs in collision prone channels, in IEEE Vehicular Technology Conference, 26. (VTC 26-Spring)., Melbourne, Australia, May 26, pp vol.3. [14] S. Garg and M. Kappes, Can I add a VoIP call? in IEEE International Conference on Communications, 23 (ICC 23). Anchorage, Alaska, USA: IEEE, May 23, pp vol.2. [15] J. Gu, D. Zhang, A new measurement-based admission control method for IEEE82.11 wireless local area networks, in IEEE Personal, Indoor and Mobile Radio Communications, 23. (PIMRC 23). Beijing, China: IEEE, September 23, pp vol.3. [16] A. Bazzi, M. Diolaiti, C. Gambetti, and G. Pasolini, WLAN call admission control strategies for voice traffic over integrated 3G/WLAN networks, in IEEE Consumer Communications and Networking Conference, 26. (CCNC 6), Las Vegas, Nevada, USA, January 26, pp