Consecutive Group Paging for LTE Networks Supporting Machine-type Communications Services

Transcription

1 Consecutive Group Paging for LTE Networks Supporting achine-type Communications Services Ruki Harwahyu +, Ray-Guang Cheng +, and Riri Fitri Sari ++ + Dept. of Electronic and Computer Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan ++ Dept. of Electrical Engineering, University of Indonesia crg@mail.ntust.edu.tw Abstract Group paging is one of the mechanisms proposed to resolve the radio access network (RAN) overload problem resulted from simultaneous channel access from huge machine devices [1]. This paper presents a consecutive group paging to enhance the performance of group paging. In consecutive group paging, the base station reserves multiple paging cycles to page the same group of users consecutively such that which failed in the earlier paging cycles can perform random-access procedure again in the later paging cycles. We utilize the analytical model presented in [2] to derive the performance metrics of consecutive group paging. The optimal group size, total number of reserved radio resource and the maximum number of preamble transmission are derived for each value of paging cycle threshold. The accuracy of the proposed analytical model is verified via computer simulations. Simulation results demonstrate that the proposed consecutive group paging enhances the performance of group paging. Keywords consecutive group paging, group paging, randomaccess, machine type communication, overload control I. INTRODUCTION achine-type communications (TC), or machine-tomachine communications (2), is a new service defined by 3 rd Generation Partnership Project (3GPP) to facilitate machines communicating each other over current cellular networks [3]. The TC enables wide range and improved applications such as smart grid, public security monitoring, environment monitoring, smart building, as well as consumer electronic devices. Simultaneous access of the random-access channels (RACHs) by a large number of user equipment (UE) in a cellular network may overload the network and severely degrade with the service quality of existing human-to-human communications (H2H) services. Several radio access network (RAN) overload control schemes are proposed to protect H2H services from being affected by 2 traffic [4]. Paging and group paging are two RAN overload control schemes proposed in 3GPP. Current paging mechanism is originally designed for H2H services. A base station (BS) can only page up to 16 with a single paging message, and only two paging occasions are available per 10 ms radio frame [1]. Therefore, a base station, which is referred as enhanced node B (enb) in LTE, must transmit multiple paging messages over a long period to activate a large number of. In group paging, an UE is assigned by a unique group identity (GID) after camping on a network and joining a group. All of the in a group listen to the same paging channel at the same paging occasion derived from the GID. The group of shall simultaneously perform the standard LTE randomaccess procedure to access the network when they find their GID in a group paging message. The which randomaccesses are failed shall follow the standard LTE random backoff procedure to retransmit their random-access attempts during a paging access interval until the retry limitation exceeds. Note that the network may use the group paging message to notify the paging access interval and the dedicated random-access resources reserved for group paging. An important work has been addressed by [2] for assessing the performance of group paging in LTE network. Some analytical models are presented to derive the performance metrics of collision probability, access success probability, average access delay, statistics of preamble transmissions, statistics of access delay, and utilization of random-access opportunities (RAOs) for group paging with various combinations of group sizes and reserved radio resources in a paging access interval. While the pure group paging has low resource utilization for smaller group size, [5] proposes dynamic resource allocation to increase resource utilization, especially in smaller group size. This work extends the model that has been elaborated in [2] by implementing consecutive group paging. In consecutive group paging, the enb reserves more than one paging cycle consecutively for paging one group, providing more chance for failed UE in the earlier paging cycles to conduct random-access procedure again in the later paging cycles. This method increases the resource utilization for larger group size. In this paper, we present a consecutive group paging to increase the access success probability of TC traffic. We extend the analytical model presented in [2] to analyze the performance of consecutive group paging and optimize the setting of the design parameters subject to a given quality-ofservice (QoS) constraint. The rest of this paper is organized as follows: standard LTE random-access procedure is briefly discussed in Sec. II. The system model and the concept of consecutive group paging are summarized in Sec. III. The analytical models of the proposed consecutive group paging are then presented in Sec. IV. Sec. V shows the simulation results, which demonstrate the effectiveness of consecutive group paging and the accuracy of the analytical model. Conclusions and future work are drawn in Sec. VI. II. LTE RANDO ACCESS PROCEDURE The LTE random-access procedure consists of 4 stage, i.e. preamble transmission, random-access reply (RAR) reception, sg3 transmission and sg4 reception. In the 1 st stage, all that want to access the network randomly choose one preamble and transmit it to enb. In LTE, a paging cycle is divided into several radio frames. Each frame consists of T RA_REP sub-frames; one of them is assigned for random-access

2 (RA) slot. Preamble transmission can only be transmitted at RA slot. For the 2 nd stage, enb requires up to T RAR sub-frame to decode received preamble, then broadcasts RARs in W RAR sub- are frames, one RARR each sub-frame, informing which acknowledged and granted for resource. Each RAR contains up to N RAR acknowledgements. Hence, the total resource, N U UL, is equal to N RAR W RAR. An UE considers preamble transmission failure if it receives no RARR after waiting T RAR +W RAR R sub-frames. The failure may be caused by low (not detected) power transmission, detected but collided (more than 1 UE choosing 1 preamble) or not collided but not selected by enb, since enb will randomly choose N UL preambles to acknowledge when total successfully-decoded in preamble transmission choosee new preamble, increases its power and retransmit preamblee at immediatee RA preamble is more than N UL. UEE that failed slot after uniform back-off with window of W BO. Preamble transmission for each UE is limited to N PTmax times. In the 3 rd stage, after receiving RAR, UE sends sg3 to enb. sg3 is a scheduled message used for sendingg UE identity and the RRC connection request message [6]. Thee UE transmits sg3 in the synchronized resource indicated in RAR. Note that sg3 and sg4 are sent according to non-adaptive hybrid automatic retransmission request (HARQ) procedure. Each time UE sends sg3, it starts (or restarts) contention resolution timer, T CR. In some cases, enb may successfully decode 1 preamble transmitted by multiple and replies a RAR. These will transmit their own sg3 on the same resource and then realize the random-access failure afterr T CR expired. The enb waits for sg3 up to T HARQ Q sub-frames after sending RAR. It sends ACK if it receives sg3 or sends NACK if no sg3 received until T HARQ elapses. After receiving NACK, UE retransmits sg3 T 3 subafter frames later. UE goes back to the 1 st stage iff it still failed sending sg3 N HA ARQ times. UE waits for sg4 up to T A_4 4 sub- the frames if it receives sg3 s ACK. In the 4 th stage, the enb sends sg4 to the UE usingg same resource used for sg3 transmission. sg4 is a contention resolution message to decide which devices gains access for data transmission. After sending sg4, enb waits for ACK within T HARQ sub-frames. The random-access procedure for this particular UE is finished iff ACK is received. enb retransmits sg4 after T 4 sub-frames if no ACK is received. UE can send sg4 up to N HARQ times. UE goes back to the 1 st stage if it still failed after sending sg4 N HARQ times. III. SYSTE ODEL This work studies an LTE-Advance network that utilizes a consecutive group paging scheme to activate a group of stationary in a cell. enb consecutively reserves C max paging cycles for this group, as illustrated in Fig 1. A paging cycle consists of I max RA slotss with R preambles in eachh RA slot. The random-access procedure conducted in each paging cycle is according to standard random-accesss procedure in LTE. In group paging, only one paging cycle is allocated for a group. To start paging, a paging message is broadcasted by enb in dedicatedd downlink channel to thee entire cell. The paging message specifies which GID is being paged. All that have the same GID as the GID mentioned in paging message will activate itself. Any of them that want to access the network will conduct random-access procedure in uplink channel at immediate RA slot, i.e. the first RA slot on the paging cycle. Some thatt are still failed even after the paging cycle elapses cannot access the network. In consecutive group paging, by providing more than one paging cycle, failed from previous paging cycle should conduct random-access procedure again in the next paging cycle. All RACH parameters are kept constant for all paging cycles. c Some may end up not able a to access the network at a the end of Cmax paging cycles if the group size,, is too large. IV. ANALYTICAL ODEL Let c be the total contending in cth paging cycle and c,i be the total contending UE at ith RA slot in cth paging cycle. Forr any value off c and i, theree possibly some of them are success and some of f them are failed. The total that are success in this RA slot, i.e. receivee RAR message, is c,i, S, meanwhile the total s that are failed in this RA slot is c,i, F. c,i,s [n] is the total success UE thatt transmitting their nth preamble at the ith RA slot in cth paging cycle. For a preamble to be successfully decoded by enb, it (1) should not be collided and (2) its transmission power should be high enough to be detected. For condition (1), based on [4], the success probability of c,i [n] among c,i that are contending for R RAOs can bee approximated by,,. For condition (2), let p n be thee preamble detection probability of the nth preamble transmission according to the power ramping effect. Inn LTE, p n is modeled as 1-( (1/e n ) [3]. However, when the total decoded preamble is more than what enb can acknowledge at that time, i.e. higher than N UL, enb will choose N UL preambles randomly and send RAR message to the whose preamble is being chosen. In thiss case, c,i,s [n] can be defined as,,,, if,,,,,,, otherwise. Subsequently, c,i,f[n] c can be defined as c,,i[n]- c,i,s [n]. Fig. 1 Consecutive paging cyclee consisting of Cmax m paging cycles The that receive RAR messagee will proceed to message part transmission which consists of sg3 and sg4 transmission. Each sg3 and sg4 transmission has failure probability of P f, and each of them can only be transmitted up to N HARQ times inn one attempt. UE should conduct preamble (1)

3 transmission again if in one attempt either sg3 or sg4 transmission exceeds N HARQ times. Hence, c,i [n] depends on the previous c,i,s [n] and also the error probability of message part transmission. We showed in [2] that the effect of the message part transmission error is negligible. Thus, c,i [n] can be expressed as if 1, 1, 1;, if 1, 1, 1;, α,,, 1 if 1, 1; 0 otherwise, where K min, K max and α k,i were given in [Eqs. (8), (9), (10), 2]. Performance etrics The collision probability, access success probability, average access delay, statistics of number of preamble transmissions, and statistics of access delay are chosen in this paper as the performance metrics to evaluate the performance of our proposed system. Collision probability, P C, is defined as the total collided RAO divided by total reserved RAO for a group in all paging cycle. Collided RAO are counted each time two or more UE uses the same preamble at the same frequency band at the same RA slot. For each RA slot in each paging cycle, total collided RAO is equal to the total RAO minus the total whose RAO is not collided minus the total idle RAO. Total reserved RAO equals to the RAO times the number of RA slot in each paging cycle times the total number of paging cycle. Hence, P S is expressed as,,, (3) Access success probability, P S, is the total UE that finish the whole random-access procedure at any paging cycle divided by the total contending UE at the first paging cycle. P S is expressed as,, (4) Average access delay,, is defined as the average time required by each successfully accessed to finish the whole random-access procedure. The delay is measured from the beginning of the first paging cycle. Let T c,i be the access delay for UE that transmit preambles at ith RA slot in cth paging cycle and complete the RACH procedure. T c,i includes waiting time for previous paging cycle (i.e. (c- 1) I max T RA_REP ), the time required to transmit preamble (i.e. (i-1) T RA_REP ), receive RAR (i.e. T RAR +W RAR ), and the average time required to successfully conduct the whole message part transmission,. and T i are expressed as (2),,,,, 1 _ 1 _ (6) where can be obtained from [Eq. (17), 2]. Remember that message part transmission is not limited by paging cycle duration since it uses dedicated channel. Hence, Eq. (5) can be used to estimate the delay for any value of 1 i I max. Let d be the access delay, which is the interval between the first random-access attempt and the RACH procedure completion for UE that successfully finish its RACH procedure. Let G(d) be the CDF of the access delay. Remember that d is the time interval relative to the very beginning of the paging and it can be at any paging cycle. For the successfully access that complete their RACH procedure at time d, the average time they complete the preamble transmission relative to the beginning of the paging cycle where it belongs to is (7) sub-frames or at the 1e th RA slot relative to the _ beginning of the paging cycle where it belongs to. Hence, G(d) can be estimated by,,,, Resource utilization, U, is defined as ratio between the total number of successfully accessed and the total number of reserved RAOs during the whole paging cycle for this group. U is expressed as,, V. SIULATION RESULTS The performance of the proposed consecutive group paging scheme is investigated through computer simulations. In the simulation, a group size of 10 to 1000 with no background traffic is considered. The same random-access parameters used in [2]. N PTmax = 16 is considered herein. To have a fair comparison, the design parameters of R and N PTmax, are adjusted such that the total number of RAOs used in the consecutive group paging are identical. For simplicity, W BO was fixed in all simulations. Two scenarios are investigated. Scenario I compares the performance of group paging and two examples of (5) (8) (9)

4 consecutive group paging. The performance metrics of collision probability [3], access success probability [3], statistic of access delay [3], average access delay and resource utilization are shown in Figs. 2 to 6, respectively. Scenario II is designed to evaluate the effect of design parameters on the QoS constraint of P S and the result is shown in Fig 7. From Fig. 7, enb can properly select R, N PTmax, and C max to ensure 90% of P S for different value of.. In scenario I, the effect of different C max to the performance metrics can be best observed when the other parameters are fixed as in the baseline case. The baseline case is a normal group paging case taken from [2] with C max =1, R=54, W BO =21 and the original N PTmax of 10 is adjusted to 16 to make it easier to find consecutive paging cases with equal total RAO. Remember that total RAO is equal to C max R I max, and I max depends on N PTmax and W BO. Thus, increasing C max can be done by decreasing R and keeping N PTmax and W BO constant, by decreasing N PTmax and keeping R and W BO constant, or by decreasing W BO and keeping R and N PTmax constant. For the given baseline, there are only 2 consecutive group paging cases to compare. The first case has higher C max and smaller R while N PTmax is fixed, i.e. C max =2, R=27, N PTmax =16. The second case has higher C max and smaller N PTmax while R is fixed, i.e. C max =7, R=54, N PTmax =3. In Fig 2 to 6, analytical results are shown in line while simulation results are shown in dot representing the average value of 1000-times simulation trial. Fig 2 and Fig 3 show the preamble collision probability and access success probability respectively. As shown in both figures, keeping reasonable total RAO in increased C max is better conducted by decreasing N PTmax, i.e. case 2, rather than decreasing R, i.e. case 1. In case 1, smaller R yields more preamble collision and decrease total success UE in each RA slots. In case 2, smaller N PTmax decrease total success UE in each paging cycle, but with higher C max, failed can conduct random-access again in the later paging cycles, resulting higher success probability. The CDF of access delay and average access delay for successfully accessed UE are shown in Fig 4 and 5 respectively. Note that the CDF is compared for =500, when all cases has equal number of successful UE. In Fig 4 and 5, case 1 yields higher access delay than case 2. The paging cycle in case 1 lasts longer than in case 2 because of higher N PTmax (higher I max ). However, in case 1, with smaller R, more conduct random-access in later RA slots, and having longer access delay when they finish. In case 2, although some also conduct random-access in later paging cycle, the delay is smaller since 7 paging cycles in case 2 is shorter than 2 paging cycles in case 1. The resource utilization is shown in Fig 6. In this figure, case 2 has better resource utilization for >700. In larger, case 3 has more successfully accessed UE compared to the other cases. On the contrary, case 1 yields worse resource utilization although more paging cycle is allocated compared to the baseline case due to prone preamble collision in each RA slot. In Scenario II, the analytical results of consecutive group paging for various. C max, R, and N PTmax are shown in Fig 7. This figure shows surfaces indicating the minimum R to ensure at least 90% access success probability (P S =0.9) as the function of N PTmax and for various C max with constant W BO of 21 and N UL of 15. This figure demonstrates that the system can provide 90% access success probability with fewer R at the bigger C max. Note that some peak values are not shown in C max =2 and C max =3 since there it is impossible to provide 90% of access success probability at those particular condition. In those cases, although infinite R is reserved in each RA slot, the number of dedicated channel for message part transmission is limited to N UL. In the best case, the total UE that can finish the RACH procedure in every RA slot is equal to N UL. Hence the access success probability would be equal to N UL C max I max /. For instance, it is impossible to obtain P S 0.9 for any value of R as long when C max =2, =1000, N PTmax =7 because N UL =15. Fig 7 can be used to obtain the proper values of C max and N PTmax to minimize R to ensure 90% success probability when there are contending for access in the cell. VI. CONCLUSIONS AND FUTURE WORK This paper proposes a consecutive group paging scheme and an analytical model to optimize the setting of design parameters. The performance metrics of collision probability, access success probability, average access delay, CDF of access delay, and resource utilization of the consecutive group paging scheme are derived. In the simulation, we demonstrate that the consecutive group paging may achieve a better performance than that of group paging if the total number of RAOs used by the two schemes are identical. We also demonstrate that the proposed model can accurately predict the performance metrics of the consecutive group paging scheme. However, we find that the proposed consecutive group paging may not always achieve a better performance than that of group paging. It is because that we assume that a constant value of R is used in all paging cycles in the consecutive group paging. Therefore, one of the future works is to develop a dynamic resource allocation algorithm for consecutive group paging such that we can assign the optimal value of R in each of the paging cycle of the consecutive group paging. REFERENCE This work was supported in part by the National Science Council, Taiwan under Contracts, NSC E and NSC E Y3. REFERENCE [1] 3GPP R , "Pull based RAN overload control", Huawei and China Unicom, RAN2#71, August [2] Chia-Hung Wei, Ray-Guang Cheng, and and Shiao-Li Tsao, "Performance Analysis of Group Paging for achine-type Communications in LTE Networks," IEEE Transactions on Vehicular Technology, vol. PP, no. 99, arch [3] 3GPP TR , "RAN improvements for machine-type communication", v.1.0.0, August [4] Chia-Hung Wei, Ray-Guang Cheng, and and Shiao-Li Tsao, "odeling and estimation of one-shot random access for finite-user multichannel slotted ALOHA systems," IEEE Commun. Letter, vol. 16, no. 8, pp , August [5] Chia-Hung Wei, Ray-Guang Cheng, and Firas Al-Taee, "Dynamic Radio Resource Allocation for Group Paging Supporting Smart eter

5 Communications," in IEEE International Conference on Smart Grid Communications, Tainan, 2012, pp [6] ing-yuan Cheng, Guan-Yu Lin, Hung-Yu Wei, and Alex Chia-Chun Hsu, "Overload control for achine-type-communications in LTE- Advanced system," IEEE Communications agazine, vol. 50, no. 6, pp , June P C D A Fig. 5 Average access delay, Fig. 2 Preamble collision probability, P C U 5 P S Fig. 3 Access success probability, P S Fig. 6 Resource utilization, U 1 CDF Fig. 7 inimum R to ensure 90% of P S for various C max, N PTmax and for W BO = access delay (subframe) Fig. 4 CDF of access delay, G(d)