Yali Wu , Guixia Kang ＊, Ningbo Zhang

1 Key Laboratory of Universal Wireless Communications, Beijing University of Posts and Telecommunications, Beijing 100876, China

2 Science and Technology on Information Transmission and Dissemination in Communication Networks Laboratory,Shijiazhuang 050081, China

* The corresponding author, email: gxkang@bupt.edu.cn

I. INTRODUCTION

With the rapid development of internet of things (IoT), machine-to-machine (M2M)communications, also known as machine-type communications (MTC), has been widely applied in various aspects [1]. Due to the diverse set of applications and services, three different aspects should be investigated in the design of future fifth generation (5G) system: (1) enhanced mobile broadband (eMBB) supporting both high data rate and low latency communications, and extreme coverage, (2) ultra-reliable MTC (uMTC) supporting network service with high reliability and low latency, and (3)massive MTC (mMTC) supporting the connectivity for increasing number of devices [2].In this paper, the goal is to provide insights for the design of mMTC services in a 5G setting.According to a report released by the 5G Infrastructure Public Private Partnership (5G-PPP)community, the next generation of wireless networks need to support more than 10,000 devices per square kilometer [3]. However, a large number of M2M user equipments (UEs)connected to the network within a very short period time will inevitably lead to the radio access network (RAN) overload and degrade the performance of existing human-to-human(H2H) communications [4]. There have been some orthogonal resource allocation methods to alleviate the RAN congestion and improve the access success probability [5-8]. Orthogonal multiple access (OMA) techniques, e.g.orthogonal frequency division multiple access(OFDMA), is adopted in the long term evolution (LTE) systems, where each resource block (RB) serves one UE to keep UE’s orthogonality. However, the continuous expansion of M2M UEs still results in a shortage problem of radio resources. OMA is not optimal in terms of spectral efficiency, and cannot achieve the system upper bound [9].

This paper studies a novel RA and resource allocation scheme for the coexistence of NOMA-based and OMA-based M2M communications.

In wireless communication networks, how to efficiently utilize the channel resource has become a critical issue due to the scarcity of spectrum [10,11]. Non-orthogonal multiple access (NOMA) is highly recommended to be involved in 5G systems, which can improve spectrum efficiency and network capacity by allowing multiple UEs to share the same radio resource in power domain at the transmitter with successive interference cancellation(SIC) at the receivers [12]. Theoretically,recent research has been proved that NOMA can achieve system capacity limit compared to OMA [13,14], thus making it a desirable enabling technology to facilitate M2M communications.

Random access (RA) is the key step for M2M UE to access the cellular network. An UE starts a RA by sending a preamble on the physical random access channel (PRACH) to Evolved Node B (eNB), and transmitting the connection setup request or data packet on the physical uplink shared channel (PUSCH) [15].OMA allows only one UE to use the PUSCH,we call this kind of RA as OMA-based RA.NOMA with SIC receiver permits multiple UEs sharing the same PUSCH, we call this kind of RA as NOMA-based RA. However,the performance of NOMA will be degraded if SIC receiver cannot perfectly cancel co-channel interferences, and may not satisfy the quality of service (QoS) of M2M communications in some scenarios. Therefore, the contributions of this work mainly focus on:

1) The coexisting RA (i.e., the coexistence of the NOMA-based RA and OMA-based RA)for M2M communications is proposed. A joint UE paring and power allocation algorithm is designed for the coexisting RA, which considers three criteria, i.e., the maximum transmission power constraints, minimum data rate constraints and the order of SIC in NOMA. In addition, the number of UEs for NOMA-based RA and OMA-based RA are determined by this algorithm.

2) To improve the number of successful data packet transmissions for the coexisting RA, all available radio resources are dynamically separated between NOMA-based RA and OMA-based RA. Furthermore, a reasonable resource tradeoff between PRACH and PUSCH is achieved.

The remainder of this paper is organized as follows. Section II discusses the resource allocation for the coexisting RA. Section III presents the joint UE paring and power allocation for the coexisting RA. Performance evaluations are given in Section IV, while the paper is concluded in Section V.

II. RESOURCE ALLOCATION FOR THE COEXISTING RA

The NOMA-based RA and OMA-based RA are detailed following, respectively. Different from OMA-based RA, when NOMA-based RA is selected, a UE group is first established,and the transmission power of each multiplexing UE is decided. The result of UE paring and power allocation is given by Section III.Denoteu1andu2as the number of UE groups(i.e., the number of group center UEs) for NOMA-based RA and the number of OMA UEs(i.e., the number of UEs for OMA-based RA),respectively. The value ofu1andu2are derived from Section III.

2.1 NOMA-based RA

When NOMA-based RA is selected, an access class barring (ACB) mechanism is adopted to control the number of UE groups. The ACB parameter determines the probability that the group center UE is allowed to access to the PRACH. The ACB parameter, denoted byR,is derived in Section 2.3. The eNB broadcasts the ACB parameter to all group center UEs. A group center UE randomly selectsxout of the uniform distribution between zero and one. If, the group center UE is allowed to access to the PRACH. Ifx＞R, the group center UE reattempts RA after a random back-off.The NOMA-based RA consists of four steps as follows:

Step 1: Preamble transmission. The allowed group center UE selects the specified preamble and transmits it on PRACH on behalf of its UE group. After that, the group center UE broadcasts the index of the specified preamble,path loss, power back-off factor and the list of non-group center UEs’ identity (ID) to the non-group center UEs.

Step 2: After detecting the specified preamble, eNB sends random access response(RAR) through downlink channel. The number of RARs equals the number of allowed UE groupsThe preamble information of these RARs is the specified preamble while the assigned PUSCH is different in each RAR.

Fig. 1 NOMA-based RA

Step 3: Power back-off and data transmission. The group center UEs will receiveRARs corresponding to the specified preamble. The group center UEs are sorted in accordance with the value of the group center UEs’ IDs. The group center UE randomly selects one RAR out of the available RARs in line until all theRARs are selected. The RAR, which is selected by one group center UE, cannot be selected by other group center UEs. The multiplexing UEs in the same UE group are expected to receive the same RAR.Upon receiving the RAR, the multiplexing UE adjusts the transmission power based on path loss and power back-off factor, then transmits both UE ID and data packet on the PUSCH which is indicated by the received RAR.

Step 4: SIC reception and acknowledge(ACK). As all the PUSCH for different UE groups are pairwise orthogonal, eNB could decode messages from each PUSCH independently. It performs SIC and decodes the data packet one by one on each PUSCH. After that, eNB sends the ACK message and corresponding UE ID via control channel. The NOMA-based RA is illustrated in Fig. 1.

2.2 OMA-based RA

To avoid the resource wasting caused by preamble collision, we extend the preamble transmission by transmitting UE ID and a cyclic redundancy check (CRC) information on a part of the subcarriers reserved for guard band [8].Division Multiple Access (CDMA) is used to distinguish each UE, i.e., the UE ID and CRC are encoded by gold sequence. Every preamble has a unique gold sequence. When more than one UE select the same preamble, their UE ID will be multiplied by the same sequence.When more than one UE transmit their IDs on the same subcarriers, eNB cannot decode UE ID. Then eNB thinks collision occurs and will not schedule PUSCH to the preamble, thus saving resource. The OMA-based RA consists of four steps as follows:

Step 1: An OMA UE randomly selects one preamble out of non-specified preambles with equal probability. The OMA UE transmits the preamble, UE ID and CRC on PRACH.

Step 2: After detecting the preamble, eNB begins to decode the UE ID. After successfully decoding the UE ID, eNB sends the corresponding RAR. The RAR conveys the preamble information and uplink resource grant for data packet transmission in Step 3. If eNB cannot decode the ID, this preamble is regarded as a collision and eNB will stop sending RAR.

Step 3: After receiving the corresponding RAR, the OMA UE adjusts the transmission power and transmits a data packet to eNB on the PUSCH indicated by the received RAR.

Step 4: If eNB correctly decodes the data packet transmitted in Step 3, it transmits an ACK message to the corresponding UE.

2.3 Resource allocation for the coexisting RA

Assuming that the total number of RBs allocated to PRACH and PUSCH in a RA cycle isQ, and the number of RBs allocated to PRACH isN. A PRACH consists of 6 physical RBs in a subframe, which occupies 864 subcarriers [15]. Assumingκpreamble sequences are mapped to the central 839 subcarriers while the rest 25 subcarriers are reserved for guard band in each RA cycle. Thereforepreambles are constructed, andNis integer multiples of 6. eNB chooses one of the preambles as a specified preamble, and the remainingpreambles as the non-specified preambles.

The number of RBs allocated to PUSCH isQ-N. Suppose the number of RBs constituting one subchannel issubchannels can be constructed.denotes the bottom integer function. Assuming an M2M UE usesRBs for a fix-sized data packet transmission. Since multiple UEs can share the same subchannel for data packet transmissions in NOMA-based RA, eNB will schedule enough subchannels for NOMA-based RA. In extreme case, the number of subchannels iswhenN=0. Therefore, the ACB mechanism is adopted to control the number of UE groups, and the ACB parameter is

Fig. 2 OMA-based RA

The expected number of UE groups selecting the specified preambleeNB will schedulesubchannels for NOMA-based RA, and schedulesubchannels for OMA-based RA.

DenoteMkas the number of multiplexing UEs on thek-th subchannelThen the number of successful data packet transmissions contributed by NOMA-based RA is

An OMA UE randomly selects a preamble out ofnon-specified preambles. This selection follows the Binomial distribution with meanDenoteas the probability of a successful preamble that is selected by one OMA UE.is given by

Considering the number of contending OMA UEs is very large, the number of non-specified preambles selected by one OMA UE follows the Binomial distribution. Therefore, the expected number of OMA UEs that select the successful preambles is

To maximize the uplink resource utilization, RBs should be allocated to PRACH and PUSCH such that the number of PUSCH is equal to the expected number of successful preambles. By this choice, the number of successful data packet transmissions would be equal to the number of successful preambles since sufficient PUSCH resources are available for assigning to successful preambles.DenoteN*as the optimal number of RBs allocated to PRACH, which is achieved when the gap between the number of successful preambles and the number of available subchannels is minimum.

Consequently, the number of subchannels scheduled to OMA-based RA isThe number of successful data packet transmissions contributed by OMA-based RA is

Therefore, the total number of successful data packet transmissions contributed by the coexisting RA is

III. JOINT UE PARING AND POWER ALLOCATION FOR THE COEXISTING RA

3.1 Uplink NOMA transmission

All the M2M UEs and the eNB are equipped with a single omnidirectional antenna. From Section 2.3,Mk(Mk＞1) UEs are paired as a UE group and transmit data packets on thek-th subchannel, whereandDenotemas the index for them-th UE. Letsk,mbe them-th UE’s signal transmitted on thek-th subchannel with.pk,mis the transmission power of them-th UE on thek-th subchannel. The received data packet on thek-th subchannel at eNB is

Wherehk,m = gk,mlk,mis the channel from them-th UE in thek-th UE group to eNB.lk,mis the large-scale path loss, andgk,mis Rayleigh fading coefficient.nkis the additive white Gaussian noise observed on thek-th subchannel at eNB. To simplify the analysis,lk,mis modeled by Free-Space path loss model [16],i.e.,whereGlis the product of the transmit and receive antenna field radiation patterns in the line-of-sight(LOS) direction, λ is the signal wavelength anddk,mis the distance between them-th UE in thek-th UE group and eNB.gk,mis modeled as zero-mean, independent and circularly-symmetric complex Gaussian random variables with variance μ. The probability density function (PDF)is

3.2 Uplink power control for NOMA

The difference betweenpk,mandpk,1can be expressed in watt by

PLk,mis modeled by Free-Space path loss model, and[18].Therefore, from (10),pk,mis given by

Wherem= 2,…,Mk.

3.3 Joint UE paring and power allocation for the coexisting RA

Fig. 3 illustrates a two-UE uplink NOMA system and gives the specific signal detecting process of SIC receiver.

Assuming each UE has its QoS requirement guaranteed by a minimum required data rate, which is denoted asOur design of UE paring and power allocation is based on providing QoS guarantees that each UE meets the corresponding rate requirement, i.e.,Substitutingandinto (15) andwe can get

Wherem= 1, 2,…,Mk-1.

Fig. 3 Illustration of a two-UE uplink NOMA system

Algorithm 1 Joint UE Paring and Power Allocation for the coexisting RA

Wherem= 1, 2,…,Mk-1.

IfMkincreases, the implementation complexity of SIC in uplink NOMA and the error propagation (EP) would be increase [19]. In order to keep the receiver complexity comparatively low, we consider the simple case where only two UEs can be allocated on the same subchannel. SubstitutingMk= 2 into (18) and(19), we can get

We focus on the following factors when proposing a joint UE paring and power allocation algorithm.

a) The order of SIC in NOMA is usually based on the descending order of the Rayleigh fading coefficient in (14).

b) The UEs can be paired together if the channel conditions of the UEs belong to the condition of (12), which satisfies the transmission power constraints of each multiplexing UEs.

c) The UEs can be paired together if the channel conditions of the UEs belong to the condition of (20), which satisfies the data rate constraint of each multiplexing UE.

Assuming there areNmactive M2M UEs for UE paring. The number of RBs allocated to thei-th subchannel isThe noise power on thei-th subchannel at eNB isThe target arrived power of the group center UE on thei-th subchannel ispi,u=pu. DefineX(Y) as theY-th element ofX. We consider the wireless channels are independent and identically distributed (i.i.d.) block Rayleigh fading,which means the fading channel gain is constant during a frame. Assuming that perfect channel state information (CSI) is available at eNB for each frame [20,21]. eNB can use the CSI for UE pairing and power allocation.Then eNB broadcasts the result of UE pairing and power allocation to UEs through downlink channel. The joint UE paring and power allocation for the coexisting RA consists of two phases as follows:

IV. SIMULATION RESULTS

In this section, simulations are conducted to illustrate and verify the performance of the number of successful data packet transmissions for the proposed RA and resource allocation scheme for the coexistence of NOMA-based and OMA-based M2M communications. To show the gain of the proposed scheme, the performance of OMA-based RA is also addressed. In the simulation,w=1,Gl=1,andλ=v/fc, wherevis the speed of light andfc= 2GHz. Two types of M2M services exist,each with the different target data rate and maximum uplink transmission power. The simulation parameters are given in Table I.

Denote the largest arrived SNR asL_SNR,and(dB), where the unit ofpuandis watt.

Table II shows the comparison of the number of successful data packet transmissions contributed by NOMA-based RA, OMA-based RA and the coexisting RA with different number of active UEs, respectively. According to the resource allocation for the coexisting RA in Section II, the ACB parameter isR=1.According to the UE paring and power allocation for the coexisting RA,u1andu2are determined. Sinceu1is the expected number of UE groups selecting the specified preamble. eNBwill scheduleu1subchannels for UE groups,therefore,. With the increase ofNm,significantly exceeds. This is because NOMA can use spectrum more efficiently by opportunistically exploring UE’ channel conditions.andcontribute toBs.

Table I Simulation parameters

Table II Comparison of, Bs with different Nm

Table II Comparison of, Bs with different Nm

?

Fig. 4 The comparison of the number of successful data packet transmissions between theoretical and simulation results for the coexisting RA

Fig. 4 presents the comparison of the number of successful data packet transmissions between theoretical and simulation results for the coexisting RA in a random RA cycle.We can see the rising trend of the number of successful data packet transmissions with the increased number of active UEs in a random RA cycle. In addition, the theoretical results also match the simulation results well.

Fig. 5 The comparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA

Fig. 6 The comparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA with different κ and

Fig. 5 illustrates the comparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA in a random RA cycle.We can see the number of successful data packet transmissions of the coexisting RA significantly outperforms the number of successful data packet transmissions of OMA-based RA. In addition, the average number of successful data packet transmissions of the coexisting RA of multiple RA cycles is presented in the figure. The number of successful data packet transmissions in a random RA cycle changes around the average result.

To better illustrate the performance of the coexisting RA, Figs. 6-8 are simulated based on the average number of successful data packet transmissions of multiple RA cycles.

Fig. 6 shows thecomparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA with differentκand. We can see that if, the number of successful data packet transmissions forκ=24,36,48 closes to each other. While ifκ=24, with the increase of,the number of successful data packet transmissions decreases.That is because more RBs are needed for transmitting a data packet whenincreases.This proves that the value ofhas more effect than the value ofκin the coexisting RA. Furthermore, with the increased number of active UEs, the number of successful data packet transmissions of coexisting RA significantly outperforms the number of successful data packet transmissions of OMA-based RA with differentκand.

Fig. 7 presents the comparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA with different target data rate. As expected,lower target data rate results in more number of successful data packet transmissions as it is more likely to be satisfied. With the increased number of active UEs, the number of successful data packet transmissions of the coexisting RA significantly increases, and exceeds the number of successful data packet transmissions of OMA-based RA with different target data rate. That is because NOMA can utilize resource more efficiently by opportunistically exploring UE’ channel conditions.

Fig. 8 shows the comparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA with different largest arrived SNR. With the increase ofNm, the number of successful data packet transmissions sharply increases with the increase of largest arrived SNR until it reaches a certain saturation point. The value of largest arrived SNR corresponding to the saturation point is optimal, since the number of successful data packet transmissions keeps constant even when the largest arrived SNR exceeds the optimal value. Furthermore,with the increase ofNm, the gap between the coexisting RA and OMA-based RA sharply increases with the increase of largest arrived SNR until it keeps constant.

V. CONCLUSIONS

This paper studies a novel RA and resource allocation scheme for the coexistence of NOMA-based and OMA-based M2M communications. The joint UE paring and power allocation algorithm for the coexisting RA is proposed, by which the number of UEs for NOMA-based RA and OMA-based RA are determined. The eNB dynamically changes the resource allocation between NOMA-based RA and OMA-based RA, as well as the resource allocation between PRACH and PUSCH according to the number of UEs for NOMA-based RA and OMA-based RA. Simulation results show the number of successful data packet transmissions is significantly improved by the coexisting RA than OMA-based RA. Therefore, the coexisting RA and resource allocation scheme in this paper provides an important insight for future M2M communications.

ACKNOWLEDGEMENTS

Fig. 7 The comparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA with different target data rate

Fig. 8 The comparison of the number of successful data packet transmissions between the coexisting RA and OMA-based RA with different largest arrived SNR

This work was supported by the National Natural Science Foundation of China (61501056),National Science and Technology Major Project of China (No. 2016ZX03001012), and the Research Fund of ZTE Corporation.

[1] 3GPP TR 37.868, V11.0.0, “Study on RAN improvements for machine-type communications,”O(jiān)ct. 2011.

[2] A. Osseiran, J. F. Monserrat, P. Marsch, 5G Mobile and Wireless Communications Technology.Cambridge University Press, 2016.

[3] The 5G Infrastructure Public Private Partnership,“5G PPP Use Cases and Performance Evaluation Models,” Apr. 25, 2016.

[4] M. Lvesque, F. Aurzada, M. Maier,et al., “Coexistence Analysis of H2H and M2M Traffic in FiWi Smart Grid Communications Infrastructures Based on Multi-Tier Business Models,” IEEE Trans.Commun., vol, 62, no. 11, pp. 3931-3941, 2014.

[5] W. Li, Q. Du, L. Liu,et al., “Dynamic Allocation of RACH Resource for Clustered M2M Communications in LTE Networks,” International Conference on Identification, Information, and Knowledge in the Internet of Things, pp. 140-145, 2015.

[6] Y. Pang, G. Lin, H. Wei, “Context-Aware Dynamic Resource Allocation for Cellular M2M Communications,” IEEE Internet of Things, vol. 3, no. 3,pp. 318-326, 2016.

[7] C. Oh, D. Hwang, T. Lee, “Joint Access Control and Resource Allocation for Concurrent and Massive Access of M2M Devices,” IEEE Trans.Wireless Commun., vol. 14, no. 8, pp. 4182-4192, 2015.

[8] N. Zhang, G. Kang, J. Wang,et al., “Resource Allocation in a New Random Access for M2M Communications,” IEEE Commun. Lett., vol.19,no. 5, pp. 843-846, 2015.

[9] H. Li, H. Liu, “An analysis of uplink OFDMA optimality,” IEEE Trans. Wireless Commun., vol. 6,no. 8, pp. 2972–2983, 2007.

[10] X. Li, Y. Shi, X. Wang,et al., “Efficient link scheduling with joint power control and successive interference cancellation in wireless networks,”SCIENCE CHINA Information Sciences, vol. 59,no. 12, pp. 1-15, 2016.

[11] N. Wang, S. Gong, Z. Fei,et al., “A QoE-based jointly subcarrier and power allocation for multiuser multiservice networks,” SCIENCE CHINA Information Sciences, vol. 59, no. 12, pp. 1-13,2016.

[12] Z. Wei, J. Yuan, D. Ng,et al., “A survey of downlink non-orthogonal multiple access for 5g wireless communication networks,” ZTE Communications, vol. 14, no. 4, pp. 17-25, 2016.

[13] Y. Yuan, X. Zhao, “5G: vision, scenarios and enabling technologies,” ZTE Communications, vol.13, no. 1, pp. 3-10, 2015.

[14] X. Chen, A. Benjebbour, A. Li, “Multi-User proportional fair scheduling for uplink non-orthogonal multiple access (NOMA),” IEEE VTC Spring,pp. 1-5, 2014.

[15] 3GPP TR 36.211, V12.2.0, “Evolved universal terrestrial radio access (E-UTRA): Physical channels and modulation,” Jun. 2014.

[16] A. Goldsmith, Wireless Communications. Cambridge, U.K.: Cambridge Univ. Press, 2005.

[17] N. Zhang, J. Wang, G. Kang,et al., “Uplink Nonorthogonal Multiple Access in 5G Systems,” IEEE Commun. Lett., vol. 20, no. 3, pp. 458-461, 2016.

[18] 3GPP TR 36.213, V10.12.0, “Physical layer procedures,” Sep. 2015.

[19] A. Benjebbour, Y. Saito, Y. Kishiyama,et al.,“Concept and practical considerations of nonorthogonal multiple access (NOMA) for future radio access,” IEEE Intell. Signal Process. Commun. Syst., pp. 770–774, 2013.

[20] X. Ge, J. Chen, C. Wang,et al., “5G green cellular networks considering power allocation schemes,” SCIENCE CHINA Information Sciences, vol. 59, no. 2, pp. 1-14, 2016.

[21] W. Han, Y. Zhang, X. Wang,et al., “Orthogonal Power Division Multiple Access: A Green Communication Perspective,” IEEE Journal on Selected Areas in Communications, vol. 34, no. 12, pp.3828-3842, 2016.