Flexible Uplink MU-MIMO Scheduling in Unlicensed Spectrum

Boqi Jia, Ting Zhou＊, Honglin Hu Yang Yang, Zhenhong Li, Sandrine Boumard

1 Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai, China

2 Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai, China

3 Shanghai Research Center for Wireless Communication, Shanghai, China

4 TCL Communication Technology Holdings Ltd. Shanghai R&D Center, Shanghai, China

5 VTT Technical Research Centre of Finland Ltd, Oulu, Finland

6 University of Chinese Academy of Sciences, Beijing, China

I. INTRODUCTION

The fifth-generation (5G) wireless communication system is designed to resolve several unprecedented technical challenges, which has attracted growing interests from both academic and industrial communities [1-3]. Based on the worldwide research, a broad consensus on the vision of the 5G wireless network was reached, which is defined by three usage scenarios and a group of key performance indicators (KPIs) [4]. The key usage scenarios depicted by International Telecommunication Union (ITU) include enhanced mobile broadband (eMBB), massive machine type communications (mMTC) and ultra-reliable and low latency communications (URLLC). A group of KPIs indicate that the technologies of 5G systems are more practical and comprehensive than those of IMT-Advanced systems, which mainly focuses on the KPI of data rate.

Though the key techniques of 5G wireless network in licensed spectrum, such as massive multiple-input multiple-output (Massive MIMO) [5] and mmWave [6], can improve spectrum efficiency considerably as one of the most important KPIs, the need for more licensed spectrum is never-ending [7]. With the rapid development and the increasing demand for wireless communications, the scarcity of spectrum resources, especially the licensed spectrum resources, has become an important challenge for wireless communication. Therefore, it is imperative for 5G wireless network to consider more spectrum to meet the considerable increasing of mobile data traffic and connections [8-9]. According to the existing spectrum planning, 5G wireless network sharing with unlicensed systems is being deeply studied from the view of technique and management.

In this paper, we investigate the uplink multi-user multiple-input multiple-output(UL MU-MIMO)technique to achieve multiplexing gain in unlicensed spectrum.

Different from licensed spectrum, unlicensed spectrum is open to various radio access technologies (RATs) as long as they abide by the unlicensed regulations. Among all the unlicensed systems, wireless local area network (WLAN) has been the most popular one.In the last decade, unlicensed spectrum has been used to offload data traffic via WLAN systems. Besides, long term evolution license assisted access (LTE-LAA), which uses the LTE radio communications technology in unlicensed spectrum, enables the mobile operators to make more efficient use of unlicensed spectrum. Furthermore, the 3rd generation partnership project (3GPP) is considering the 5G new radio (NR) access to unlicensed spectrum[10]. With the support of the state-of-the-art medium access technology in cellular systems,unlicensed spectrum can be utilized efficiently[11]. Especially, the available unlicensed spectrum at 5 GHz is fairly broad.

Besides the unlicensed spectrum, MIMO is a key technique to tackle the challenge of the growing data traffic. The use of multiple antennas at both the transmitter and receiver sides achieves a significant improvement in system capacity dramatically [12]. However,the limited size and cost of user equipment(UE) and the complexity of implementing multiple antennas have posted great difficulty to the realization of MIMO in uplink wireless communication systems. The uplink multi-user MIMO (UL MU-MIMO) or virtual MIMO technique is an alternative efficient approach to this limitation [13-16]. In UL MU-MIMO systems, multiple UEs, each with single antenna, are scheduled to be paired and transmit data at the same time-frequency resource blocks (RBs). Thus, this technique can provide the MIMO benefits for the uplink systems while mitigating the unpleasant restriction.

User pairing in UL MU-MIMO systems is a critical issue, and a well-designed user pairing scheme helps to achieve a multi-user diversity gain for a significant performance improvement. In addition, in conventional UL MU-MIMO systems, the number of the scheduled UEs is fixed, and after receiving the scheduling grant from base station (BS),the scheduled UEs transmit data to the BS immediately. There are many literatures studied the UL MU-MIMO in licensed spectrum[17-19]. The authors in [17] studied the fair UL MU-MIMO scheduling and proposed a double proportional fair algorithm based on fairness and spectrum efficiency (SE), which is the main issue researched in UL MU-MIMO systems in the beginning. Low-complexity scheduling with maintaining SE and fairness performance was investigated in [18],while the tradeoff between SE and energy efficiency (EE) in UL MU-MIMO systems was researched in [19], where complexity and EE are the focused issues in UL MU-MIMO systems recently. Furthermore, the authors in[20] studied the throughput and delay performances based on carrier sense multiple access and collision avoidance (CSMA/CA) mechanism in MU-MIMO WLANs. In [21], a joint access point (AP) selection and UE scheduling problem was investigated as an auction game in UL MU-MIMO WLANs. Though the UL MU-MIMO has been investigated in licensed systems and WLANs, there are few literatures researched the UL MU-MIMO issues for cellular network in unlicensed spectrum.

In unlicensed spectrum, user pairing algorithms proposed in licensed spectrum also can be adopted. However, besides user pairing problem, there is another critical issue. Dif-ferent from the deterministic UL MU-MIMO scheduling in licensed spectrum, the UL MU-MIMO transmission in unlicensed spectrum should comply with the listen-before-talk(LBT) operation, where a UE applies a clear channel assessment (CCA) check before using the channel. The CCA utilizes at least energy detection to determine the presence or absence of other signals on a channel in order to determine if a channel is occupied or clear, respectively. In unlicensed systems, the scheduled UEs need to do LBT operation after receiving the scheduling grants from BS to occupy the channel. Unfortunately, each UE may fail to occupy the channel due to CCA failure or collision especially when the channel is busy, i.e., other techniques like WLANs are sharing the unlicensed channel. Furthermore, in order to exploit the multiplexing gain of the channel and achieve higher system throughput, BS should schedule as many UEs as UL MU-MIMO channels support. Doing LBT operation before access the channel is the usage rule of unlicensed spectrum. However,due to LBT operation in unlicensed spectrum,if the number of the scheduled UE just equals to the independent MIMO channel number,then once any UEs fail to occupy the channel,the channel capacity is wasted, i.e., failed LBT has a major impact on the access in UL MIMO systems. Thus, conventional deterministic UL MU-MIMO scheduling scheme cannot fully exploit the multiplexing gain in unlicensed spectrum and may degrade the system throughput especially when the unlicensed band is busy.

In this paper, we propose an idea that adjusting the number of the scheduled UEs flexibly in unlicensed spectrum and design an adaptive scheduling algorithm to maximize throughput in unlicensed UL MU-MIMO systems. Specifically, the system model and problem description for UL MU-MIMO in unlicensed spectrum are given in Section II. The proposed flexible scheduling scheme and the corresponding adaptive scheduling algorithm are presented in Section III. In Section IV, the simulation results and performance comparisons are provided, followed by the conclusions in Section V.

II. UPLINK MU-MIMO IN UNLICENSED SPECTRUM

Fig. 1 Coexistence scenario among several unlicensed systems

As illustrated in figure 1, an MU-MIMO system with single BS is considered, which can be an LTE-LAA eNB or a 5G gNB. In the uplink, the BS serves several UEs by sharing the unlicensed channels with other unlicensed systems, for example, WLANs or other cellular systems in unlicensed spectrum which are located in the coverage of the BS. An uplink multi-user transmission procedure is considered that total U single-antenna UEs indexed from 1 to U have uplink data to send to the BS, which is equipped with M receive antennas, in the unlicensed channel l. It is noteworthy that our work can be extended to the multiple BSs scenario where each BS can work with the proposed schemes independently.

2.1 Uplink scheduling procedure

Fig. 2 Uplink scheduling procedure (a) in licensed spectrum (b) in unlicensed spectrum (c) in unlicensed MU-MIMO systems

In the conventional uplink wireless communication systems, each UE is scheduled under the control of BS to transmit data as illustrated in figure 2(a). Such a deterministic scheduling scheme can avoid the conflict among UEs, and particularly in wide-band communication systems. It helps UEs to select the most suitable time-frequency resource and achieves higher spectrum efficiency than accessing the channel by contention as WLAN does. However, there exists a serious problem if the deterministic scheduling scheme is adopted in unlicensed spectrum as described in figure 2(b). After the BS schedules UE to transmit data, the scheduled UE needs to do LBT to occupy the channel first. If the scheduled UE fails to occupy the unlicensed channel, then it will waste the resource and degrade the throughput performance. Especially, considering the UL MU-MIMO in unlicensed spectrum, the failed channel occupation will leads to the decreasing of the multiplexing gain. For example as shown in figure 2(c), there are K UEs scheduled by BS, and each scheduled UE may fail to occupy the unlicensed channel such as UE 2 in the figure, which reduces the number of the active UEs and degrades the capacity performance.

2.2 UL MU-MIMO in unlicensed spectrum

In a conventional licensed system, when a scheduling event happens in the uplink, the BS equipped with M antennas chooses K single-antenna UEs to share the same time-frequency RBs. Hence, an uplink multi-user MIMO system is constructed with K virtual transmitting antennas and M receiving antennas.

In the M×K UL MU-MIMO systems, each UE transmits a data stream xiwith powerand when applying the zero forcing (ZF) decoder to decouple the K data streams [22], the received signal y collected at the unlicensed BS can be detected by the pseudo-inverse of the channel matrix, i.e.,where H ∈ℂK×Mdenotes the channel matrix from each UE to all M antennas of the unlicensed BS, and we assume a Rayleigh quasi-flat fading channel in this channel model.n∈ℂM×1is the additive white Gaussian noise(AWGN) with zero mean and covariance matrixThe ZF receiver converts the joint decoding problem into K single stream decoding problems thereby significantly reducing receiver complexity. Multiplying the received signal vector y on the Left Hand Side byK decoupled sub streams is obtained with output SNR of the ith stream given as

Where [A]iidenotes the ith diagonal element of matrix A. Thus, the throughput of the K-input M-output UL MU-MIMO system combined by K streams can be written as

In uplink unlicensed systems, the unlicensed BS with M active antennas will selects K from U UEs to construct a K×M UL MU-MIMO at every uplink scheduling interval. However, unlike the conventional licensed system, all selected UEs need to do LBT operation first. The UEs which successfully occupy the channel will be permitted to transmit data to the unlicensed BS. We denote the number of the UEs which successfully occupy the channel as K’. Thus, K － K’UEs are scheduled abortively, and actually K’UEs sent data to the BS. Therefore, a UL MU-MIMO system is actually constructed with K’transmitting antennas and M receiving antennas.

Considering the indeterminacy of the actual number of scheduled UEs, the actual achievable throughput of the K×M UL MU-MIMO system in unlicensed spectrum can be expressed as

where K’ ≤ K degrades the throughput due to the failed channel occupancy.

2.3 Coexistence analysis among different unlicensed systems

There are amount of literatures have studied the coexistence among unlicensed systems,which mainly focus on the coexistence between two typically kind of unlicensed systems, i.e., LTE-LAAs and WLANs [23-26]. In the WLAN systems, distributed coordination function (DCF) with the exponential backoff scheme is employed in the WLAN UEs(WUEs) to access the unlicensed channel. To guarantee the coexistence with the existing WLAN systems, an adaptive backoff scheme to access the unlicensed channel l has also been adopted in other unlicensed systems, e.g.,LTE-LAA or 5G NR.

In the scenario of this paper, UEs in the coverage of the unlicensed BS are coexistence with several WUEs and UEs of other unlicensed systems. Based on the throughput analysis in [27] and [28], the successful transmission probability for the WUEs on unlicensed channel l when considering the effect of other unlicensed systems can be given by

Accordingly, focusing on the effects of WLAN on unlicensed BS, the successful transmission probability for the UEs on the unlicensed channel l can be expressed as

III. FLEXIBLE UL MU-MIMO SCHEDULING

In this section, we discuss the UL MU-MIMO scheduling in unlicensed spectrum, and propose a flexible scheduling scheme to statistically offset the degraded throughput due to the failed channel occupancy.

3.1 Flexible scheduling scheme for unlicensed UL MU-MIMO systems

Compared to the conventional UL MU-MIMO system where the number of the scheduled UEs is fixed, the indeterminacy of the actual scheduled UEs in unlicensed systems leads to the decreasing of the throughput as described in Eqn. (3). Such a throughput gap, arising from coexistence between the unlicensed systems and WLANs, needs to be filled for unlicensed UL MU-MIMO systems. Therefore, we develop a flexible scheduling scheme where the number of the scheduled UEs, decided by the BS, is not fixed but can be dynamically adjusted according to the channel condition. The procedures of the flexible scheduling scheme is described in figure 3, after the BS gather the channel condition and scheduling request, the BS decides the number of the scheduled UEs,i.e., Kf, and KfUEs are scheduled to do LBT operation. The number of the scheduled UEs,Kf, can be larger than the fixed number in the deterministic scheduling scheme, i.e., Kf≥ K.

In the proposed scheme, the BS decides the number of the scheduled UEs, Kf, according to the channel condition of active UEs. Thus, we give a definition of channel condition first.

Definition 1.Channel busy ratio (CBR) is defined as the probability to occupy the unlicensed channel successfully. The CBR parameter is user-specific and is calculated over a period of time in past. Based on the successful transmission probability for the UEs, i.e.,CBR can be described as

Actually, the achievable average throughput is related to the number of the scheduled UEs and the CBR, i.e., Kfand β. The CBR is the key parameter decided by the access environment, In order to maximize the throughput,the number of the scheduled UEs should be optimized according to the foregone CBR.To investigate the relationship between the optimal number of the scheduled UEs and the CBR, we formulate an optimization problem which dynamically adjust the number of the scheduled UEs as (8) shown in the top at next page, where Koptis the optimal number of the scheduled UEs. In each uplink scheduling interval, KoptUEs are scheduled and on account of the channel access probability on channel l, i.e., βi(l), the number of UEs which actually transmit data is less than Kopt.

The difference between the deterministic scheduling scheme and the proposed flexible scheduling scheme in unlicensed spectrum is illustrated with an example in figure 4. In the first scheduling step, UE 1-4 are scheduled by deterministic scheduling scheme in which UE 3 fails to occupy the unlicensed channel,and the actual number of the scheduled UE is three which degrades the throughput. In the proposed flexible scheduling scheme, six UEs are scheduled, and though UE 3 and UE 6 fail to access the channel, there still have four UEs to transmit data to the BS.

Fig. 3 The proposed adaptive UL MU-MIMO scheduling scheme

Fig. 4 An example for various UL MU-MIMO scheduling method

3.2 Adaptive scheduling algorithm for flexible UL MU-MIMO scheduling

The formulated problem as described in Eqn.(8) is an NP-hard problem and even that it is complex to get the solution by exhaustive search method. In this subsection, we develop a low-complexity adaptive scheduling algorithm to get the optimal number of the scheduled UEs and achieve the maximum throughput in the unlicensed spectrum.

The idea of the algorithm comes from the concept of H-index [30], which is an author-level metric that attempts to measure both the productivity and citation impact of the publications of a scientist or scholar. A scientist has index h if h of his or her total Nppapers have at least h citations each and the other Np– h papers have ≤h citations each.There are total U UEs in the coverage of the BS, and because of the various locations and conditions, all of the UEs have diverse CBRs,which is complex for the BS to get the optimal number of the scheduled UEs. In the proposed adaptive scheduling algorithm, the H-index is used to represent the channel condition. If h of total U active UEs have at least h probability each that occupy the unlicensed channel successfully, and the other 1 – h of total U active UEs have ≤ h probability each that occupy the unlicensed channel successfully, then the H-index of this channel condition is h.

The pseudo-code of the proposed algorithm is described as Table 1, where the H-index of the CBRs is calculated first, and the optimal number of the scheduled UEs is acquired by step 6 to step 9. The essential H-index of CBRs, i.e., hbasis, is preset by the BS and if h ＜ hbasis, which means the whole channel condition in the coverage is too bad, then the optimal number of the scheduled UEs, i.e.,Kopt, needs to be increased to offset the capacity degradation. The increasing of Koptis stepby-step by using the CBRs which is updated by the step length of CBRs updating, i.e.,α. In brief, Koptwill be decided until the H-index of CBRs is not less than hbasis. Furthermore, the period of updating the optimal number of the scheduled UEs can be preset manually as T,which denotes the granularity of the proposed algorithm.

IV. NUMERICAL RESULTS AND PERFORMANCE COMPARISONS

In this section, we provide a numerical simulation to evaluate the effectiveness of the proposed algorithm and compare it with existed schemes as benchmarks. The simulation are evaluated with extensive MATLAB-based system level simulation tools. We assume a single unlicensed BS scenario equipped with 4 antennas for simplicity where total 100 UEs each equipped with single antenna are assigned with all RBs, and all the UEs are uniformly distributed in the coverage. Furthermore, the unlicensed BS coexists with several other unlicensed systems or WLANs. We set the number of the total scheduling interval as a hundred-fold number of the total UEs. At every scheduling interval, 4 UEs from total U UEs are selected to transmit data at same RBs with deterministic scheduling scheme, and KfUEs are selected with the proposed algorithm where Kfis updated in every ten scheduling intervals. The detailed parameters are listed in Table 2.

The throughput performances of the proposed flexible scheduling scheme (FSS) with various fixed number of the scheduled UEs,the adaptive scheduling algorithm (ASA),and benchmarks are presented in figure 5(a)with random scheduling and figure 5(b) with proportional fairness (PF) scheduling. Benchmarks include the deterministic scheduling scheme (DSS) in unlicensed spectrum and the DSS with ideal channel condition where each scheduled UE can occupy the unlicensed channel successfully. Thus, note that DSS with ideal channel condition can achieve the maximum throughput in unlicensed spectrum

As described in figure 5(a), all the throughput performances of the FSS with K=5,6,7,8 are larger than that of the DSS (K=4), and in figure 5(b), the throughput of the FSS with K=5,6 are larger than that of the DSS (K=4).However, the FSS with K =7,8 decrease the throughput, the reason of which is that the number of the scheduled UEs is too large and the scheduled UEs which occupy the unlicensed channel successfully is often larger than the number of the antennas. Thus, the FSS with some specific number of scheduled UEs can partially offset the throughput degradation because of the failed unlicensed channel occupation. Furthermore, the proposed ASA algorithm increases the throughput in unlicensed spectrum. For example, as described in figure 5(a), the throughput of the ASA algorithm with random scheduling is slightly better than that of the optimal FSS, i.e., the FSS with K =8, and in figure 5(b) it is better than the optimal FSS of PF scheduling, i.e., the FSS with K=5. Furthermore, the throughput of random scheduling is less than that of PF scheduling because PF scheduling does not only consider the fairness but also pursue the data rate.

Table I Proposed Adaptive Scheduling Algorithm

Table II System-level simulation parameters

Fig. 5 Throughput of various schemes with (a) Random scheduling, and (b) PF scheduling

Fig. 6 Fairness of various schemes with (a) Random scheduling, and (b) PF scheduling

Fig. 7 Throughput with different H-index

The fairness performances are illustrated in figure 6(a) with random scheduling and in figure 6(b) with PF scheduling. The fairness index is evaluated by using the Jain’s fairness index [31], i.e.,

The PF scheduling is fairer than the random scheduling while the random scheduling also embody the fairness of UEs because of randomness. Furthermore, the proposed ASA algorithm maintains the advantage of fairness meanwhile offset the throughput performance.

In the above simulations and comparisons,the essential H-index in the proposed SAS algorithm is fixed as 0.5. As shown in the figure 7, we present the effect of the essential H-index for the throughput in the proposed SAS algorithm. Note that in such a simulation scenario, the essential H-index for optimal throughput is about 0.5, and increase or reduce the essential H-index will degrade the throughput performance.

VI. CONCLUSIONS

In this paper, we have discussed the UL MU-MIMO scheduling in unlicensed spectrum. A flexible scheduling scheme was devised for the problem that the failure to occupy the unlicensed channel leads to the throughput degradation. Then, a low-complexity adaptive scheduling algorithm based on H-index was developed to maximize the throughput in unlicensed UL MU-MIMO systems. By adjusting the scheduled user number flexibly, the degraded performance related to the failed channel occupation can be offset statistically. The proposed scheme can be used as a guide for the UL MU-MIMO technology in unlicensed spectrum.

The authors are very grateful to all the experts from TCL Communication Technology Holdings Limited for their hard work and valuable discussions. Jia Boqi’s work was partly supported by the NSFC International Cooperation and Exchange Program (Grant No.61461136001). Zhou Ting’s work was partly supported by the Shanghai Rising-Star Program (grant no.17QA1403800) and the NSFC International Cooperation and Exchange Program (Grant No. 61461136004). Hu Honglin’s work was partly supported by the ESEC project of Tekes and the National Natural Science Foundation of China (Grant No.61401440).Yang Yang’s work was partly supported by the NSFC International Cooperation and Exchange Program (Grant No. 61461136003).

[1] Y. Liu, X. She, P. Chen, J. Zhu and F. Yang, “Easy network: the way to go for 5G,” in China Communications, vol. 12, no. Supplement, 2015, pp.113-120.

[2] P. Fan, J. Zhao and C. L. I, “5G high mobility wireless communications: Challenges and solutions,” in China Communications, vol. 13, no.Supplement2, 2016, pp. 1-13.

[3] J. Wang, J. Xu, Y. Yang, and H. Xu, “GPP based open cellular network towards 5G,” in China Communications, vol. 14, no. 6, 2017, pp. 189-198.

[4] ITU, Document 5D/TEMP/625-E. IMT Vision:Framework and overall objectives of the future development of IMT for 2020 and beyond [R].Geneva: ITU, 2015.

[5] P. Patcharamaneepakorn et al., “Spectral, Energy, and Economic Efficiency of 5G Multi-cell Massive MIMO Systems With Generalized Spatial Modulation,” in IEEE Transactions on Vehicular Technology, vol. 65, no. 12, 2016, pp. 9715-9731.

[6] W. Feng, Y. Wang, D. Lin, N. Ge, J. Lu, and S. Li,“When mmWave communications meet network densification: a scalable interference coordination perspective,” IEEE Journal on Selected Areas in Communications, vol. 35, no. 7, 2017,pp. 1459-1471.

[7] X. Wang, T. Q. S. Quek, M. Sheng and J. Li,“Throughput and Fairness Analysis of Wi-fiand LTE-U in Unlicensed Band,” in IEEE Journal on Selected Areas in Communications, vol. 35, no.1, 2017, pp. 63-78.

[8] X. Meng, J. Li, D. Zhou and D. Yang, “5G technology requirements and related test environments for evaluation,” in China Communications, vol. 13, no. Supplement2, 2016, pp. 42-51.

[9] H. Tan, W. Li, T. Wang, J. Fang and Z. Feng, “The analysis on the candidate frequency bands of future mobile communication systems,” in China Communications, vol. 12, no. Supplement,2015, pp. 140-149.

[10] RP-170828, “NR-based Access to Unlicensed spectrum,” Qualcomm Contribution to 3GPP TSG RAN Meeting #75, Dubrovnik, Croatia,March 6 - 9, 2017.

[11] H. Cui, V. C. M. Leung, S. Li, and X. Wang, “LTE in the Unlicensed Band: Overview, Challenges,and Opportunities,” in IEEE Wireless Communications, vol. PP, no.99, 2016, pp.2-8.

[12] I. E. Telatar, “Capacity of multi-antenna Gaussian channels,” European Transactions On Telecommunications, vol. 10, no. 6, 1999, pp. 585-595.

[13] M. Dohler, “Virtual antenna arrays,” Ph.D. dissertation, Dept. Elect. Electron. Eng., King’s College London, London, U.K., Nov. 2003.

[14] W. Feng, Y. Wang, N. Ge, J. Lu, and J. Zhang,“Virtual MIMO in multi-cell distributed antenna systems: coordinated transmissions with largescale CSIT,” IEEE Journal on Selected Areas in Communications, vol. 31, no. 10, 2013, pp.2067-2081.

[15] W. Feng, Y. Li, J. Gan, S. Zhou, J. Wang, and M.Xia, “On the deployment of antenna elements in generalized multi-user distributed antenna systems,” ACM Mobile Networks and Applications, vol. 16, no. 1, 2011, pp. 35-45.

[16] Y. Wang, W. Feng, L. Xiao, Y. Zhao, and S. Zhou,“Coordinated multi-cell transmission for distributed antenna systems with partial CSIT,” IEEE Communications Letters, vol. 16, no. 7, 2012, pp.1044-1047.

[17] X. Chen, H. Hu, H. Wang, H. Chen, and M. Guizani, “Double proportional fair user pairing algorithm for uplink virtual mimo systems,” IEEE Transactions on Wireless Communications, vol.7, no. 7, 2008, pp. 2425-2429.

[18] L. Wang and H. Chang, “A Low-Complexity Uplink Multiuser Scheduling for Virtual MIMO Systems,” IEEE Transactions on. Vehicular Technology, vol. 65, no. 1, 2016, pp. 463-466.

[19] B. Jia, H. Hu, Y. Zeng, and T. Xu, “The Upper Bound of Energy Efficiency for Virtual MIMO System with User Pairing,” Proc. IEEE GLOBECOM, 2016, pp. 1-6.

[20] S. Wu, W. Mao and X. Wang, “Performance Study on a CSMA/CA-Based MAC Protocol for Multi-User MIMO Wireless LANs,” IEEE Transactions on Wireless Communications, vol. 13, no.6, 2014, pp. 3153-3166.

[21] M. Xie and T. M. Lok, “Access point selection and auction-based scheduling in uplink MU-MIMO WLANs,” Proc. IEEE International Conference on Communications (ICC), 2016, pp.1-6.

[22] Y. Wu et al., “Green transmission technologies for balancing the energy efficiency and spectrum efficiency trade-off,” IEEE Communications Magazine, vol. 52, no. 11, 2014, pp. 112-120.

[23] A. Mukherjee et al., “Licensed-Assisted Access LTE: coexistence with IEEE 802.11 and the evolution toward 5G,” IEEE Communications Magazine, vol. 54, no. 6, 2016, pp. 50-57.

[24] B. Chen, J. Chen, Y. Gao and J. Zhang, “Coexistence of LTE-LAA and Wi-fion 5 GHz With Corresponding Deployment Scenarios: A Survey,”IEEE Communications Surveys & Tutorials, vol.19, no. 1, 2017, pp. 7-32.

[25] V. Valls, A. Garcia-Saavedra, X. Costa and D. J.Leith, “Maximizing LTE Capacity in Unlicensed Bands (LTE-U/LAA) While Fairly Coexisting With 802.11 WLANs,” IEEE Communications Letters,vol. 20, no. 6, 2016, pp. 1219-1222.

[26] Y. Li, T. Zhou, Y. Yang, H. Hu and M. Hamalainen,“Fair Downlink Traffic Management for Hybrid LAA-LTE/Wi-fiNetworks,” IEEE Access, vol. 5,2017, pp. 7031-7041.

[27] G. Bianchi, “Performance analysis of the IEEE 802.11 distributed coordination function,” IEEE Journal on Selected Areas in Communications,vol. 18, no. 3, 2000, pp. 535-547.

[28] R. Yin, G. Yu, A. Maaref and G. Y. Li, “Tradeoff between co-channel Interference and collision probability in LAA systems,” Proc. IEEE International Conference on Communications (ICC),Kuala Lumpur, 2016, pp. 1-7.

[29] C. Croarkin, P.Tobias, “NIST/SEMATECH e-handbook of statistical methods,” [J] American association for clinical chemistry Inc, 2010, :49(:6).

[30] Hirsch, and E. Jorge, “An index to quantify an individual’s scientific research output,” Proc.National Academy of Sciences (PNAS), 2005,pp.16569-16572.

[31] R. Jain, D. M. Chiu, and W. Hawe, “A quantitative measure of fairness and discrimination for resource allocation in shared computer system,”Tech. Rep., Digital Equipment Corp., Sep. 1984.