Weight Based Channel Selection Towards 5G in the Unlicensed Spectrum

Yu Zeng, Ting Zhou＊, Honglin Hu Yang Yang, Jinfeng Tian Zhenhong Li Shanghai Institute of Microsystem and Information Technology (SIMIT), Chinese Academy of Sciences,No. 865 Changning Road, Changning District, Shanghai 200050, China

2 University of Chinese Academy of Sciences, No. 19A Yuquan Road, Shijingshan District, Beijing 100049, China

3 Shanghai Advanced Research Institute (SARI), Chinese Academy of Sciences, No. 99 Haike Road,Pudong District, Shanghai 201210, China

4 ShanghaiTech University, Shanghai 201210, China

5 TCL Communication Technology Holdings Ltd. Shanghai R&D Center, Building C, 5-9F, No.232, Liangjing Road,Zhangjiang High-tech Park, Shanghai 201203, China

* The corresponding author, email: zhouting@sari.ac.cn

Abstract: The growth of the wireless and mobile communication data traffic has brought severe challenges to the present telecommunication systems. To meet the ever-increasing mobile traffic demand in the next 5th generation (5G) communication systems, deploying 5G in the unlicensed spectrum (5G-U), has been regarded as a promising technology. The Third Generation Partnership Project (3GPP)has specified the standardization of the Licensed Assisted Access (LAA) and its extension enhanced LAA (eLAA), to opportunistically transmit in the unlicensed spectrum. The LAA/eLAA systems share unlicensed spectrum resource with other networks, e.g., the Wi-Fi systems. In this article, we analyze the coexistence between the eLAA and the Wi-Fi systems in the unlicensed spectrum. Wefirstly establish the system model where the eLAA coexists with the Wi-Fi systems. Then, we theoretically derive and figure out the unfairness in the multi-channel occupancy rate between the eLAA and the Wi-Fi systems. After that,we propose a weight based channel selection method to improve the fairness of the coexistence. The numerical results demonstrate that by avoiding contentions and declining collisions, our method not only enhances the fairness, but also improves the overall unlicensed spectrum usage rate.

Keywords: 5G; unlicensed spectrum; eLAA;Wi-Fi; multi-channel transmission

I. INTRODUCTION

The nextfifth generation (5G) communication systems will be based on both the New Radio(NR) access technologies and the evolved existing wireless technologies, e.g., the long term evolution (LTE) and the Wireless Fidelity(Wi-Fi) [1]. Currently in the Third Generation Partnership Project (3GPP), both the smooth evolution of LTE and the development of the NR interface are in process. Nonetheless, due to the continuous growth of the mobile traffic demand as well as the service requirements from users, the limited licensed spectrum is not able to satisfy the future demands and the sufficient spectrum resource becomes significant for the practical utilization of 5G. The unlicensed spectrum, which has a wide bandwidth and is costless, is now becoming attractive in both academia and industrialfields. According to the recent research, efficient utilization of the unlicensed spectrum is regarded as a promising solution to meet the requirements of the high data rate and large capacity in the 5G communication systems [2-4].

Under such a circumstance, the 3GPP has already started the projects of the deployments of LTE, LTE-A and even NR in the unlicensed spectrum. In LTE Release 13, 3GPP has extended the downlink (DL) transmissions of LTE in the unlicensed spectrum, i.e., the Licensed Assisted Access (LAA) [5-6]. By utilizing the carrier aggregation (CA) technique in LTE, the LAA is able to aggregate the channels of both licensed and unlicensed spectrum. Among the transmissions of the LAA,the conventional LTE transmissions occur in the licensed spectrum as normal, while the unlicensed spectrum is opportunistically utilized.More specifically, the signals of the control plane are always transmitted via licensed spectrum while the signals of the data plane can be transmitted via either licensed or unlicensed spectrum. In LTE Release 14, the 3GPP established the work item of enhanced-LAA(eLAA), where the uplink (UL) transmissions are also supported [7]. According to the theoretical analysis and simulations, the eLAA can significantly improve system throughput[8-9]. In this article, we consider the eLAA in which both UL and DL transmissions in the unlicensed spectrum are supported.

Since the unlicensed spectrum is accessible for all the wireless systems, the nearby systems could cause interference to each other if they simultaneously transmit in the same channel, as shown in figure 1. Thus, a system needs to detect and ensure that there are no other systems already occupying the unlicensed channels, before its own transmission.In this condition, the fair coexistence, i.e.,the equipments in different systems have the equal chances to access and occupy the same unlicensed spectrum for transmission, must be maintained [10-11]. The widely used IEEE Wi-Fi systems use Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA) to ensure fairness with other systems. Meanwhile, to achieve a high data rate as well as the efficient spectrum utilization, the multiple channel transmissions are supported in the Wi-Fi systems, where the channel bonding mechanism is used to keep the fair coexistence[12]. The multiple channel transmissions are also supported in the eLAA [13], where the multi-channel detection needs to be performed before the transmission in any unlicensed channel. Nonetheless, since the multi-channel access mechanism of the eLAA is more flexible than that of the Wi-Fi, the eLAA has higher probabilities to occupy more unlicensed channels, which brings unfairness to the Wi-Fi systems [14-15]. To solve this problem, several member corporations of the 3GPP have proposed proposals [14-16]. However, most of the proposed methods merely modify the eLAA channel access mechanisms to align with the Wi-Fi. Thus, theflexibility from the LTE CA is not fully utilized. This critical issue also becomes a hotspot in academia recently. In [17],the authors propose a multi-objective optimization framework to balance the performances of Wi-Fi and LAA in the unlicensed spectrum.Paper [18] introduces a time-frequency structure for the channel access of both LAA and Wi-Fi users. The authors in [19] propose a hyper access point, which improves the system throughput and user fairness against traditional LAA nodes deployment. In [20], based on the stochastic geometry, the authors reveal the effects of different channel access priorities on the performance of different coexisting LAA networks in the unlicensed spectrum. However, these existing works mentioned above have not simultaneously considered the multi-channel transmissions of both the Wi-Fi and the eLAA in the unlicensed spectrum.

Motivated by this circumstance, in this article, we analyze the scenario of the multi-channel transmissions in the unlicensed spectrum.We first theoretically derive the unfairness between the eLAA and the Wi-Fi in the unlicensed spectrum occupancy. After the theoret-ical analysis, we propose a weight based channel selection method for the multi-channel contention and access in the unlicensed spectrum, to improve the fairness of the coexistence. In our method, before the transmissions,the eLAA collects the information about the channel distribution of nearby Wi-Fi systems and arranges channel contention weights for every candidate channel. The channels with relatively lighter weights are prior selected by the eLAA for transmissions. Then, we provide the numerical results which conform to our analysis and demonstrate the effectiveness of the proposed method. The main contributions of this article are summarized as follows. 1.We establish the system model where eLAA coexists with Wi-Fi systems and mathematically derive the channel occupancy rate of each system to reveal the essential reason of the unfairness among their contentions in the unlicensed channels. 2. We propose a weight based channel selection method for the eLAA system and provide theoretical demonstration as well as numerical results, which prove the effectiveness of this method. Due to the avoidance of contentions in the channels with heavy weights, comparing to the current standard in[13], the proposed method not only enhances the fairness among the eLAA and the Wi-Fi systems, but also improves the overall unlicensed spectrum usage rate.

Fig. 1. Coexistence and interference scenario of eLAA and Wi-Fi.

The rests of this article are organized as follows. In Section-II, we illustrate the channel access mechanisms of both the eLAA and the Wi-Fi systems in detail. Then in Section-III,we theoretically analyze their coexistence fairness during the multi-channel transmissions in the unlicensed spectrum. Our proposed method is then stated in Section-IV. After that,the numerical results and the relevant analysis are in Section-V. The conclusions are made in Section-VI.

II. MULTI-CHANNEL ACCESS IN THE UNLICENSED SPECTRUM

2.1 Wi-Fi multi-channel access mechanism

Multi-channel transmissions are supported in the Wi-Fi systems. For example, the IEEE 802.11 ac can support multi-channel transmissions with an overall bandwidth of 160MHz[12]. For effective channel utilization as well as to avoid collisions between different nearby Wi-Fi systems, the channel bonding mechanism is adopted, in which channels are arranged with different levels.

As shown in figure 2, a simple example in 802.11ac is used to demonstrate how the channel bonding works. Consider the channel bonding in four 20MHz channels with an overall bandwidth equal to 80MHz, one of the 20MHz sub-channels is selected as the primary 20MHz channel. The 40MHz valid channel which contains the primary 20MHz channel mentioned above is defined as the primary 40MHz channel. Then the 20MHz channel next to the primary 20MHz channel and within the primary 40MHz channel is defined as the secondary 20MHz channel. The primary 80MHz channel and secondary 40MHz channel are similarly defined as the primary 40MHz channel and the secondary 20MHz channel, respectively.

Before multi-channel transmissions, the normal channel detection procedures arefirstly performed in the primary 20MHz channel. If the channel is sensed to be idle for a distributed inter frame space (DIFS), the transmission begins. Otherwise, the Wi-Fi system randomly chooses a back-off interval between 0 and the initial contention window size (CWS). The back-off counter decreases if the channel is sensed to be idle for a slot, whose length is specified by the IEEE 802.11 protocol. However, if the channel is busy during the back-off counter decline, this procedure is frozen and will be resumed until the channel is idle for another DIFS period. If any collision happens during the transmission, the CWS doubles. If CWS has reached the upper bound, it stays in the maximum value for the rest time.

Then as for the secondary 20MHz channel,the system needs to check the activity in it for a point coordination function inter frame space(PIFS) period, prior to the intended transmission time in the primary 20MHz channel. The system can transmit in the primary 40MHz channel if both the primary 20MHz and secondary 20MHz channels are idle. If the primary 20MHz channel is idle and the secondary 20MHz channel is busy, only the former one can be used. Otherwise, no channel is available. In similar, the primary 80MHz channel can be used if the primary 20MHz, secondary 20MHz and secondary 40MHz are all idle simultaneously. Besides, for fairness consideration, IEEE has specified that the primary channel belong to different nearby Wi-Fi networks, should be aligned if they can detect the existence of each other.

We arrange the primary 20MHz channel with the highest level while the secondary 20MHz with relatively lower level and the rest are by the same analogy. It can be obviously found that the detection results of the channels in higher levels decide whether the channels in lower levels can be occupied for transmissions.

2.2 eLAA multi-channel access mechanism

Fig. 2. Channel bonding on 80MHz in 802.11ac.

The eLAA system needs to perform listen-before-talk (LBT) before transmitting in an unlicensed channel. In [13], two types of LBT, i.e.,Type 1 and Type 2, are specified for the eLAA transmission in one channel. In the Type 1 LBT, the eLAA may transmit in a channel afterfirst sensing the channel to be idle for the slot duration of a defer duration Td, which is similar to the DIFS in Wi-Fi. If the channel is busy during Td, the eLAA randomly chooses a back-off counter, which ranges between 0 and the initial CWS. The back-off counter decreases by 1 if the channel is sensed to be idle for a slot duration, whose length is specified by the 3GPP specifications [13]. However, if the channel is busy during the back-off counter decline, this procedure is frozen and will be resumed until the channel is idle for another Tdperiod. In similar to the Wi-Fi systems, the CWS doubles if collisions are detected and it stays if it has reached its upper bound. The Type 2 LBT allows the eLAA system to occupy a channel for transmission if the channel is idle for at least one Tshortulduration, which is similar to PIFS in Wi-Fi. Based on the above explanations, we canfind that the Type 1 and the Type 2 in eLAA are corresponding to the DIFS together with the possible following back-off slots and the PIFS in Wi-Fi, respectively.

For the eLAA multi-channel transmissions in the unlicensed spectrum, the procedures of channel access mechanism are specified in[13]. A simple example in figure 3 is used to demonstrate how this mechanism works. Consider that an eLAA system intends to transmit in a candidate channel groupwhich contains four channels, e.g., C1C4. To maintain the fairness in the occupancy of different unlicensed channels, the random channel selection is used, where one channel, e.g., Ci, will be randomly selected fromto perform the Type 1 LBT. Then the Type 2 LBT will be per-formed before the intended transmission time.If the Type 1 LBT has succeeded, the channels which succeed in the Type 2 LBT will also be accessed, while the transmission periods are aligned with Ci. As shown in figure 3, since the Type 1 LBT is successful in the randomly selected C1, the C4which passes the Type 2 LBT will also be utilized by the eLAA, while the C2and C3which fail in the Type 2 LBT,will not be accessed.

Fig. 3. The channel access mechanism for multi-channel transmissions in eLAA.

Fig. 4. Comparison of eLAA with Wi-Fi.

2.3 The coexistence of the multichannel access in the unlicensed spectrum

Comparing the multi-channel access mechanism of Wi-Fi in Section 2.1 with that of the eLAA in Section 2.2, the eLAA is able to monitor and occupancy channels more flexibly, while the Wi-Fi is strictly restricted by the channel levels. Besides, the difference of the two kinds channel access mechanisms can lead to spectrum resource waste.

A simple example in figure 4 is used to better illustrate the above situations. In figure 4(a), assuming a Wi-Fi system is already transmitting in channel C1and C2, which are selected as the primary 20MHz channel and the secondary 20MHz channel, respectively. Then a nearby eLAA system intends to utilize C1C4and performs the Type 1 LBT in the randomly selected channel, which is assumed to be C3. Besides, the Type 2 LBT is performed in the rest three channels before the intended transmission time. Since C3is detected to be idle during the Type 1 LBT, the channels which is idle during the Type 2 LBT will also be utilized by the eLAA. Thus, though Type 2 LBT fails in C1and C2, the rest C3and C4can still be utilized by the eLAA.

On the other hand, assuming an eLAA system already occupies C1and C2and a nearby Wi-Fi system intends to utilize C1C4, while the channel level distributions are as in figure 4(b). Since the Wi-Fi detects that C1is busy during DIFS and following back-off slots,according to the channel bonding mechanism,the rest channels cannot be used even C3and C4are idle during PIFS. Therefore, the spectrum resources in C3and C4are wasted.

III. THEORETICAL ANALYSIS OF THE COEXISTENCE FAIRNESS IN THE UNLICENSED SPECTRUM

In this section, we provide theoretical analysis of the coexistence fairness of the eLAA and the Wi-Fi systems in the unlicensed spec-trum. We assume that both the Wi-Fi and the eLAA systems are in the full buffer state, i.e.,they have immediately a packet available for transmission after the completion of each successful transmission [21]. According to the channel access procedures in Section 2.1 and 2.2, we can plot a revised Markovchain based on the classical one proposed in [21], to model the back-off procedures of the general Wi-Fi or eLAA systems. We assume the two kinds systems are trying to occupy N channels within the same frequency range in the unlicensed spectrum. Considering that beforemulti-channel transmissions, the eLAA and the Wi-Fi both need to perform channel detection in one specific channel, i.e., the primary channel in the Wi-Fi and the channel which is selected to perform the Type 1 LBT in the eLAA. Therefore, we firstly analyze the contention in the special single channel as follows.

As shown in figure 5, the system state is denoted as (m, b, c), where m is the number of the retransmission attempts, b isthe present back-off counter and c is the number of the channels occupied for transmission. As in [21]and [22], a key approximation isthat at each transmission attempt, regardless of the number of retransmissions suffered, eachpacket can be successfully transmitted with constant and independent probability pidle. Hence, the state of b＞0 will remain unchangedwith probability (1−pidle) and the back-offcounter can be decremented by 1 with probability pidle.When the back-off counter is zero, the system begins transmission and the statewill change to (0,b,0) with a probability, which represents that after a successful transmission,the back-off stage is reset to 0 and the new back-off counter will be uniformly selected in the range [0, W0-1]. If a collision happens during the transmission, due to the increased back-off stage and the doubled CWS, the state will transit to (m+1,b,0) with aprobability. Moreover, the upper bound of the back-off stage is set as M. Based on the above statements, we can derive the one-step transition probabilities as (1).

By denoting qm,b,cas the steady probability of state (m, b, c) and based on the Markov states transition, we can derive the steady probability of the transmission state, i.e., qtr,as the following equation (2),

From eq. (3), we can find that except for the special case when Pidle=0.5, the P( tr)is monotonically increasing with Pidle. Nonetheless, in the IEEE 802.11 specifications, the primary channels of the neighbor cells belonging to different Wi-Fi networks are aligned,for fairness consideration. Nonetheless, based on [13], the primary channel in the Wi-Fi and the randomly selected channel to perform the Type-1 LBT in the eLAA are highly possibly staggered.

The details in the multi-channel transmissions state are specific to the Wi-Fi and the eLAA. To focus on the influence from the two kinds multi-channel access mechanisms, we assume the probability that a channel is idle during the Type 2 LBT is. Then for the eLAA, the probability when its state transits from (m,0,0) to (m,0,c), i.e., c channels are occupied for transmission, is

On the other hand, the probability that the Wi-Fi occupies c channels is based on the following equation, i.e.,

Fig. 5. Markov model of channel access in the unlicensed spectrum.

According to (4), in the eLAA system, the LBT results of the channels which perform the Type 2 LBT, are independent and cause no influence to each other. In the Wi-Fi system,on the other handin (5) reflects that the channels with lower levels are affected by channels with relatively higher levels. Therefore, the eLAA multi-channel access mechanism is more flexible. More specifically, if a channel with relatively higher level in the Wi-Fi is occupied by the eLAA, the channels with lower levels cannot be used by the Wi-Fi even if they are idle. Thus, in order to promote the fairness in channel access as well as to achieve a high spectrum efficiency, the eLAA should prior select the channels which have relatively lighter influence to the nearby Wi-Fi systems.

IV. WEIGHT BASED CHANNEL SELECTION METHOD FOR FAIR COEXISTENCE

We use the average channel occupancy rate Oc, to evaluate the channel occupancy condition for a system. Specifically, the Oc for a Wi-Fi system, indicated by Ocw, can be calculated by

where T is the overall time for the observation, N is the total number of the channels in the unlicensed spectrum, andis the overall occupancy time of the Wi-Fi system in channel Ci. In similar, the Oc for an eLAA system, indicated by Oce, can be calculated by

We assume the Wi-Fi systems and eLAA systems are contending for the channels from the same unlicensed channel pool. As such, in eq. (6) and (7), N and T are constant parameters. Supposing both the two kinds systems are in full buffer status, they will transmit in the occupied channels for the maximum permitted time in their respective standards. Besides,once a transmission process is complete, the related system immediately starts the next channel access. Under such a circumstance,both Ocwand Oceare determined by the number of the occupied channels and the time of occupancy. To realize the fair coexistence of the channel contentions in the unlicensed spectrum, we need to balance Ocwand Oce. According to (5), based on an appropriate arrangement of the channels eLAA intends to occupy, i.e., the eLAA system avoids contentions with Wi-Fi system in the channels with high levels, the probability that channel with the lower level is idle during the period of a PIFS increases, and thus Wi-Fi is able to has a larger probability to occupy more channels.

With the above consideration in mind, in this section, we propose a weight based channel selection method (WBS) to appropriately arrange channels for eLAA to utilize and improve the coexistence fairness between the eLAA and the Wi-Fi in the unlicensed spectrum. The main idea is to map the levels of channels in Wi-Fi systems to specific weights,where the channels with relatively lower levels in Wi-Fi are arranged with lighter weights and vice versa. Then based on the information of channel weights, the eLAA system prior selects the channels with lighter weights for transmission and declines the collisions with Wi-Fi systems in their high-level channels.Thus, the channel occupancy rate of Wi-Fi is improved while the eLAA can still maintain its own channel occupancy rate andfinally the entire channel usage rate is promoted.

Assuming based on the channel condition evaluation, the eLAA system intends to transmit in N channels, i.e., C1, C2, ..., CN, while the nearby Wi-Fi systems also utilize these channels. The detailed procedures of the proposed WBS are as follows. Firstly, the eLAA system collects the information of the channel distributions in the Wi-Fi systems, e.g., φifor the i-th nearby Wi-Fi system. Then based on φi, every candidate channel in the channel pool, i.e., from C1to CN, will be arranged with a weight., e.g.,for Cnin i-th Wi-Fi system, whileis equal to the number of channels influenced by the channel monitor result of Cnin that Wi-Fi system. Hence, the primary channels, i.e., the higher-level channels, will be arranged with relatively heavier weights, comparing to the secondary channels(the lower level channels) and vice versa. For example, assuming a channel, e.g., Cn, determines whether 4 channels are available for the i-th Wi-Fi system. Then Cnwill be arranged with a weight equal to 4 in the i-th Wi-Fi system, i.e.,=4. The channel in which the i-th Wi-Fi system does not contend will be arranged with a weight equal to 0. Thus, we can obtain all the channel weights, i.e.from the i-th Wi-Fi system. If there are multiple adjacent Wi-Fi systems with different channel level distributions, multiple weights of a channel are set according to the same principle and then thefinal channel weight is the summary of the multiple weights from different systems.Finally, the channels with lighter weights will be prior considered for the channel access of the eLAA multi-channel transmissions. The proposed WBS can prior select the channels for transmission and causes less influence to the nearby Wi-Fi systems. Moreover, since our method works in the channel selection before the actual LBT, it will not affect the current standard.

We assume that there are WNnearby Wi-Fi networks in contention with the eLAA in the unlicensed spectrum and the weights for candidate channels are indicated by vectorwhere ωirepresents the weight for the i-th channel. The details of the proposed method are summarized in algorithm 1.

For a better illustration, a simple example is provided in figure 6, where 8 channels in the unlicensed spectrum are considered. In Wi-Fi system 1, i.e., Wi-Fi 1, Ch1 is the primary channel which means if Ch1 is busy for Wi-Fi 1, the rest 7 channels are not available for Wi-Fi 1 even if they are idle. Therefore, the detection result in Ch1 decides whether Wi-Fi 1 has the opportunity to access the overall 8 channels. Hence, for Wi-Fi 1, we arrange Ch1with a weight equal to 8. On the other hand,the detection result in Ch1 decides whether Wi-Fi 2 has the opportunity to access the overall 4 channels and thus for Wi-Fi 2, Ch1 is assigned with a weight equal to 4. In similar, the rest channels are assigned with specific values of the weight, respectively. Then, the weight in each channel from different Wi-Fi systems is summarized. In this case, the channels with lighter weights will be prior selected by the eLAA. For example, if the eLAA intends to transmit in 4 channels, Ch5 ～ Ch8 will be prior selected. In similar, if the eLAA intends to transmit in 6 channels, Ch3 ～ Ch8 are prior considered.

Algorithm 1. Weight based channel selection method (WBS).

Fig. 6. An illustration of weight based channel selection method.

Table I. Key parameters of simulations.

We need to note that though our proposed method is based on the eLAA system, it is not limited for the extensions of LTE. It can be easily revised to suit the future wireless systems, e.g., the NR, as long as the NR-based network follows the multi-channel access in eLAA for transmissions, which is the trend of the current standardization process. We provide the numerical results in Section-V to demonstrate the effectiveness of the proposed method.

V. NUMERICAL RESULTS AND ANALYSIS

In this section, we use numerical simulations to evaluate the effectiveness of the proposed method. As in figure 1, we consider an eLAA system sharing 8 continuous and orthogonal channels in the unlicensed spectrum with two different and nonadjacent Wi-Fi systems,while Wi-Fi 1 intends to transmit in all the channels and Wi-Fi 2 randomly contends for 4 continuous channels. Without loss of generality, the typical channel model is applied [14-15], while the parameters are listed in table 1.The Wi-Fi systems are working according to IEEE 802.11ac protocol. We assume all the three wireless systems have the full buffer data traffic. To emphasize the impact from the difference of channel access mechanism, the maximum times for retransmission attempt and the occupancy time for once transmission are set to be identical. To better reveal the in-fl uence from the traffic load of the eLAA, the number of channels which the eLAA intends to occupy varies from 2 to 6, which indicates the different levels of its traffic load. The original random (RA) channel selection from [13]is regarded as a benchmark. We use the average channel occupancy rate to evaluate the channel occupancy condition. The entire Wi-Fi systems average channel occupancy rate is defined as

In figure 7, we plot the average channel occupancy rate of the Wi-Fi systems. In general,the proposed WBS can significantly improve the channel occupancy rate of the Wi-Fi systems. That is because the channels with relatively higher levels in the Wi-Fi systems are assigned with heavier weights and thus avoiding the severe contentions with the eLAA by effective channel selection. Hence, the Wi-Fi systems are less affected comparing to when the eLAA using RA. More specifically, when K varies from 2 to 4, the occupancy rate gap between proposed WBS and RA increases.Since that in these two cases, the channels are not fully utilized. Thus, by avoiding the unnecessary contentions between systems,the WBS can achieve a larger performance gain. On the other hand, we can alsofind that when K increases to 6, the rate gap between WBS and RA declines on the contrary. That is because in this case, the contentions inevitably begin in the high-level channels in Wi-Fi systems, which limits the effectiveness of the proposed WBS.

In figure 8, we plot the average channel occupancy rate of the eLAA system. By applying WBS, the eLAA system also avoids plenty of contentions with Wi-Fi systems.Therefore, WBS also outperforms RA in the eLAA channel occupancy rate. In similar to figure 7, the gap of the occupancy rate by utilizing RA and WBS also varies with K. When K increases from 2 to 4, the eLAA intends to transmit in more channels. In this case, WBS selects the channels which are with relatively lighter weights in the Wi-Fi systems and effectively reduces collisions. Therefore, the channel occupancy rate gain of WBS is increased significantly comparing to the traditional RA.However, when K increases from 4 to 6, the eLAA has to contend with Wi-Fi systems on their higher-level channels and thus limits the gain brought by WBS.

Fig. 7. Wi-Fi average channel occupancy.

Fig. 8. eLAA average channel occupancy rate.

We compare the overall channel occupancy rate of RA and WBS in figure 9. Obviously,WBS brings a higher overall channel occupancy rate of the unlicensed spectrum than the traditional RA. The reason can be referred from figure 4, where the situation of channel resource wasting is avoided to a certain extent. Besides, we can alsofind that the overall channel occupancy rate is higher when the eLAA intends to occupancy 4 channels, rather than when the eLAA intends to occupancy 6 channels. That is because if the eLAA system is with quite heavy load, i.e., when K=6,much time will be consumed in channel detection and contention among the three systems in simulation. Besides, the frequent collisions between the eLAA and Wi-Fi also limit the effective channel occupancy rate.

VI. CONCLUSIONS

Fig. 9. Average effective channel occupancy rate.

In this article, we analyze the coexistence fairness between the eLAA and W-Fi systems in the unlicensed spectrum. We theoretically derive and figure out that under the current standardization specifications, the eLAA has more chances to occupy channels than the Wi-Fi during multi-channel transmissions, which brings unfairness. After that, we propose a weight based channel selection method to solve this issue. The numerical results demonstrate that based on effective channel selection and by avoiding unnecessary contentions, our method not only enhances the fairness between the eLAA and Wi-Fi systems, but also significantly improves the overall channel occupancy rate.

ACKNOWLEDGEMENTS

The authors are very grateful to all the experts from TCL Communication Technology Holdings Limited for their hard work and valuable discussions. This work was partly supported by the National Science and Technology Major Project (grant no. 16510711600) and the National Natural Science Foundation of China(grant no. 61631013). This work was also partly supported by the National Natural Science Foundation of China (grant no. 61401440).