Keywords

1 Introduction

Every year, user needs and the number of device connections are increasing. According to the Cisco statistics [1], mobile video traffic has been believed to increase by 50 percent annually. Due to the growing popularity of multifunctional multimedia applications and services, wireless mobile networks must constantly evolve, offering higher data rates, as well as reducing data transfer delays and increasing energy efficiency, which leads to an increase in the quality of service for end users. 3GPP New Radio (NR) radio access technology standardized by 3rd Generation Partnership Project (3GPP) [2] is expected to play a key role in 5G systems. NR systems promise multi-gigabit rates together with reduced latency at the air interface compared to 4G Networks. Despite the enormous available bandwidth potential, mmWave signal transmissions suffer from fundamental technical challenges like severe path loss, sensitivity to blockage, directivity, and narrow beamwidth, due to its short wavelengths. To effectively support system design and deployment, accurate channel modeling comprising several 5G technologies and scenarios is essential [3, 4]. Though mmWave frequency band offers huge amount of available resources, the lack of PTM capabilities may imply a future limitation of 5G networks, leading to inefficient service provisioning and utilization of the network and spectrum resources [5,6,7].

In indoor scenarios 5G NR systems mostly suffer from obstacles such as humans and vehicles, which generally tend to be mobile and are often termed “blockers” [8]. For the sake of block error probability mitigation at the air interface a user equipment (UE) experiencing such type of blockage may lower its modulation and coding scheme (MCS) that depends on propagation environment and the distance to NR base station (BS), or enter outage conditions in case further lowering of MCS is impossible [9]. Multi-connectivity operation recently proposed by 3GPP is aimed at solving the problem of outage conditions by maintaining several simultaneously active links for adjacent NR BSs so that the connection is redirected between them in case of blockage [10, 11]. However, when lowering MCS, additional physical resources are utilized to support the required rate at the air interface. Once the required amount of resource in not available, a session should either reduce its rate, or it should be dropped [3].

In [12] authors introduce a joint scheduling framework based on dynamic MC to satisfy the different requirements of enhanced mobile broadband (eMBB) and ultra-reliability low-latency communication (URLLC) traffics in 5G. The framework aims to optimize multiple traffics in independent link dynamically. BSs will form MC clusters, which vary with network characteristics to accomplish high rate. Maximum capacity of eMBB traffic is achieved while the latency of the uRLLC traffics is guaranteed. The authors of [13] characterize the outage probability and spectral efficiency associated with different degrees of MC in a typical 5G urban scenario, where the line-of-sight propagation path can be blocked by buildings as well as humans. These results demonstrate that the degrees of MC of up to 4 offer higher relative gains.

In this paper, we consider a 5G NR BS deployment serving mixture of unicast and multicast sessions. MC techniques is proposed to be used for cell-edge users. First, by means of stochastic geometry we derive the amount of required resource in terms of average values that will further simplify our computations. Then, we present our simulation approach and finally validate our model against computer simulations. Our findings showed the advantages of multicast sessions over unicast ones in terms of reliability and resource consumption.

The remainder is organized as follows. First, our system model of NR BS operation under unicast and multicast traffic load introduced in Section 2. The simulation approach is offered in Sect. 3. Numerical results illustrating the resource occupation effect of multicast sessions are presented in Section 4. Conclusions are drawn in the last section.

2 System Model

This section provides an overview of the considered scenario with 5G NR Base Station that serves unicast (point-to-point, PTP) and multicast (point-to-multipoint, PTM) sessions[14]. As we assume only rather small mobile obstacles in our model, the BS has circularly-shaped coverage area of radius \(d_{LoS}^E\). We assume random distribution of users that follows Poisson Point Process (PPP) with density parameter \(\rho \). The coverage area radius can be evaluated using the mmWave propagation model mentioned below together with MCSs what gives us the SNR threshold at which the session can be no longer served at current BS. The mapping between CQIs and spectral efficiency for 3GPP NR systems is given in Table 1 where LoS ad n nLoS ranges are estimated for the default system paramenters presented in Sect. 4. This approach follows [3], where authors compute mean demand for the model without multiconnectivity feature.

Table 1. CQI, MCS, and distance mapping.
Full size table

When establishing a user session, BS allocates the amount of physical resource that is generally a random variable and depends on propagation model and the distance from UE to the serving BS. According to [15, 16], the mmWave linear path loss L in dBs for UEs in LoS and nLoS conditions is given by:

$$\begin{aligned} L(x) = {\left\{ \begin{array}{ll} 32.4 + 21 \log (x) + 20 \log f_c, \text {non-blocked,}\\ 47.4 + 21 \log (x) + 20 \log f_c, \text {blocked,} \end{array}\right. } \end{aligned}$$

where \(f_c\) is operational frequency measured in GHz, and x is the distance between BS and UE. These expressions allow us deriving maximum distances \(d_{nLoS}^E\) and \(d_{LoS}^E\) at which a UE is able to start a session in blocked and non-blocked conditions correspondingly. This can be done by taking the lowest SNR defined by MCS as the value of L threshold [9, 15].

In our model the coverage area is divided into two sectors: “inner sector (zone)" with radius \(R_I = d_{nLoS}^E\); and the “outer sector (zone)" with width \(R_O = R_C - R_I\), where \(R_O\) is restricted by the radius of the coverage area \(R_C\). This division allows us for improving resource demand estimation as demands of sessions from inner zone that are always served at the same BS considerably different from sessions from outer zone that are subject of MC mechanism.

Then, to define the resource demands \(b^{(n)LoS}_{I/O}\) that depend on both UE location (inner or outer zone) and LoS conditions we use approximations of spectral efficiency for each of the four propagation scenarios. The approximations are performed as the following: first, using the conventional MCS we define the longest possible distances \(x_i\) for each given i-CQI in LoS and nLoS conditions (1); then we calculate the coverage area for the CQIs as \(\varOmega _i = \pi (x_{i}^2 - x_{i-1}^2)\) for i-CQI, \(x_0=0\); finally, we derive the average resource demand for each of the scenarios (2).

$$\begin{aligned} x_i = {\left\{ \begin{array}{ll} {\left( {\frac{{{P_A}{G_A}{G_U}}}{{S_i{N_0}W\left( {{{10}^{2{{\log }_{10}}{f_c} + 3.24}}} \right) }}}\right) ^{\frac{1}{\gamma }}}, \text {non-blocked}, \\ {\left( {\frac{{{P_A}{G_A}{G_U}}}{{S_i{N_0}W\left( {{{10}^{2{{\log }_{10}}{f_c} + 4.74}}} \right) }}}\right) ^{\frac{1}{\gamma }}}, \text {blocked,} \end{array}\right. } \end{aligned}$$
(1)

where \(P_A\) is the NR BS transmit power, \(S_i\) is the worst possible SNR given by MCS for i-CQI, \(G_A\) and \(G_U\) are the antenna array gains at the NR BS and the UE ends, respectively, \(N_0\) is the power spectral density of noise, W is the given bandwidth, \(\gamma \) – path loss exponent.

$$\begin{aligned} b^{(n)LoS} = {\left\{ \begin{array}{ll} \left\lceil {v \cdot {{\left( {{s_A} \cdot \sum \limits _{i:0< {x_i} \le {R_I}} {\frac{{\varOmega _i^{(n)LoS} \cdot {E_i}}}{{{R_I}}}} } \right) }^{ - 1}}} \right\rceil , \text {inner zone},\\ \\ \left\lceil {v \cdot {{\left( {{s_A} \cdot \sum \limits _{i:{R_I} < {x_i} \le {R_C}} {\frac{{\varOmega _i^{(n)LoS} \cdot {E_i}}}{{{R_C}}}} } \right) }^{ - 1}}} \right\rceil , \text {outer zone}, \end{array}\right. } \end{aligned}$$
(2)

where \(s_A\) is the physical resource block (PRB) measured in MHz, \(E_i\) is the spectral efficiency corresponding to i-CQI, v is the required service data rate.

Mobile obstacles may appear on LoS towards the BS and thus reduce the SNR of the established connection. In this paper we consider the typical type of blockers for indoor scenarios which is humans with their blockage radius \(r_B\) approximating the width of human body. We model blockers’ movement with two exponentially distributed random variables with parameters \(\theta _{LoS}\) and \(\theta _{nLoS}\) that correspond to the time intervals when a blocker is passing the LoS and periods between two consecutive blockage events [3]. It should be noted that the \(\theta _{LoS}\) and \(\theta _{nLoS}\) intensities generally depend on density of users and their behaviour, i.e. their velocity and mobility model. It means that in case of dense networks, system performance may be severely compromised not only by the number of active devices, but also the increased LoS blocking probability.

Fig. 1.
figure 1

Communication scenario.

Full size image

Figure 1 illustrates the communication scenario considered in our paper. We observe the resource allocation process at a single NR BS. We suppose there are no interfering beams that could worsen channel quality. Each blocker has its own radio shadow area with length \(d_B=\frac{d(h_B-h_U)}{(h_A-h_U)}+r_B\) [3]. Inside inner zone UE has only one active link with the target NR BS, resources are allocated even if UE in the radio shadow area. As the MCS does not support connection in nLoS conditions whenever UE at the distance \(d>R^I\), multiple links are maintained towards the target and nearby BSs; once it gets into the shadow, user session is handed over to the BS with the greatest SNR at the moment.

As we assume that NR BS maintains constant bitrate for an established session, in our model session is only dropped when its new demand is more than remaining resource. Session requires additional resource in the following cases: (i) new session is initiated; (ii) LoS towards UE is blocked in inner zone; (iii) ongoing session is handed over back to original NR BS.

We also assume that there is no prioritization among traffic types. Whenever a session is attempting for establishment, NR BS allocates unoccupied resource without any reservation mechanisms [17].

3 Simulation Approach

In this section we present the system-level tool that we used to validate the above-mentioned model. As we have limited system capacity in terms of resource, this guarantees accessibility of steady-state conditions [18] that are detected using exponentially-weighted moving average technique with smoothing parameter set to 0.05.

The simulation tool utilizes an event-driven engine that allows for processing predefined set of events [19, 20], such as session establishment and aborting, blockages, handover, resource management, etc. This approach provides both much flexibility and powerful capabilities for statistic data collection. High-level algorithm of the simulation is described as UML activity diagram presented in Fig. 2.

Fig. 2.
figure 2

Simulation algorithm.

Full size image

Data is collected during the steady-state period only providing better accuracy compared to analytical results. In what follows, we demonstrate only point estimates of the metrics of interest.

4 Numerical Assessment

In this paper, we analyze NR BS resource allocation process needed for establishing and maintaining unicast and multicast [21] sessions composing mixed traffic load. Our metrics of interest are unicast and multicast session drop probabilities and mean NR BS resource utilization which are the key indicators of system reliability. The input system parameters [22] are given in Table 2.

Table 2. Input data
Full size table

The traffic profile (Table 3) complies with the global traffic forecast for 2025 [23], while service characteristics provided in [24]. We consider six types of services: streaming video [25], audio, file sharing, social networking, Web, and Machine-to-Machine [26] traffic. Data is transmitted by unicast sessions for all the services, but video can be also streamed via multicast sessions to reduce the amount of utilized resource.

Table 3. Service characteristics.
Full size table

In Fig. 3 and Fig. 4 we present session drop probabilities for different density values. One may observe extremely rapid degradation of unicast video service compared to other services. This can be explained by the huge resource demand of video traffic coupled with long-lasting service delivery time. It is much easier for services with smaller demands to fit into the remainder of the resources in case of high traffic load at NR BS.

Figure 4 shows that in case of handover sessions the gap between video and other types of services in terms of session continuity becomes even greater.

Fig. 3.
figure 3

Session drop probabilities.

Full size image
Fig. 4.
figure 4

Handover failure probabilities.

Full size image
Fig. 5.
figure 5

Resource utilization.

Full size image
Fig. 6.
figure 6

Session drop probabilities as function of ISD.

Full size image

In Fig. 5 we demonstrate resource distribution among the above-mentioned services. Along with the network densification the share of resource occupied by video traffic declines, ceding it to the more light-weight and short-time services.

We also present numerical results for session drop probabilities (Fig. 6) as function of inter-site distance for different ratio between unicast and multicast traffic shares. As may be observed, the trend of unicast numerical superiority over multicast in drop probabilities is mainly kept within all the variety of traffic shares. However, when the number of multicast sessions becomes insufficient to continuously occupy allocated amount of resource, the multicast session drop probabilities start to outnumber unicast ones.

By setting a threshold at some maximum allowed value of drop probability, it is possible to estimate the longest possible distance between adjacent NR BSs. Such approach allows for optimization of NR BS deployment minimizing the number of BSs to provide acceptable coverage within a given area. This may significantly reduce capital and operations expenses of network operator, but in this case one should give thorough consideration to the network scalability.

Our findings show that multicast sessions offer more reliability compared to unicast communications providing four-times less drop rates while occupying less amount of resource. This can be explained by the concept of resource occupation while organizing a multicast session. Once amount of resource is allocated to a multicast session, all the successive multicast sessions providing the access to the same content will keep on capturing this resource until the last session comes to its end.

5 Conclusion

In this paper, we considered the radio frequency resource allocation by a 5G NR BS with multi-connectivity mechanism that allows for reducing density of BS deployment. By means of the developed simulation tool we provided the numerical analysis of 5G NR BS deployment serving mixture of unicast and multicast sessions. Our findings revealed the advantages of multicast sessions over unicast ones in terms of reliability and resource consumption.

Our future work is to study fair radio resources distribution policies between unicast and multicast multimedia services in next-generation mobile networks that allow to protect disadvantaged types of services to fulfil QoE requirements.