Efficient dynamic relay probing and concurrent backhaul link scheduling for mmWave cellular networks

Abstract

Exploiting the enormous chunks of mmWave spectrum between 30 GHz and 300 GHz have the potential to facilitate gigabit rate services to the future 5G cellular networks, and help in alleviating the current spectrum crisis. Conventional backhaul links such as Digital Subscriber Line (DSL) and Asymmetric Digital Subscriber Line (ADSL) have been proved to be a major bottleneck in satisfying these high data rate demands of indoor user equipments associated with traditional Femto Base Stations (FBS). One possible solution is to deploy higher capacity optical fiber cable to satisfy such demands. However, it is a costly and non-flexible solution. Thus, mmWave wireless backhaul links can be utilized at the FBSs. But due to their high-frequency, mmWave carrier signals are highly susceptible to obstacles and thus suffer a high attenuation in signal strength when passed through the obstacles. In order to alleviate the losses incurred due to blockages and to improve the signal reachability, in this paper, we propose an efficient distributed mode selection and dynamic relay probing scheme. We also propose an efficient scheduling scheme, for scheduling wireless backhaul links, which works jointly with the proposed mode selection and relay probing scheme to further improve the system throughput. Our proposed scheduling scheme permits non-interfering links to schedule and transmit concurrently. An expression for calculating the expected number of concurrent transmissions for our proposed scheduling scheme is derived and validated. Through extensive simulations under various system parameters, we have demonstrated the superiority of our proposed mode selection and relay probing scheme over the fixed relay probing scheme.

Introduction

The exponential growth in wireless services because of billions of devices, users, and connections have driven the need for a transition from 4G to 5G cellular network standards. Standards for the 5G cellular networks deployment have not been finalized till date. However, by considering the business and the consumer demands Next Generation Mobile Networks (NGMN) Alliance forecasted that the first phase of 5G networks should be deployed by 2020 [1]. The NGMN Alliance has also outlined a few important requirements that 5G cellular network standards should meet. These requirements include:

• High data rates of tens of Mbps for tens of thousands of users.

• Simultaneously multi-gigabit data rates to many workers on the same office floor.

• Data rates of 100 Mbps for metropolitan areas.

• Several hundreds of thousands of concurrent connections for wireless sensors.

• Better coverage, improved spectral efficiency, and significantly low latency as compared to Long Term Evolution (LTE).

The unprecedented growth in the number of cellular broadband users and data-intensive applications (high definition video streaming, ultra-fast file transfer, augmented reality, and virtual reality games, etc.) resulted in multi-gigabit rate requirements [2]. However, most of the data traffic because of these applications is generated by indoor User Equipments (UEs) associated with low powered Femto Base Stations (FBSs) [3]. In order to satisfy the data rate requirements, the radio access (FBS-UE) link capacity can be increased with the help of millimeter Wave (mmWave) techniques, higher cellular spectrum reuse, and unlicensed carrier aggregation. On the other hand, in order to carry the indoor aggregated data traffic because of the new data-intensive applications, traditional broadband connection based backhaul (such as Digital Subscriber Line (DSL) and Asymmetric Digital Subscriber Line (ADSL)) cannot be used. The traditional broadband-based connection can provide data rates up to tens of Mbps only and can become a bottleneck [4]. One possible solution in order to satisfy the backhaul requirements can be the use of optical fiber links. However, in order to cater to exponentially increasing users and their requirements, a dense deployment of FBSs is required [5]. Connecting FBSs to the other FBSs and to the Micro Base Station (MiBS) by optical fiber cable will increase the deployment and maintenance cost. Thus, mmWave based wireless backhaul at FBS can be a good solution compared to traditional broadband connections and optical fiber cables. A large amount of free bandwidth available at mmWave band is capable of solving the bottleneck issue. Wireless mmWave based backhaul not only fulfills the data rate requirements but is also cost-effective, flexible, and easy to deploy as compared to optical fiber backhaul links. Benefits of using mmWave based backhaul at FBSs is well discussed in our previous work [6]. Fig. 1 shows the comparison between traditional DSL and ADSL based backhaul architecture and mmWave based backhaul architecture.

In some of the future use cases, wireless backhaul is the only possible solution because of mobility of FBSs. For example, big vehicles (buses, trucks, etc.) supported by FBSs can provide better connectivity to other small fast moving autonomous vehicles in VANETs (Vehicular Ad Hoc Networks) [7], fast-moving trains supported by FBSs provide connectivity to passengers [8], etc. Fig. 2 shows various application scenarios where FBS-to-FBS (F2F) communication can be beneficial. It also shows usage of two modes of communication, viz. direct and relay. Huge chunks of bandwidth available in the mmWave frequency band (30–300 GHz) can be utilized to satisfy the backhaul links’ rate requirements of future 5G cellular networks [9], [10], [11], [12]. mmWave 60 GHz band, E-band (71–76 GHz and 81–86 GHz), etc. have the capabilities to satisfy these rate requirements, and can be promising backhaul solutions to meet the data rate demands of future cellular networks. Motivated by all these facts, we consider mmWave wireless backhaul links between FBSs, and also between an FBS and MiBS.

mmWave signals suffer from high free space path loss, considerable oxygen absorption, and attenuation due to blockages. For example, mmWave signals get attenuated from 20 dB to 35 dB when passed through human body [13]. Hence, both MiBS and FBSs are expected to be equipped with directional antennas that will help in achieving high data rate and longer transmission range. However, direct communication between FBS-MiBS may not always provide required data rate because of the unusual characteristics of mmWave (for example, large building blockages coming in between FBSs and MiBS). One promising strategy to get rid of signal losses due to blockages is to employ relays. Any FBS can act as a relay which is not performing any data transmission. Relays can provide an alternative multi-hop path, and increase the probability of signals to reach the receiver (MiBS). Thus, relays help in increasing the system throughput thereby increasing backhaul flows (single hop or multi-hop paths) throughput.

Though relays can provide an alternative multi-hop path for signals to reach the destination, determining whether a relay is appropriate or not needs learning of both source–relay (i.e., FBS-FBS (Relay)) and relay–destination (i.e., FBS (Relay)-MiBS) channels. For channel estimation, mmWave transmitter and receiver need to align their antenna beam towards each other and this beam alignment takes a non-negligible overhead. In this paper, we call probing as the process of exploring whether a particular transmitter and receiver pair can provide a required link capacity or not, after performing beam alignment. Hence, probing more relays may increase the probability of finding a better relay but at the cost of more probing overhead. Thus, there is a tradeoff in searching for a good relay in mmWave systems. Hence, in order to maximize the backhaul flow throughput, proper mode selection is important. It is also essential to determine how many relays should be probed in mmWave systems or in other words it is important to determine a bound on the number of probes. Our paper addresses these issues and determines the bound on the number of probes.

In this paper when we say probe it means probing single transmitter and receiver pair. Hence, probing FBS-FBS (Relay) and FBS (Relay)-MiBS requires two independent probes.

Most of the current research studies on FBSs were done using the sub-6-GHz band. Chen et al. [14] have proposed backhaul-constrained resource optimization for distributed femtocell interference mitigation. An energy efficient cell selection framework for femtocell networks with limited backhaul link capacity is proposed in [15]. Chu et al. [4] have discussed backhaul-constrained cooperative management strategy for interference management in dense femtocell networks. These studies consider backhaul-constrained resource allocation or interference management because of limited backhaul capacity of broadband connection such as ADSL.

The existing relaying and link scheduling techniques for 4G cellular networks are designed considering omnidirectional antennas, which may be sub-optimal for mmWave 5G cellular networks. This is due to the different channel characteristics and usage of directional antennas in mmWave communication as compared to 4G LTE.

The existing research work on mmWave transmission and link scheduling is mostly focused on Wireless Personal Area Network(WPAN) [16], Wireless Local Area Network (WLAN) [17], and ECMA-387 [18]. Cai et al. [19] have derived the exclusive region conditions and used these conditions for concurrently scheduling the links efficiently in Wireless Personal Area Network (WPAN). The derived conditions ensure better performance than Time Division Multiple Access (TDMA). In the protocols proposed in [20], [21] for WPAN, multiple links are concurrently scheduled to communicate in the same slot if the interference due to multiple links is below a specific threshold. Rehman et al. [22] have considered an ideal directional antenna and proposed a vertex multiple coloring concurrent transmission scheme for Device-to-Device (D2D) network over unlicensed 60 GHz mmWave band. They have also proposed that while scheduling different flows, preference should be given to the flows having better chances for higher data rate. Spatial-time division multiple-access scheduling algorithm for WPAN considering the mmWave band has been studied in [23]. The algorithm allows both interfering and non-interfering links to transmit concurrently while satisfying the Quality of Service (QoS) requirements of each link. He et al. [24] have developed a decomposition principle in order to minimize the maximum expected delivery time for the link and relay selection in centralized dual-hop 60 GHz networks. Qiao et al. [25] have addressed a resource sharing scheme by allowing non-interfering D2D links to operate concurrently in an mmWave 5G cellular network.

Gia Khanh Tran et al. [26] have proposed a method to control mmWave meshed backhaul for efficient operation of mmWave overlay 5G HetNet through Software-Defined Network (SDN) technology. Their proposed method is featured by two functionalities, i.e., backhauling route multiplexing for overloaded mmWave small cell base stations (SC-BSs) and mmWave SC-BSs’ ON/OFF status switching for an underloaded spot. They assume backhaul interfaces of SC-BSs are located at the streets lamp posts and Line-of-Sight (LOS) conditions are always guaranteed. However, in practical scenarios it might not be true. H. Ogawa et al. [27] have proposed a traffic-adaptive formation for outdoor hotspots. They have proposed a load balancing mechanism called route multiplexing to support certain mmWave APs which need to accommodate traffic higher than what their mmWave gateway (GW) sectors can support. Especially, the architecture employs multi-hop relay to support mmWave APs located far from the GW. The proposed mechanism assumes zero interference among backhaul links. However, in practical deployment scenarios, interference among backhaul links cannot be ignored as it will create network bottleneck and affects the whole system rate. A. Mesodiakaki et al. [28] propose an optimization framework for the design of policies that help in solving the problem of where to associate a user, subsequent routing of packets over the backhaul, and identifying which backhaul links and base stations to be switched off such that energy cost is minimized without hampering the user demands. They assume LOS conditions are always guaranteed between mmWave backhaul links. Also, they consider the interference among adjacent backhaul links is negligible.

To the best of our knowledge, no work in the literature focuses on the scheduling of mmWave based backhaul for FBSs by jointly considering the mode selection and relay probing costs. The mmWave band can provide the required capacity, but proper scheduling of mmWave based backhaul link for FBS is very important in order to achieve the maximum benefit. Our paper targets the same as one of the major objectives.

One of the promising works for relay probing in mmWave cellular networks is proposed in [29] for D2D relaying. It was shown that throughput-optimal relay probing strategy is a pure threshold policy and relay probing can certainly be stopped once the optimized spectral efficiency threshold is achieved by a two-hop link. However, they have considered a single user scenario and left the multiuser scenario for future work. Apart from this, they have considered a full signal loss due to blockages, i.e., they consider received signal strength to be zero when an obstacle is present in between the transmitter and receiver, which might not capture the actual system scenario. Rebato et al. [30] have compared and examined several resource sharing configurations in 5G cellular networks in order to achieve the enhanced potential of mmWave communication. They have shown the advantage of full infrastructure and spectrum sharing configuration in terms of increase in user data rate, and economical advantages for each of the service providers. The problem of joint beamwidth selection and power control is well studied in [31]. Authors have shown the tradeoff between the beam alignment and throughput. They have proposed standard compliant algorithms for short wave mmWave scenarios considering the alignment-throughput tradeoff. However, they have left the extension of proposed work for the mmWave based cellular networks as a future work. Our work well studies the future work insights and intricacies of [29] and [31] considering cellular networks.

The existing research work has been mainly focused on mode selection, relay selection, and link scheduling in WPAN, WLAN, ECMA-387, etc. using mmWave spectrum or FBSs using broadband connection based backhaul. In contrast, our work particularly focuses on mode selection, relay probing, and efficient concurrent backhaul scheduling for FBSs using mmWave based backhaul. The main contributions of this paper are summarized as follows:

• There exists a non-negligible beam-alignment overhead inmmWave systems because of the use of directional antennas. We consider a more practical and realistic scenario in scheduling the backhaul links for FBSs by explicitly considering the overhead incurred due to beam alignment.

• Proper mode selection (direct or relay) of data transfer is important in order to maximize the system throughput. Our proposed solution efficiently determines the mode of data transfer and the best relay in case relay mode is selected.

• Probing more relays may increase the chances of finding the better relay but at a cost of more probing overhead because of beam alignment. We identify the tradeoff between relay probing and throughput for a single FBS scenario and provide a heuristic for multiple FBSs scenario.

• mmWave band has huge chunks of bandwidth available in order to cater FBS backhaul requirements. However, proper scheduling of backhaul links or flows is important. We propose an efficient polynomial time scheduling scheme which schedules non-interfering backhaul links concurrently.

• We derive an expression for calculating the expected number of concurrent transmissions for our proposed scheduling scheme and validate it using extensive simulations.

In this work, we consider the maximum of two-hop communication links only because further increasing the hops may increase the delay and thereby reduce the system throughput. Also, it increases the overhead due to beam-searching overhead at each step in mmWave systems.

The paper is organized as follows. In Section 2, we explain system model for mmWave communication in cellular networks. Section 3 discusses the optimization problems for determining best stopping rule criteria and to maximize the number of concurrent backhaul link transmissions. Then, we discuss our proposed solutions for both the optimization problems in Section 4 by deriving various heuristics. In Section 5, we derive an expression for calculating the expected number of concurrent transmissions for our proposed concurrent backhaul link scheduling scheme. Section 6 presents the simulation and analytical results, and comparison with other schemes. Finally, conclusions are drawn in Section 7.