Mobile M2M communication architectures, upcoming challenges, applications, and future directions

Unlike traditional mobile networks such as global system for mobile communications (GSM), LTE is the first fully packet-switched mobile network proposed by 3GPP in 2004. 3GPP standardized LTE in its Release 8 with small modifications in Release 9 [13, 29]. The standard is designed on the basis of the GSM and universal mobile telecommunications system (UMTS) networks to provide improved energy and spectral efficiency. The LTE radio and non-radio enhancements include evolved UMTS terrestrial radio network (E-UTRAN) and system architecture evolution (SAE), respectively. The evolved packet system (EPS) is the combination of LTE and SAE. The function of the EPS bearer² is to route traffic from gateways (GWs) in the packet data networks (PDNs) to user equipments (UEs). A high level E-UTRAN architecture is presented in Fig. 5 (right). According to the 3GPP Technical Report TR 25.913 [30], the details of the performance requirements of 3GPP LTE networks can be described as follows:

LTE interfaces between UE, E-UTRAN, EPC, and S-GWs are associated with protocol stacks which are used to exchange data and signaling messages [31]. Therefore, the LTE protocol stack can be divided into the following categories:

On the air interface (Uu), the UE high-level functionalities are controlled by the MME [32, 33]. However, there is no direct communication path between UE and MME. The communication path between UE and MME is established using the evolved node base stations (E-UTRAN eNB) which enhance the level of the hardware complexity in the network. To reduce this complexity, the Uu interface is further divided into two levels of protocols. One is called access stratum (AS) whereas the second is the non-access stratum (NAS). The MME high-level signaling lies in the NAS level but is transported within the network using AS protocols. The control and user plane termination protocols are supported by the eNB. The user plane protocols include the packet data convergence protocol (PDCP), the medium access control (MAC), and the physical (PHY) layer protocols. In addition, the protocols of the control plane include the radio resource control (RRC) protocols. A high level overview of the protocol stack is depicted in Fig. 5 (left). The major functionalities of these protocols are given below:

LTE transmission schemes support orthogonal frequency division multiple access (OFDMA) and SC-FDMA in downlink and uplink, respectively. The radio frame structure in the LTE uplink and downlink is similar. In LTE evolved universal terrestrial radio access (E-UTRA), the channel bandwidth ranges from 1 to 20 MHz [40]. The E-UTRA channel bandwidths include 1.4, 3, 5, 10, 15, and 20 MHz. Each LTE frame (10 ms) is divided into ten subframes of 1 ms each [34], see Fig. 6. Every single subframe consists of two equal time slots of 0.5 ms. Each slot is further divided into six and seven OFDM symbols in the extended and the normal cyclic prefix (CP), respectively. For uplink and downlink communications in LTE, the smallest radio resource which can be allocated to each UE is called PRB which has a dimension of 180 KHz in the frequency domain and 1 ms in the time domain. A single PRB is further composed of 12 consecutive sub-carriers. In normal cyclic prefix, there are 84 resource elements per PRB whereas an extended cyclic prefix is composed of 72 resource elements per PRB.

LTE radio resource management (RRM) enables efficient utilization of the scarce radio resources among several UEs. RRM consists of a set of procedures to distribute available PRBs among multiple users while ensuring their QoS as well as seamless connectivity [41]. The scheduling of radio resources in LTE is performed by the MAC layer of the eNB. Downlink and uplink scheduling procedures are initiated in every transmission time interval (TTI) of 1 ms depending on the amount of data in the buffers of eNB and UE, respectively. Furthermore, the RRM initiates the scheduling process when sufficient PRBs are available in the network to meet the QoS demands of other users as well. The transport block size (TBS) represents the capacity to transmit data with respect to the RRM procedure. In addition, the capacity of TBS depends on the available modulation and coding scheme (MCS) as well as the number of allocated PRBs to a particular UE. Figure 7 represents the capacity of TBS (in bits) with respect to the MCS index and the number of allocated PRBs. It can be seen from Fig. 7 that TBS capacity increases for higher values of MCS and larger number of allocated PRBs. For instance, 328 bits can be sent per PRB with an MCS index of 16. However, the PRB capacity increases to 712 bits when the index of MCS is raised to 26.

In LTE networks, each UE performs random access (RA) procedures to transmit data during the access grant time interval (AGTI) or RA slots over a time-frequency radio resource called physical random access channel (PRACH). A UE initiates the RA mechanism in two cases [42]. Firstly, when a UE is turned on and the uplink timing synchronization is either lost or not yet achieved. Secondly, whenever a handover mechanism is performed from a serving eNB to a target eNB. In LTE, UEs perform the RA procedure for several reasons such as for establishing a communication link, re-initiating a radio link after a failure, for radio resources in case of no uplink resource grant, and scheduling requests in the absence of dedicated physical uplink control channels (PUCCHs).

The RA procedure can be classified into a contention-based and a contention-free random procedure [19, 42‐44]. Initially, a UE uses a contention-based RA procedure to select a preamble. A preamble is an OFDM-based signal with a narrower sub-carrier spacing. It is possible that several devices can select the same preamble, which demands a further contention resolution mechanism. In a contention-free RA procedure, an eNB uses dedicated preambles to initiate the RA mechanism. Hence, it ensures faster response than the contention-based approach. Contention-free RA procedures are usually used during critical situations such as handover. The major steps of contention-based and contention-free RA procedures are given in Fig. 8. Furthermore, a contention-free RA procedure is simpler than the contention-based RA procedure and consists of three steps by allocating dedicated preamble signatures to a device (see Fig. 8). The procedure starts with the RA preamble assignment by the eNB. After the transmission of an assigned RA preamble by the UE, the eNB responds with the random access response (RAR). This is the last step of the contention-free RA since there is no need to resolve further collision.

Battery life optimization for mobile UEs, especially smartphone users, has always remained a challenging task for communication researchers. In the case of smartphones, users charge their phones frequently, usually every day. However, they always demand for the optimum power consumption of phones in order to increase its battery lifetime. In order to fulfill the increasing demand of improved UE power consumption, 3GPP introduced the power-saving mode (PSM) such as discontinuous reception (DRX) and idle state in its Release 12 TS 23.682 [45]. Under the PSM, a UE requests the network for an active timer duration during tracking area updates (TAUs) procedures. In this time, a UE remains attached, however all the AS layer functions are stopped. The DL data can be received by a UE in the RRC connected state, and during the active time period. The active timer is initiated by the UE when it moves from the connected to the idle state. The device moves to PSM with the expiry of the timer. The value of the active timer represents the time in which the device is accessible. In PSM, a device remains attached/registered with the network. However, it is not accessible due to the fact that it does not check for paging in PSM. Moreover, the device stays in PSM until it starts an activity such as UL data transmission. Figure 9 depicts the state transitions of a UE from connected to idle, and vice versa.

LTE-A is the latest 3GPP standard which defines further advancements with respect to the existing LTE standard to approach the needs of IMT-Advanced which has been set by the International Telecommunication Union Radio communication (ITU-R) [46]. According to these requirements, LTE-A is capable to provide throughput of up to 1 Gbps in downlink and 500 Mbps in uplink. The peak spectrum efficiency in LTE-A is also improved by 15 Bps/Hz in uplink and 30 Bps/Hz in downlink with the maximum antenna configurations of 4×4 and 8×8, respectively. Moreover, the LTE-A standard is also designed to reduce the latency in user and control planes compared to the existing LTE standard. To achieve the defined tasks for LTE-A, the problems such as low signal-to-interference-plus-noise ratio (SINR) and network coverage issues due to shadowing and non-line-of-sight (NLOS) are considered. According to [47], the capacity of the 3GPP LTE Release 8 network is already very close to the Shannon limit and it is required to upgrade the existing technologies with new standards to support the increasing future data traffic. One idea is to increase the number of eNBs which is not a cost-efficient approach and not an appealing option for network operators.

In LTE-A networks, the evolved UMTS terrestrial radio network (E-UTRAN) consists of eNBs with a similar role as that of the base station (BS) in GSM but with additional functionalities [41, 48]. The E-UTRAN handles communication between the UE and the EPC, whereas an eNB handles a UE in one or more cells. The serving eNB (SeNB) is the eNB which currently provides services to a UE. Moreover, the eNB controls low-level operations such as handover functions. The eNB communicates with the core network through the S1 interface. Several eNBs are also inter-connected with each other through the X2 interface (see Fig. 5, right) which is mainly used for exchanging information such as packet forwarding during handover. The S1 interface between an eNB and EPC can also perform handover functions at the cost of more delay. The S1 and X2 interfaces operate using an IP-based transport network. Moreover, E-UTRAN also provides home eNB (HeNB) services. HeNB is an eNB which is particularly dedicated to a user within a home for femtocell coverage. It can be connected either directly to the EPC in the same way as a standard eNB is connected or using some intermediate terminals such as gateways that forward information from multiple eNBs.

According to ETSI specifications discussed in [15, 49, 50], the M2M architecture consists of the following five key elements which are shown in Fig. 10:

M2M area networks The M2M area networks are responsible for establishing the communication path between the M2M devices and the M2M gateways. These networks are usually called subnets which collect and route information from the M2M devices to the M2M gateways. There are several subnets which are used for generating the communication link between the M2M devices and the M2M gateways [50]. Generally, the use of subnets is dependent on the network technology. In fully distributed networks, all M2M devices are connected as peers to the network. One of the nodes which is connected to the network, e.g., via wired or wireless connectivity, acts as a router. In client-server networks, all nodes or devices communicate directly with the server. Whereas in cooperative networks, all nodes communicate with each other using some intermediate gateways. Some of the major subnets are presented in Table 2.

Communication networks The main role of the communication network is to create a communication path between the M2M devices and application servers either through wired or wireless communication networks. The main communication networks are listed in Table 3.

In the 3GPP MTC reference model given in [45, 51‐53], the communication between MTC devices and application servers use 3GPP services and optionally services capability server (SCS)³ services. The application server (AS) particularly hosts an MTC application in the external network. Furthermore, the AS manipulates an SCS for supplementary value-added services. The 3GPP provides several services to support MTC, e.g., transport, user management as well as architectural enhancements, e.g., initializing control plane functionalities of an MTC device [52]. 3GPP envisages various models for MTC communications [52]. These models are related to the SCS provider-based communication between the AS and the 3GPP system. The various architectural models proposed by 3GPP include the direct model, the indirect model, and the hybrid model [45, 52, 53]. In the case of the direct model, the AS directly performs data communication with the UE by connecting directly to the network and does not use any SCS service (see Fig. 11). In addition, the 3GPP services can be used by the application in the outer network. In the case of the indirect model, the AS uses SCS services and connects indirectly to the network. Thus, the indirect model utilizes supplementary value-added services for MTC such as control plane device triggering. In addition, the SCS is either controlled by the MTC service providers or network operators in order to provide several services, e.g., value-added services and communication services related to control and user plane. In case of the hybrid model, both the aforementioned models (direct and indirect) are used by the AS concurrently. The AS performs direct data communications with the device by connecting directly to the network. However, it also uses SCS services. From the 3GPP network aspect, the AS direct data communication and the SCS value-added services, e.g., control plane communication are fully independent, although both can service the same AS MTC application. In addition, the MTC applications can receive distinct services/facilities from the SCS controlled by the MTC service provider as well as the 3GPP operator. Since the abovementioned three models are complementary, 3GPP operators can combine them for various applications. For instance, the SCS based on the MTC service provider can be combined with the 3GPP network operator controlled SCS, both communicating with the similar PLMN.

Architectural reference model The architectural reference model for MTC is depicted in Fig. 12, which shows an MTC device communicating with 3GPP networks, i.e., E-UTRA, E-UTRAN, and GERAN using Um, Uu, and LTE-Uu interfaces, respectively. Moreover, Tsp is used for 3GPP network services to SCS and AS (see Fig. 12). 3GPP has standardized the Tsp interface in order to support value-added services for MTC such as control plane device triggering. In addition, an SCS usually provides Tsp interface. Moreover, the MTC interworking function (MTC-IWF) is a functional entity used to support indirect and hybrid models by residing in the home public land mobile network (HPLMN). It usually shows the internal PLMN network topology. Furthermore, it also translates Tsp-related signaling protocols to bring particular functionality in the PLMN. The descriptions of various MTC-related reference points and network elements are summarized in Table 4 (taken from [53]).

The service requirements vary significantly due to the diversity of MTC applications. For this reason, the authors of [15] and [54] stated that MTC traffic types significantly depend on the type of technologies used for communication between devices and MTC application domains. This subsection presents MTC service requirements by focusing on 3GPP networks.

3GPP MTC features The MTC traffic characteristics differ significantly from application to application. Resultantly, system as well as service optimizations also vary, thus a particular optimization approach might not be applicable to the vast M2M application range [55]. 3GPP has defined a set of features for MTC in order to provide a structure to optimize systems according to varying MTC applications. In general, 3GPP identifies the following MTC features in its technical specifications TS 22.368 [55]:

3GPP MTC common service requirements 3GPP has set certain service requirements for MTC in its Release 10 and onward. The major 3GPP-based MTC service requirements are summarized below (taken from [55‐58]):

MTC device triggering 3GPP-based device triggering requirements of an MTC device are summarized as follows [55, 58]:

According to [15], M2M communications significantly rely on numerous technologies spanning across several industries. Due to the wide application range, standardization in M2M communication is more challenging than the traditional standardization for a particular platform. Consequently, the required standardization scope is much higher than the existing standardization developments and entails to upgrade/modify existing standards to support future M2M traffic on a large scale. 3GPP, ETSI, IEEE, and Telecommunications Industry Association (TIA) have played a significant role in developing new standards to support M2M communications. The standardization work related to M2M communication is on-going in several standardization organizations outside 3GPP (e.g., ETSI, IEEE, TIA, and oneM2M). Moreover, these standardization bodies have also described the required enhancements in the existing communication standards in their specifications. One important effort for M2M standardization in 2012 is the oneM2M partnership by the global standard development organizations (SDOs) [59].

This section presents the major standardization efforts made by several SDOs to facilitate M2M communications and services. The major efforts made by several standardization bodies are described below. Moreover, a summary of their major contributions is given in Table 5.

3GPP 3GPP has made progress for network developments for inter-MTC device communication in Release 12. 3GPP proposed several solutions to handle large numbers of MTC devices in GSM and UMTS networks. The study focused on handling a massive number of MTC devices which are expected to communicate either directly with each other or with the backend server simultaneously. Moreover, the service requirements for MTC are addressed in [54]. Several key issues such as network congestion due to the large number of devices are discussed. In addition, numerous possible solutions such as group-based communication are given in order to tackle the aforementioned issues. 3GPP in its Release 12 [61] described various enhancements as well as use cases for MTC. Several applications for MTC such as transportation and logistics, surveillance, and e-healthcare are addressed. The security aspects in MTC are addressed in [62] and [63]. Furthermore, a detailed study of RAN enhancements and GERAN improvements for MTC is also given.

In addition to the aforementioned contributions, 3GPP started interacting with third parties in [64] and empowered them to facilitate users through third party services using 3GPP functions. In order to accomplish it, third parties also enable MTC services after making an agreement with operators. Furthermore, 3GPP has started working on an MTC architecture in Release 10. In Release 11, 3GPP defined several interfaces between the core network (CN), MTC service, and application enablement platforms. 3GPP is currently working on small data transmissions and low power consumption devices in Release 12 SA2. The major existing and expected efforts to support MTC in upcoming Releases are summarized in Table 6.

ETSI ETSI has made serious efforts in developing standards and protocols for M2M communication. ETSI has established a new committee with goals to provide an overall M2M communication overview. Moreover, the group has also focused on relating M2M communication closely with the ETSI development for the latest communication standards. Moreover, ETSI focuses on to relate M2M communication with the advanced 3GPP new communication technologies. The detailed E2E network architecture for M2M communication by ETSI is given in [28]. Furthermore, a detailed study of the inter-communication of the proposed ETSI M2M architecture with local area networks such as ZigBee is given in [65]. In addition, several communication interfaces such as mla, dla, and mld for M2M communications are provided in [66]. Moreover, the future requirements for M2M communication such as mobility support are discussed in [67]. Test specifications for the constraint application protocol (CoAP) binding of M2M primitives is presented in [68]. In addition, several use cases of M2M communication such as e-healthcare and smart metering are explained in [69] and [70], respectively. Moreover, M2M automotive applications and smart grid applications are discussed in [71] and [72], respectively.

IEEE IEEE has started studying M2M communication by announcing working groups for M2M communication [73, 74]. The working group IEEE 802.16p 10/0005 describes the usage of M2M communication and various standard modifications which are required to support future M2M communications. In addition, the working group IEEE 802.16p-10/0014 provides various methodologies which seem to be helpful for M2M communications. Furthermore, IEEE also focuses on air interface optimization for mass device communication, the use of sub-GHz spectrum (Wi-Fi) for M2M communication, and optimizing air interfaces for the smart grid IEEE 802.15.4.

TIA [75] This SDO provides bi-directional communication between smart devices and the networks in its TR-50 framework [76]. Furthermore, it provides an extension for e-healthcare scenarios in which e-healthcare sensors can access the converged personal network service (CPNS) server in the global network via a mobile gateway. This extension in e-healthcare provides monitoring and simple daily care services. Hospitals and pharmacies can access information received from the sensors deployed on the body of the patient.

Open Mobile Alliance (OMA) [80] The key enablers for M2M devices and services include M2M device management, M2M gateway management, personal area network services to fulfill ETSI M2M area network requirements, and location detection services [80].

Due to the emerging use of mobile networks for M2M communications, several research projects such as EXpAnding LTE for Devices (EXALTED) and Achieving LOw-LAtency in Wireless Communications (LOLA) focused M2M communications for supporting increasing M2M traffic efficiently. Furthermore, several of the latest EU research projects such as METIS-2020 have extensively addressed existing and upcoming M2M-related issues in their research. Consequently, several recommendations and solutions have been proposed in order to handle M2M small-sized transmissions efficiently on a large scale. METIS-2020, one of the largest EU research projects addressed M2M upcoming issues as a front line research topic, within the list of its horizontal topics. Resultantly, several solutions have been proposed for handling future mobile M2M traffic efficiently. This subsection presents an overview of these projects and highlights their major contributions, for the ease of the reader to understand the existing and upcoming issues and research gaps. In the end, a summary of all major contributions related to M2M communications is given in Table 7.

METIS-2020 [82] Mobile and Wireless Communications Enablers for the Twenty-Twenty Information Society (METIS-2020) is an EU research project, initiated in 2013. Twenty six partners including major operators such as Ericsson, NTT Docomo, and NOKIA have participated in this project. The primary goal of this project is to lay the foundation of upcoming 5G mobile and wireless systems. Since cellular-based M2M communications is one of the latest research topics, several massive machine communication (MMC)-related issues are addressed in METIS [83]. Furthermore, future MMC scenarios are discussed, and consequently several novel radio link schemes are proposed to handle the increasing mobile M2M traffic. Since handling MMC in future mobile networks is a challenging task especially for the radio link layer, METIS focused on providing multiple access and multiple control schemes, advanced MAC and HARQ algorithms, synchronization issues, and MMC-related signaling limitations for MMC scenarios in its radio link analysis.

EXALTED [3] EXALTED is an EU FP7⁴ ICT research project commenced in 2010 (September) for a duration of 30 months. The project was coordinated by Sagemcom SAS, France, along with the other 12 partners including well-known operators and module manufacturers such as Vodafone, Ericsson, and Gemalto. One of the key issues of EXALTED is to provide secure, cost- and energy-efficient M2M communications in the context of future mobile technologies. Therefore, EXALTED also considered existing challenges in mobile networks to finally deliver the foundation of a new scalable network architecture. Furthermore, the project aimed at the specifications of massive machine communications under the name of LTE-M networks particularly in mesh networks. The resulting system provides a low cost and spectrally efficient M2M system with extended coverage in LTE networks. Approximately 1000 devices were expected to be supported per cell with the proposed architecture and schemes by assuming quite relaxed latency requirements greater than 100 ms.

Existing mobile standards are neither designed to handle small-sized payloads efficiently nor to support large number of devices simultaneously [83]. Consequently, this will lead to network congestion. Current mobile networks must be extended or revised to support massive M2M devices in the future. Since mobile radio resources are valuable assets and scarcely available, it is required to ensure efficient utilization of radio resources for M2M communications. According to present 3GPP standardization, the smallest unit of the radio spectrum allocatable to a single device is 1 PRB which is capable to transmit several hundred bits under favorable channel conditions (see TB size for 1 PRB in Fig. 7). However, allocating 1 PRB to a single M2M device could significantly degrade radio spectrum utilization. This is due to the fact that the capacity of a PRB can be much higher than the actual size of the device payload.

Figure 13 illustrates how inefficiently a PRB can be used when a device sends small-sized packets experiencing good channel conditions. Two exemplary device payloads (i.e., 4 and 8 bytes) are considered. Moreover, a PRB utilization efficiency is evaluated in terms of percentage of the actual payload and the padded zeros with respect to the MCS index. It can be seen from Fig. 13 (left) that with the lower MCS values, e.g., 0–6 depicting unfavorable channel conditions, the PRB capacity is smaller, and therefore it is fully utilized to transmit the given payload. On the other hand, with the higher MCS values, e.g., 20–26 representing favorable channel conditions, the capacity of the PRB increases. However, small packets with increasing percentage of padded zeros are transmitted. Similarly, Fig. 13 (right) depicts the percentage of PRB utilization and padded zeros for the second scenario. However, it can be seen from Fig. 13 (right) that for the lower MCS values, the PRB is used efficiently as compared to Fig. 13 (left). However, with the higher values of MCS, the percentage of padded zeros is still significantly large. This shows that approximately 10–20 times more packets can be accommodated per PRB under favorable channel conditions in the given exemplary scenarios.

Due to the inefficient utilization of scarce radio resources, M2M traffic is expected to degrade the performance of traditional mobile traffic. For instance, the authors of [89, 90] evaluated the influence of increasing M2M traffic on LTE data traffic through system level simulations using an LTE-A-based model developed in the OPNET Modeler. The authors concluded that the performance of LTE/LTE-A low priority traffic such as FTP and HTTP is degraded significantly when the number of M2M devices was raised to 750 (Fig. 4 in [89]). This happened due to the fact that e-healthcare traffic was prioritized over LTE traffic.

In LTE/LTE-A networks, radio resources are distributed among several users in order to accomplish user plane and control plane services. The massive number of M2M devices is expected to influence the performance of LTE/LTE-A networks for both the user and the control plane due to the exchange of excessive signaling messages for sending/receiving small packets. Therefore, these topics demand serious attention to enhance M2M services efficiently using LTE/LTE-A networks.

In communication networks, efficient radio resource management has always remained a challenging topic. It has become more challenging for mobile networks to handle traditional traffic co-existing with cellular-based MTC traffic. Therefore, new and optimized radio resource management schemes are required to support future MTC traffic, for which LTE and LTE-A networks are not yet optimized [83]. Several proposals have been presented so far for efficient radio resource allocation for LTE/LTE-A networks.

For instance, authors of [18] proposed two methods for allocating radio resources between MTC and H2H devices. The first method is based on the orthogonal radio resource allocation. The strength of this resource allocation scheme is its simplicity, whereas the drawback is the low spectral efficiency from the system level aspect. However, the second method, i.e., shared radio resource allocation improves spectral efficiency of the network in a way that MTC devices can reuse the bandwidth allocated to H2H/H2M users. However, the drawback is an increased level of interference as compared to the former approach. In 3GPP LTE/LTE-A networks, radio resources are split into PRBs also called subchannels in the context of resource allocation. For instance, if MTC traffic co-exists with traditional H2H traffic in a communication environment, five distinct links are available which are listed below, see Fig. 14 (taken from [18]):

In a mixed traffic environment, one of the major challenges is handling of interference. Therefore, it requires a sophisticated resource partitioning mechanism. The radio resource partitioning enables radio resource allocation among different devices in a coordinated way. The restrictions can be either on the basis of available PRBs or transmit power. As a result, such restrictions provide the possibility for improving the SINR ratio.

Several scheduling algorithms have been proposed for LTE/LTE-A networks. For instance, authors in [91] evaluated the performance of dynamic packet scheduler for LTE networks. The advantage of this approach is the effective control of throughput and fairness between users. However, the drawback of the proposed scheme is that it does not differentiate services. Additionally, authors in [92] introduced a principle of metric decoupling to maximize throughput. Moreover, authors concluded that the fairness among users can be controlled efficiently either through weighting metric in the frequency domain or handling priority in the time domain. In [93], a path selection criteria for small cells is proposed for joint transmissions on the access and backhaul links in the millimeter wave band to improve the performance of device-to-device (D2D) communication. The proposed D2DMAC scheduling scheme is centralized and outperforms traditional approaches particularly in terms of delay and system throughput. Authors in [94] studied dynamic TDD transmission mechanisms for small cells with several interference management schemes such as interference cancellation, cell clustering, as well as power control. The advantages of proposed schemes are potentially large in low-to-medium traffic loads. Boosting UL power in dynamic TDD scheme significantly outperforms static TDD scheme. Approximately 30–60 % and 210–300 % user throughput is improved in DL and UL, respectively. Moreover, authors in [95] proposed a feasible strategy to enable dynamic TDD transmissions in small cells to achieve UL and DL throughput gains. Additionally, authors in [96] made a theoretical analysis on the SINR performance in case of dynamic time-division duplex (TDD) transmission in small cells. By introducing a partial interference cancellation scheme, outage probabilities are notably decreased. Moreover, authors concluded that the risks of link failures can be significantly reduced by canceling one interfering cell in case of low-to-medium loaded scenarios.

Authors in [97] proposed a cell breathing framework in downlink to minimize breathing effect in femtocells, and thus to balance data rates and cell coverage. A voting-based mechanism, i.e., femtocell virtual election rule (FEVER) is proposed in which users notify the femtocell base station about their current channel conditions which significantly improve system efficiency and fairness among users. In [98], authors studied game dynamics and learning schemes and proposed a novel learning scheme called cost-to-learn which integrates factors such as switching cost and delay as well as user behavior. Resultantly, users concurrently learn their optimal payoff and strategy with the designed CODIPAS-RL (combined fully distributed payoff and strategy reinforcement learning) scheme. Moreover, authors created asymptotic pseudo-trajectories to solve differential equations. In [99], authors proposed a distributed and coordinated scheduling scheme for LTE/LTE-A networks where each cell tends to allocate MCS, PRB, and transmit power to users for minimizing DL power while satisfying user throughput demands. Furthermore, each cell notifies neighboring cells about PRB allocation to cell-edge users. Resultantly, the proposed scheme provides stable solutions for interference mitigation as well as efficient frequency reuse. Moreover, a simple self-organizing scheme is proposed in [100] in which cell transmit power is minimized in order to utilize frequency reuse patterns efficiently in distributed cellular networks.

To support MTC traffic, authors proposed in [101] various scheduling schemes by considering channel conditions as well as delay constraints, thus to maximize the number of supported devices per cell. In [21, 102], authors proposed clustering techniques for MTC traffic. MTC devices are classified into clusters with similar QoS requirements. Fixed AGTI is allocated to each cluster depending on its priority as well as traffic flow. The advantage of the proposed scheme is the capability to support the maximum number of devices while considering their priorities and delay restrictions. The performance of the above mentioned proposals are rather promising at the expense of high signaling overhead [103]. Additionally, there are several limitations in the aforementioned proposals such as the constant MTC traffic flows were examined (contrary to potentially random traffic). In addition, the mean delay experienced by end users is expected as a performance indicator instead of the probabilistic packet delay and loss rate which are recognized as more realistic performance indicators. The authors of [103] attempted to overcome these issues by optimizing the signaling overhead as well as by considering random MTC traffic. The signaling overhead is reduced by merely considering MTC QoS class identifiers (QCIs) which are known in the beginning of a connection and by avoiding feedback messages regarding the buffer status and channel conditions. However, considering various MTC traffic growth forecasts, efficient radio resource management is still one of the most challenging research tasks and demands serious attention [104]. New and advanced schemes are required in order to allocate radio resources efficiently to handle massive MTC traffic in the future.

In LTE/LTE-A networks, MTC devices contend for PRBs in order to transmit and receive data using an RA procedure. The wireless access schemes can be based either on capillary short-range networks such as WLAN, ZigBee, and IEEE 802.15.4x or long-range mobile networks such as GPRS, GSM, UMTS, and LTE/LTE-A. The wired access techniques are more reliable, introduce small delay, and provide higher throughput. However, it is not suitable for several MTC applications due to various factors such as lack of scalability, cost efficiency, and mobility. In order to cope with these issues, wireless networks play a prominent role. Capillary wireless access is usually used for short-range links and is less expensive, scalable, and requires low power. However, several factors such as low data rates, weak security, higher interference, and less coverage make it less suitable in most of the MTC applications. In order to provide wider coverage, higher data rates, mobility support, roaming services, and security, mobile networks are more attractive and applicable for future MTC [19]. LTE-A is the latest mobile standard which is a potential option to support MTC. Since MTC devices access the network simultaneously and contend for PRBs, this paper focuses on the contention-based RA challenges and possible proposed solutions.

PRACH overload control Per cell, a massive number of MTC devices is expected to access the network simultaneously [19]. Consequently, it causes a low RA success rate and a high PRACH network congestion. Moreover, unexpected delay, increased packet loss, high power consumption, and inefficient utilization of radio resources will occur. In addition, the PRACH will be further loaded when the increased number of devices will attempt to re-access the network after collisions. Therefore, efficient overload control mechanisms are needed to handle future MTC traffic efficiently.

QoS provisioning In a mixed communication environment, MTC and H2H traffic compete for PRACH resources. However, MTC traffic should fulfill its QoS requirements without degrading the performance of traditional H2H traffic [19]. Since MTC devices are capable to access the eNB either directly or using an intermediate gateway, the decision of selecting an eNB or a gateway can be taken based on the signal strength or channel gain particularly in the case of RA (discussed in Section 1.4.2). In addition, this decision should consider QoS performance based on the contention-based procedure. For instance, a device can receive satisfactory signal strength through direct communications with an eNB. However, if the device uses an intermediate gateway, it can potentially increase RA success probability, reduced power consumption, as well as smaller delay. Furthermore, optimum selection of transmission mode should also satisfy MTC QoS requirements such as lower delays, e.g., in the case of emergency/accidental information, and higher data rates, e.g., in surveillance applications. Therefore, the selection of the transmission mode is also crucial. Unfortunately, most of the existing literature does not consider the point selection mechanism for overload control in the RA procedure.

Cognitive M2M communications One of the major issues is the increasing signaling overhead due to the huge number of MTC devices per cell. This increases the possibility of signaling congestion in the network [105]. In addition, there may not be enough bandwidth in order to support and satisfy all devices simultaneously. One of the traditional approaches is to increase the number of eNBs. Furthermore, eNBs supporting MTC and non-MTC traffic should cooperate in order to reduce the risk of network congestion. However, interference between MTC and non-MTC traffic would be a challenging task. Centralized coordination would be effective in order to limit interference. However, it increases the signaling overhead as well as the management complexity. Distributed resource management could be an effective approach in order to limit interference between MTC and non-MTC traffic. Thus, cognitive MTC communication is a promising idea which can be helpful in order to reduce interference between MTC and non-MTC traffic [106]. In addition, the opportunistic random channel mechanisms would be beneficial in order to avoid network congestion. Therefore, enabling cognitive radio capability within an MTC device might be a promising research topic. To handle a large number of devices per cell, random access mechanisms would be beneficial in a way that MTC and non-MTC devices would share the same limited radio spectrum, thus compete for radio resources.

Group-based M2M communication Due to the increasing MTC devices, major challenges are the allocation of RA slots and signaling overhead. In such cases, group-based communication would play a vital role in managing increasing MTC traffic [21]. The primary objective of group-based communication is to reduce the signaling congestion on the air interface. A group head can be selected which collects uplink/downlink data packets, status information, etc. and then forwards the collected information to the corresponding eNB/MTC devices. Since efficient preamble allocation mechanisms are required to avoid collisions, game theory can additionally be a useful tool to optimize preamble allocation [107]. In addition to reduce the network congestion risks, it can also reduce the power consumption of MTC devices. This is due to the fact that machines need to send/receive data to/from a nearby point, e.g., a relay node. In addition, categorizing MTC devices into groups demands special attention as the grouping should be based either on logical or physical characteristics on the basis of, e.g., QoS, jitter, and other MTC traffic characteristics.

In order to control overload in LTE-A networks due to massive MTC traffic, the following solutions have been proposed for PRACHs:

Access class barring (ACB) scheme In a traditional ACB procedure, an eNB broadcasts access probability (AP) and access class (AC) barring time within a cell [19]. When a device initiates the RA mechanism, it draws a number between 0 and 1. The result of the draw is compared with AP. If the result is less than AP, the device can proceed with the RA mechanism. Otherwise, device has to wait for a duration of AC barring. In Release 10 and onward, 3GPP has proposed an extended access barring (EAB) mechanism in which a certain class of delay tolerant traffic cannot perform the RA procedure whenever an EAB is initiated [108]. The eNB can control the PRACH overload by decreasing the AP value. Resultantly, it lowers the RA attempts. However, it might increase the RA delay for certain devices. Furthermore, in a cooperative ACB mechanism [109], an eNB selects the value of ACB depending on the level of network congestion. An optimization problem is considered to balance the number of MTC devices attached to a single eNB. However, the drawback of the proposed approach is that it does not consider the priority of MTC devices. In order to cope with this limitation, a priority-aware RA (PRA) scheme is proposed in [110] which takes into account the priority of MTC devices. PRACHs are pre-allocated to multiple MTC classes by adopting a class dependent backoff procedure.

Resource partitioning mechanism The resource partitioning mechanism illustrates the separation of orthogonal PRACH resources/preambles between H2H and MTC traffic [19]. Two resource partitioning mechanisms are proposed in [111]. In the first method, the preambles are split into two subsets, and each subset is dedicated to MTC and non-MTC devices. In the second method, preambles are again split into two subsets. However, one subset is particularly dedicated to non-MTC traffic, and the second subset is shared by MTC and non-MTC traffic. Once the ACB mechanism is passed successfully, the devices can send preambles using RA. However, the drawback of the proposed approach is that it neglects the eNB selection method and backoff procedure.

Slotted access scheme In this scheme, each MTC device receives dedicated access slots to perform the RA mechanism and, other than this, always remains in sleep mode [108]. Using ID and RA cycle number, each device can calculate its allowable access slots. An eNB broadcasts the RA cycle number which is a multiple of radio frames. The advantage of the given scheme is that MTC devices can share the access slot if the total number of unique access slots is less than the number of devices within a cell. However, the drawback of this scheme is the large RA request delays if the RA cycle length is increased. However, it can significantly reduce the collisions at the cost of large RA request delays. In the LTE-A RA procedure, the influence of an increasing number of transmission attempts on throughput and delay of the slotted ALOHA-based preamble contention is also studied in [112].

Clustering mechanism for MTC In this scheme, each device can be associated particularly with one or more groups. The radio resources are allocated based on the grouping principle, e.g., MTC devices are grouped based on their QoS requirements [21]. Furthermore, grouping can also be performed based on applications or geographical location. In each group, a group head is selected which communicates with an eNB by relaying requests of the group members to an eNB (i.e., peer-to-peer communication). One of the primary advantages of this scheme is to minimize power consumption for MTC devices [113]. Moreover, it significantly reduces the signaling overhead and ultimately minimizes the risk of network congestion.

Dynamic radio resource allocation In this scheme, an eNB allocates PRACH resources among MTC devices on the basis of PRACH overload and overall traffic load in the network [19]. In a PRACH subframe, a part of the subframe is used by the eNB for performing the PRACH mechanism in order to satisfy QoS and delay requirements, and therefore cannot be used for data transmission. Authors of [114] formulated this issue by reducing the number of PRACH subframes. The expected RA delay is kept within a given delay limit. In addition, a self-optimizing algorithm is proposed in [115] based on the 3GPP proposed slotted access scheme. In the proposed approach, an eNB can dynamically change RA slots for transmitting preambles by taking into account the actual traffic load. The advantage of the proposed scheme is that it minimizes the risk of network congestion.

MTC-specific backoff scheme In this scheme, the MTC and H2H RA attempts are delayed separately in a way that a smaller value of backoff time, e.g., 20 ms, is given to H2H traffic and a larger value, e.g., 960 ms, is assigned to MTC traffic [21]. Consequently, this scheme is effective in low overload situations. Whereas in high overload situations, i.e., when more devices perform the RA mechanism simultaneously, this scheme is unable to tackle the network congestion issues. Moreover, the drawback of this proposed scheme is that it does not takes into account the priority of MTC traffic when co-existing with H2H traffic.

Pull-based scheme This centralized scheme handles situations when more devices attempt to access the network using the RA procedure [108]. In this scheme, an MTC server communicates with an eNB and requests to page MTC devices. When an MTC device receives a paging message from the eNB, it will initiate the RA procedure. Based on the PRACH load and spectrum availability, an eNB is capable to control the required number of MTC devices to be paged. However, the drawback of this scheme is the use of extra control channels to page a massive number of devices.

QoS provisioning is one of the major requirements in telecommunication systems. In the perspective of service provisioning, ITU-T defines QoS as a set of all service-related tasks associated with the service provider when a particular service is offered to the end user [116]. Additionally, QoS parameters are mainly used to evaluate whether or not technologies and offered services fulfill the expectations of end users in terms of quality, reliability, and availability [117].

Unlike traditional mobile traffic, QoS class categorization for M2M communications has remained one of the most challenging tasks for operators due to the fact that M2M applications cover a wide variety of use cases. Since mobile networks are expected as a strong candidate to support future M2M communications, there is a strong requirement for defining unique QoS class categorizations for differentiating M2M traffic, e.g., based on priority and delay. Nevertheless, a particular QoS class categorization for M2M traffic has not yet been defined in the literature [117]. Authors of [118] proposed various M2M QoS classes based on the traffic characteristics. On the other hand, 3GPP and ETSI are putting efforts into allocating specific QCI values to each category in M2M communications. In this survey, several M2M characteristics and traffic models are summarized in Table 8 with a provisional QoS categorization depending on the application type, data packet size, required priority, expected delay, and mobility demands. For instance, it can be seen from Table 8 that M2M devices mainly transmit small packets with an expected inter-arrival time distribution ranging from milliseconds to hours. Furthermore, M2M traffic can be classified based on the type of traffic, e.g., real or non-real time traffic. Table 8 depicts various M2M traffic features which can be used to classify M2M traffic into various classes. For instance, emergency alerting, i.e., accidental and/or critical e-healthcare information, is a real-time application which is delay sensitive and therefore demands strict priority. On the other hand, traffic generated by POS terminals is also real time; however, it demands low priority compared to the emergency alerting. In addition, smart metering and monitoring applications generate regular traffic patterns, therefore it can be considered as low priority traffic, but with higher data rate requirements.

In mobile M2M communications, the cost and power consumption of an M2M module are of significant importance to draw a comparison between cellular-based and non-cellular-based M2M communication services [6]. Unlike traditional mobile data traffic such as VoIP and multimedia services which demand high mobility and have no firm limitations for price and battery time of a device, M2M devices demand low price and low power consumption. 3GPP has considered this issue in its Release 12 and introduces a low-cost M2M device category called Cat-0 device. The average price of this newly introduced Cat-0 device is approximately 40–50 % of traditional LTE devices. The complexity of the upcoming MTC modem in 3GPP Release 13 is expected to be 25 % compared to a Cat-1 modem. Furthermore, this new category device called Sub Cat-0 will not support Tx diversity and MIMO operations. Moreover, narrowband RF design supporting 1.4-MHz bandwidth, data rates up to 200 kb/s and transmit power of 20 dBm would sufficiently fulfill MTC application requirements, thus will further reduce device complexity. The key reductions agreed in 3GPP Release 12 and upcoming Release 13 are summarized in Table 9.

In M2M communications, devices have strict battery power constraints peculiarly in battery-driven M2M devices [6]. Moreover, it is inconvenient to provide services in locations without a direct energy source such as water meters, sensors used in intelligent containers, on vehicles for on-board security, on the body of animals for wildlife management, massive number of metering devices deployed, e.g., in a tower. In such cases, it is expected that M2M devices should work for a longer time period (several years) which is indeed a challenging task for module manufacturers as well as service providers. Furthermore, M2M devices spend most of the time in the idle state due to comparatively longer inter-arrival times. The idle state is designed as a low power state in the LTE network in which devices usually sleep to save battery and wake up periodically to check any system information (SI) update or DL packet arrival by listening to the paging message broadcast by the LTE network [119]. In the existing LTE standards, the maximum allowed paging cycle is 2.54 s [120] which is not optimal for M2M communication due to the fact that the inter-arrival time between traffic sessions might be in the range of seconds to minutes. Therefore, consideration of longer paging cycles are under discussion in 3GPP [121, 122]. However, longer paging cycles can increase the delay of DL packets. Therefore, the range of the paging cycle should be selected carefully to keep a balance between delay and power saving efficiency. Figure 15 summarizes the power consumption in different states of a device.

In order to further improve power consumption for M2M devices, several techniques are being considered by 3GPP in upcoming Release 13. Since downlink traffic can be delay intolerant, it is not possible to use longer TAUs. Similarly, power consumption can significantly be improved through coverage extension. This is due to the fact that there is always a trade-off between power consumption and reachability. In addition, Release 13 may address functioning of optimized RRC by optimizing signaling overhead. For instance, a remarkable increase in battery life for two AA-size batteries has been observed from 13 to 111 months by increasing the DRX cycle from 2.56 s to 2 min. The basic idea to improve battery lifetime is to further reduce the activity of the idle mode over the required level for data transmissions/receptions. Thus, the sensors should not have a regular paging cycle which means that devices should only be in active mode if there is any data to transmit.

Due to the tremendous growth in the number of vehicles, transportation and logistics are considered as two of the main M2M communication domains worldwide [123]. ITS enables the exchange of information between vehicle infrastructure and ITS applications through several communication methods and technologies [5]. In ITS, vehicles communicate with other vehicles (vehicle-to-vehicle (V2V)) or with ITS infrastructure server (vehicle-to-infrastructure (V2I)) [50]. The major M2M applications in ITS include collision avoidance, safety and security, parking time management, Internet connectivity, and fuel consumption [3]. In addition, tracking and monitoring of vehicles are the main application areas of mobile M2M communication in transportation and logistics [5, 123]. Several research efforts investigated the support of M2M communications in ITS [5, 124]. Some of the major M2M applications in ITS are discussed as follows:

Collision avoidance and on-board security Collision avoidance and on-board security are the most significant M2M services in ITS. The main purpose of using mobile M2M communication is to deal with line-of-sight (LoS) limitations in order to avoid accidents [125]. In case of an emergency, the collected information is sent to other vehicles or infrastructure within the communication range. To avoid further incidents, communication between the server and the vehicles must be very fast to detect emergency messages and deliver warning messages immediately. Since the response time to emergency/warning messages is significantly small (i.e., in milliseconds), collision avoidance services demand high QoS and low latency. Furthermore, these warning messages are small in size, thus should only be sent in critical situations to ensure efficient utilization of scarce radio resources in mobile networks [126].

Traffic and infrastructure management All over the world, everyday drivers face the problem of road congestion that not only increases fuel consumption which leads to more emission and causes increased pollution but also causes stress for the drivers [127]. Traffic and infrastructure management can play a significant role in handling and controlling the problem of road congestion. ITS tackles the problem by providing bi-directional M2M communications, i.e., between vehicles and ITS server. Vehicles can send status updates about the position, speed, etc. to the infrastructure and can receive information/instructions about road accidents, emergency braking systems etc., from the infrastructure. Therefore, a better managed infrastructure improves the productivity and reduces the factors of cost and pollution in the society. Nevertheless, these applications do not demand high data rates due to the fact that only the abovementioned parameters are mainly sent [128].

The breakdown call (bCall) The bCall service simply starts a voice call and sends the current location/position of a vehicle to an adjacent roadside organization [5]. A bCall service is activated through a triggering operation which is done by the car by pushing triggers. Besides the bCall, an enhanced bCall service provides current diagnostic information along with the location of the vehicle. This allows for a remote diagnosis of faults and enables appropriate and in-time actions to be taken.

Additional M2M services in ITS In addition to the aforementioned applications, M2M communications in ITS support several operations such as tracking of a stolen vehicle, traffic reports and route planning, as well as infotainment services [5]. For instance, to recover a stolen vehicle, stolen vehicle tracking (SVT), service providers request data about the location from Telematic Control Unit (TCU) located inside the vehicle. In addition, drivers are also updated by sending reports regarding traffic in a particular region so that they can change/plan new routes in case of a traffic jam or an emergency. Furthermore, infotainment services aim to provide news/information to drivers and passengers through mobile TV, web-browsing, etc.

M2M in logistical processes Fleet management is one of the major applications of automotive M2M communications in logistical processes [123]. The movements of vehicles, containers, buses, cars, etc. are being tracked regularly through M2M devices which collect data, e.g., of the location, vehicle speed, temperature, distribution progress, fuel consumption, etc., and send this information to monitoring servers. Through regular monitoring, several activities of the system can be performed in an efficient way. For instance, the goods which are transported from one place to another are monitored regularly in order to accomplish in time delivery and to handle any undesirable situation during shipment processes. In summary, M2M communication addresses the following main issues in logistics, (1) Global Connectivity: the cargo moves across several regions/countries and thus must be monitored in order to stay updated; (2) Reporting: with the help of reporting, the exact location of the freight and the conditions of the objects can be easily monitored; and (3) Alarming: in applications such as warehouse management, fleet management, robotics, and control systems alarms are used to detect critical or emergency situations. Besides, M2M communications in logistics can provide additional services such as decreased operational cost, high inventory flexibility, increased supply chain visibility, and reduced loss of vehicles and containers [129].

ITU in 2011 reported the increasing use of M2M services for medicine and electronic health records in e-healthcare [130]. The applications of M2M communications in e-healthcare include, e.g., remote monitoring of patients, provoking of alarms for handling serious conditions, and remote control of certain medical parameters. A patient is equipped with numerous M2M sensors which collect several health indicators of a patient, e.g., blood pressure, temperature, and heart rate. These devices transmit collected data to the monitoring centers such as hospitals, and clinics which store the recorded data and reacts accordingly [69].

Generally, the recorded data is mainly sent to medical centers by using WLAN technology, which is reliable for both the patients and medical control units. However, the lower coverage areas of WLANs may generate problems for the patients in case of mobility outside the home. If the patient is outside the home, e.g., for exercising, the regular monitoring of the patient is not possible due to the limited WLAN coverage. Consequently, in case of any unusual situation, the patient will not be able to get emergency aid by the medical center. In these situations, there is a need to extend the coverage and high mobility support features using mobile networks such as LTE-A. Mobility is one of the fundamental requirements of many M2M applications such as e-healthcare, ITS, and logistics. In this part, some of the major e-healthcare applications are discussed as follows [69]:

Smart metering and monitoring are considered to be one of the main driving forces for increased M2M traffic [131]. Moreover, M2M communications play a vital role for improving the performance of the smart grid which is used to improve the performance of energy generation, distribution, as well as consumption. Smart metering and monitoring are considered to be the major elements used in a smart grid. Devices are deployed, e.g., at the residential area for monitoring of the data, e.g., electricity, gas, water consumption, monitoring and controlling utility consumption, tracing, tracking and protecting people, and assets or animals. Following are the major use cases of smart metering and monitoring [67]:

Pre-payment is one of the metering applications of gas, electricity, water, etc. With pre-payment, a person/family can buy a particular volume of gas, water, petrol, etc. This information is securely sent and stored within an M2M module. Furthermore, the actual and accurate amount of energy consumed during some particular time is sent to an M2M device. In addition, whenever the household consumes the acquired volume, the actual readings of the used energy are transmitted using an M2M module. When the limit of the purchased volume is reached, warnings in terms of messages or indicators can be sent and the supply can also be stopped.

Most of the existing POS terminals are connected through wired networks, and therefore fixed at some certain positions, e.g., in shopping malls, restaurants, super markets, railway stations, and airports. If a person is interested to use a POS terminal, then he has to visit the POS terminal. This might cause inconveniences for the users [67]. On the other hand, remotely located POS terminals (e.g., ticketing machines, and parking meters) require a wired connection which is not cost efficient and also difficult to be installed. In addition, this may also be exposed to damages. The other option is to take advantage of mobile networks to connect POS terminals. However, this approach poses some security limitations. Introducing M2M devices enable additional possibilities for applications. For instance, online transaction with credit/debit cards and secure communication can be offered with the help of M2M devices which can be deployed in, e.g., parking, stations, airports, and e-ticketing terminals. Typically, these devices have to fulfill several security requirements, e.g., for financial transactions.

M2M services provide improved intelligence in industrial plants, robots, furnaces, etc. Sensors and actuators are deployed which can collect and send information to the control room. For instance, if an emergency signal is detected by the sensors, it can be sent immediately to the relevant server. As a result, immediate remedial actions can be taken. Some of the major M2M applications in control systems are discussed below:

Controlling vending machines Due to the increasing product cost, diverse customer choice, increased demand for stock availability, and advanced regulations make a strong competition between operators of vending machines [132]. Vending machines are generally placed inside train stations, offices, public buildings, tourist spots, etc. Nowadays, people are recruited and dedicated to the maintenance of these vending machines [67]. They regularly visit the vending machines to perform several tasks such as checking the fill levels, re-fill if required, maintenance, as well as repairing damages. With the help of M2M technology, additional optimization possibilities are feasible, as a result vending machines can be managed and monitored in an efficient manner. By using devices connected with a mobile network, the current status such as details of the current stock, frequency of customers, damages, and malfunctions can be sent to monitoring centers, e.g., a quality control and maintenance team. In addition, devices can also send updates, e.g., about price variations and thus perform remote maintenance.

Controlling production machines Such machines are usually located inside the production sites which can be a harsh environment for production machines [67]. Dedicated people are hired to perform various tasks related to production machines, e.g., repair and maintenance. They visit production machines on a regular basis and perform several assignments such as repairing damages. The improved performance and maintenance of production machines can be achieved with the use of M2M communications. Thus, machines would be able to communicate with each other, and subsequently send updates to the respective control room. This can significantly increase visibility of the performance of production machines. Besides, in the case of damages or malfunctions, quick identification, and repairing tasks can be performed.

M2M services in compensation applications include the following:

M2M communications play a vital role in home automation to improve energy consumption, security, etc [134, 135]. One of the major focuses of home automation with M2M devices is to enhance energy consumption through regular metering and monitoring. For example, M2M devices can be deployed in a room to know about the presence of an individual. In the case, if there is no person present, the lights can be automatically switched off to save energy and ultimately to reduce costs. In addition, water and gas meters can be monitored and regulated using M2M devices. This will significantly improve energy consumption by increasing or decreasing the level of heat or water, e.g., in a room, office, stations, and shopping malls. M2M devices can be connected wirelessly to a gateway or directly to an eNB. Furthermore, the notion of heterogeneous technologies can also be used in such applications. It can receive data, merge with the other context information from other devices and send instructions to corresponding actuators.

Mobile M2M devices must use low-power consuming technologies in order to increase their lifetime. Consequently, the customer does not need to change batteries repeatedly. This is also needed for practical reasons, e.g., when M2M devices are installed in some places which are very difficult to approach. These applications can additionally help to inform customers about their energy consumption on a regular basis. This can be achieved with remote monitoring services, i.e., a customer/user can know about the amount of energy consumption even when the user is away from his home. This can also help to notify the user of any detected peculiarity (e.g., leakage of a gas). Consequently, immediate safety measures can be adopted.

Due to the distinctive characteristics of M2M traffic, modeling the behavior of different kinds of M2M applications is one of the major requirements of M2M communications in the future. One of the possibilities is modeling each M2M device on its own. However, there are several challenges that need to be overcome. For instance, it is very difficult to model the behavior of traffic being generated from a massive amount of devices in parallel, and also it becomes more difficult in the presence of strong spatial and temporal correlation between devices [25]. On the other hand, aggregated traffic modeling defines M2M traffic as one stream from multiple devices. Nevertheless, the aggregated traffic models demand additional intermediate gateways between devices and eNBs. In [25], a coupled Markov modulated Poisson process (CMPP) framework was introduced, mainly targeting the source traffic modeling in M2M networks. Furthermore, the authors of [138] introduced the Beta/M/1 model for M2M communication considering beta-distributed inter-arrival time for the massive M2M traffic. However, at the same time, the authors stated that a G/M/1 queuing system may be more appropriate to model the behavior of different kinds of bursty applications in M2M communication. In addition, it was shown in [139], that Markov modulated Poisson process (MMPP) models are difficult to fit because they have a large number of parameters. Therefore, it is very difficult to solve them analytically. Since M2M traffic mainly exhibits periodic characteristics, stochastic models, e.g., the one given in [140] for modeling periodic traffic in an IP based system, can be developed to analyze the behavior of M2M traffic. Nevertheless, until now, there has been very little effort towards selecting the most suitable queueing models that can be used to represent the behavior of M2M traffic. In the future, there is a strong need for developing appropriate traffic models for the better understanding of M2M traffic characteristics.

As discussed in Subsection 1.5.3, the density of M2M devices per cell is expected to be enormous. Therefore, when these devices will try to access the network concurrently, this will lead to a low RA success rate and high network congestion in the PRACH. Consequently, this may cause unexpected delays, increasing packet loss, inefficient radio resource utilization, extra energy consumption as well as service intervention. Moreover, the random access channel can be further overloaded when these devices repeat their attempts to access the network after collisions. Thus, efficient access methods and overload control mechanisms are required to perform efficient RA procedures. Additionally, the initial access procedure in LTE-A networks consists of numerous steps which cause a significant amount of overhead and latency. Authors in [17] further stated that the state-of-the-art access procedures are not appropriate for M2M traffic. Either evolutionary (simpler) access schemes should be standardized for M2M communications or a revolutionary approach, i.e., fully separated standard LTE- and M2M-based data and control channels are required.

Power consumption of an M2M device is of significant importance to compare cellular- and non-cellular-assisted M2M services. Particularly, battery driven M2M devices demand low power consumption to increase their operating (life) time (as discussed in Subsection 1.5.6). Since M2M services are expected to be one of the potential sources to increase revenue of operators, it is important for module manufactures and service providers to design and ensure efficient mechanisms to reduce power consumption for M2M devices. One of the possible approaches can be that of investigating existing battery life models for UEs as well as M2M devices and designing an optimized battery life model based on the existing models and by considering more realistic M2M traffic characteristics.

Besides battery life modeling, optimized power consumption mechanisms can significantly increase operating time of an M2M device. For instance, extending sleep mode as well as longer TAUs can be effective to ensure longer battery lifetime. Moreover, optimized RRC, i.e., by optimizing signaling overhead can also be an efficient way to improve power consumption. In general, reducing idle mode activity can also be significantly helpful to improve battery lifetime.

Existing mobile standards are particularly designed to support broadband traffic. Therefore, LTE TB sizes are only optimized for typical VoIP traffic. These mobile standards must be revised in order to support the enormous M2M traffic in the future. As mobile radio resources are a valuable asset, these limited radio resources must be efficiently utilized. Since the capacity of a PRB can be significantly larger than the need of typical M2M application (i.e., under favorable channel conditions), allocating 1 PRB to a single M2M device can notably degrade radio spectrum utilization. Hence, alternative schemes such as data aggregation and multiplexing are required for M2M communication. Aggregation and multiplexing of small M2M packets can be achieved by using an intermediate node between devices and eNB. Since relay nodes are one of the latest features of LTE-A networks, they can be used to aggregate traffic from multiple devices. In this way, small packets from different M2M applications can be added to make a comparatively larger packet to maximize radio resource utilization. Besides, it can significantly reduce the risk of network congestion due to the fact that not all devices in the vicinity of the relay node but only a relay node will communicate with the eNB.