Advertisement

Dynamic Spectrum Access for Machine to Machine Communications: Opportunities, Standards, and Open Issues

  • Luca Bedogni
  • Marco Di FeliceEmail author
  • Luciano Bononi
Living reference work entry
  • 130 Downloads

Abstract

Cognitive radio can be applied to a multitude of domains, one of which is M2M communication. Specifically, M2M communication refers to communication between devices without human intervention. Hence, devices should be able to organize themselves and run the communication protocol autonomously. If cognitive radio is used, tasks such as dynamic spectrum access (DSA), spectrum sensing, and alike present additional challenges compared to traditional network, as all the decision framework should be implemented and automatized in the devices. In this chapter, we focus on DSA techniques for M2M. The main difference from other kinds of communication is relative both to the energy efficiency and to the low protocol overhead, as devices should run for long periods of time and run without human intervention. At first we present related work from literature, categorizing the different tasks devices which want to leverage DSA on M2M have to perform. At the end of the chapter, we present a proof of concept of a general framework, which can be applied to different scenario concerning M2M, encompassing all the spectrum management and measurement tasks M2M devices should generally perform. Finally, we derive open challenges and future research directions concerning this scenario.

Keywords

Dynamic Spectrum Access (DSA) Spectrum Sensing Spectrum Management Spectrum Sharing Mobile Data Collector 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

Machine to machine (M2M) communication refers to interactions which occur between devices with no human intervention. M2M has gained attention mostly thanks to the Internet of things (IoT) paradigm, where a multitude of devices communicate together with little to no human intervention. M2M communication inherently has such unique characteristics which differ from standard communication and is typically used on resource limited devices, hence posing additional constraints on the communication.

Over the years, many proposals, technologies, and standards have been proposed and are used to realize the M2M paradigm. However, such a plethora of different technologies also raises challenges on the interoperability and ability to communicate. The risk is to create the so-called intranet of things, or islands of things, meaning that devices are able to communicate only within the same ecosystems.

This heterogeneous scenario, in which different manufacturers and consortiums develop different technologies, is due to the extreme diverse scenario in which M2M communication can take place. These can span from healthcare scenarios, in which M2M technologies should provide reliability in the first place, to video surveillance installation, in which also the data rate is an important aspect. Productive scenario, such as predictive maintenance, often requires long operational times; hence, energy efficiency is key. All of the above has contributed in creating a vast ecosystem, in which different technologies share some characteristics and differ in other.

Besides the aforementioned scenarios, also cognitive M2M technologies can play an important role. Basically, cognitive networks sense the environment and reconfigure their communication parameters in order to better adapt to the scenario and possibly communicate. Among all the facets of cognitive networks, one of the most studied and investigated is certainly dynamic spectrum access (DSA), through which cognitive networks can transmit in the so-called spectrum holes, that is, part of the spectrum which are currently free and not occupied by others. Clearly, DSA requires extra computation on the device side, hence requiring more performing hardware and possibly reducing the energy efficiency.

In this chapter we present the different technologies, standards, and applications of M2M technologies. We also derive potential challenges and future research work on the topic.

Use Cases

In this section we review the applications that would benefit from cognitive M2M communication. In general, cognitive M2M communication helps deploying services in specific scenarios. For instance, M2M communication deployed in TVWS is beneficial for applications requiring an extended range compared to classical wireless solutions and for applications requiring good propagation through obstacles like building walls. This is especially interesting for indoor/outdoor networks, in which nodes are placed inside homes and need to communicate between each other as well as with other buildings. For instance, the new smart grid network is certainly a good representative of this category, as nodes should gather the consumptions of the home and balance the load of the network by aggregating the data coming from different sources.

Smart Grid Communication

One of the most prominent and most investigated topics regarding M2M communication in general, and specifically with TVWS, is certainly smart grid communication. Nowadays, and with a foreseen increase in presence over the next years, utility meters will be deployed in homes to report consumptions and balance the load. The scenario is that of multiple sensor devices deployed in homes, which monitor and report loads either to an in-home aggregator or to the utility operator. The former is for automatic balance of the consumptions locally; the latter is to enable the operator to balance the network.

As said, TVWS have been widely studied specifically for this scenario, due to their characteristics which make them appealing for such a deployment. The possibility to cover large distances, even if the meters are located in the basement [4], and the reduced energy consumptions to send data locally [3] are just two of the characteristics of TVWS that are interesting for this specific scenario.

Specifically in [59] it has been firstly seen in the term cognitive M2M (CM2M), to enhance the flexibility and the reliability of M2M communication.

There is a natural bind between devices needed in a smart utility network (SUN) like the smart grid and the TVWS device classes. In [49] it is highlighted this, which candidates TVWS as one of the most prominent technologies to realize these kind of networks. They identify four classes of devices needed in SUNs, namely: – Utility provider base station – Data collector – Utility meter – Mobile data collector They bind to the different classes of TVWS devices foreseen by regulations. The base station is clearly the static device, which is able to transmit at high power (up to 4 W) and thus covers larger distances. The data collector and the utility meter are, respectively, Mode II and Mode I devices. They do not need sensing capabilities, which are instead required by the mobile data collector, carried by operators to gather data from meters. TVWS would be beneficial for SUNs thanks to the larger coverage range, which make them attractive particularly in rural areas.

This is also the path taken by [10], which present a test-bed deployed in Scotland specifically focused on smart grid communication. The motivation leading to the use of TVWS is mainly economical: more expensive systems such as fiber optic or DSL are not economically viable in rural areas, where few people live. Instead, TVWS are more economic and explored for this deployment, following results in [46]. The underlying protocol for this network test-bed is 802.11, which runs in the UHF bands instead of ISM bands over channel 57 (758–766 MHz). Another interesting feature, going toward the energy efficiency required by M2M communication, is that of a wind turbine for power supply mounted on top of the base station, making it autonomous from an energy supply. Wind power is an attractive option for Scotland, and clearly other options should be explored in other countries and environments, where wind may not be the most efficient method of energy production. The work presented in [10] is one of the first aimed at deploying an M2M network directly on the field.

Another work which studies the deployment of smart grid communication in TVWS is [4], which presents both a measurement study to deploy smart grid networks in TVWS and TV gray space and field results on the indoor propagation of narrowband signals in the TVWS. What emerges from the study is that the superior propagation characteristics of UHF bands are able to deliver the signal much farther compared to classical ISM bands. Moreover, it is interesting to note that if deployed indoors and at a low floor such as in the basements, the DVB-T receiver can easily compensate possible interference coming from the smart meters, thanks to the high directionality of its antenna.

The battery savings are presented in [3], where the authors foresee clusters of smart meters that communicate together through TVWS and one of them eventually reports the load measurements to the remote utility aggregator. Here the main advantage of TVWS networks relies in its duality: it can be used as a local communication technology, using low power but still penetrating obstacles better than technologies in higher-frequency bands. When needed, for instance, when the meter has to report the data to the aggregator, it can increase its transmitting power to cover a larger range.

Sensor Networks

Apart from the specific case of smart grid communication, as we already mentioned, M2M networks can also be built with the focus on energy efficiency. This allows building sensor networks in general with battery-powered devices, for a range of possible applications:
  • Environmental monitoring: through TVWS networks it is easier to build large-scale networks for battery-powered devices, which by transmitting in narrowband channels can reduce their power consumption keeping a satisfactory transmission range.

  • Industrial monitoring: wireless sensor networks have been deployed in several machines and automated tools, mainly for machinery condition-based maintenance. The use of M2M in this domain would help reduce the problems when machines are placed in hard-to-reach places, and thus shadowing effects could reduce the signal to undetectable levels.

  • eHealth: in eHealth, devices could be either implanted in patients or deployed as wearable computing devices. They report the patient health with important parameters such as the heart rate or the blood oxygen level. Doctors can then monitor the patient remotely and be alarmed whenever health parameters fall below or rise above a predefined threshold.

ITS

Another interesting application of M2M communication is related to intelligent transportation systems (ITS). Here, cars report their state and gather information about the road situation, to possibly inform the driver of alternate paths in presence of road works or accidents. An important aspect is that vehicular communication happens at the ground level, and thus possible interference to the primary receiver is naturally mitigated.

Messages in safety VANETS are typically of two kinds: Emergency safety messages (ESM) and periodic beacon messages (PBM) [37]. The aim of the work is to show how to deliver ESM messages faster than classical dedicated short-range communication (DSRC), thanks to better propagation and penetration into other cars and buildings. Examples of ESM messages are stopped cars on blind turns, cars driving in the opposite way, or cars approaching too fast. Given their nature, it is straightforward to understand how reliable and fast the communication has to be. The classical approach for DSRC communication would be multihop, with vehicles relaying the message to eventually reach farther drivers. Although it is possible to enhance the IEEE 802.11p to meet stringent delivery requirements [17], it would still be challenging to guarantee the reception of the message to all the interested cars with such a congested network. Moreover, multihop communication raises the well-known problem of the broadcast storm, in which too many relays could interfere with each other and reduce the communication performance. Thus, the benefits for V2V are mainly in the longer transmission range, which make them ideal to deliver safety messages. At the same time, the challenges reside in the spectrum availability. ESM messages cannot depend solely on opportunistic communication, since in areas with scarcity of spectrum, or when the licensee is using its spectrum band, the ESM messages should still be delivered. So the main idea is to use opportunistic communication when available, but switch back to classical DSRC when this is not possible. Another issue to consider is about the time needed to get the available channels on which the message should be delivered. For classic DSA protocols, the key challenge is to achieve this information rapidly, as the ESM message cannot be delayed. For the case of TVWS, the car should get the list of available channels from the remote spectrum database in advance, since this cannot be done when the emergency occur, so that the client already have to send the message; otherwise, it would incur in too much delay. So clients should ask in advance to the database or other clients what are the channels available on the road they are traveling [16]. This would reduce the time needed to send the ESM, as the car already has the channel list in which it can transmit when needed.

Studies in literature have also demonstrated the feasibility of the aforementioned approaches, like [2] and [29]. In [2] the test-bed is built in Massachusetts, while [29] is deployed in Japan. Both are realized thanks to the GNU radio platform and show in practice how it is possible to build VANETs using TVWS.

When we consider M2M communication for cars, it is also important to commonly agree on the protocol to be used. For vehicular communication, the IEEE standardized the well-known IEEE 802.11p protocol, in which are foreseen two periods of time: one is used to transmit safety messages over a common control channel (CCC) and the other one to offer services such as infotainment. One of the issues with 802.11p, analyzed in [17], is about the timely delivery of messages, particularly those containing safety information. A possible solution is to rely on licensed communication for mission critical messages and rely on DSA for all the other messages, such as control information. This also goes in the direction foreseen by [37], which proposes to use TVWS to transmit safety messages which could then reach farther cars and possibly use DSRC communication for critical information. Another possible protocol proposed for V2V communication is 802.11af, studied in this particular scenario in [60]. 802.11af for V2V is certainly feasible and an interesting option, which could also be used together as 802.11p, by using DSRC and 802.11af together, as we already stated before.

Industry 4.0 and IoT

An area which is not much explored but that can certainly represent one peculiar application scenario for M2M in TVWS is certainly the upcoming Industry 4.0. In this environment, machines and devices are able to talk without human intervention, to offer new and enhanced services, ranging from predictive analytics to enhanced services for the end users.

Industry 4.0 refers to the plethora of tools, technologies, and methodologies needed to revolutionize normal industries with enhanced technologies, such as wireless sensor network and big data, which will open up services which would had never been possible. For instance, wireless technologies should be adapted to this changing environment, which will likely have multiple devices running different standards, which need not to prevent others from working [53]. Here, multiple gateways, able to speak different wireless protocols, are envisioned, so that they can bring messages from one network to the other [15]. For instance, access to users can be provided either through Ethernet or WiFi; however, this is not feasible for battery-powered devices, which might also be located in hard-to-reach places [42].

Another scenario which hugely benefits from M2M communication is predictive analytics [34], defined as the ability for a system to monitor its assets and forecast the remaining life of each one, hence optimizing maintenance and reducing downtimes. This is done by monitoring components and through big data analysis and machine learning techniques recognizes unusual behaviors, such as never seen before vibrations, exceeding threshold values, and so on. Then, maintenance may be performed to multiple components at once, reducing the costs and arguably maintaining machines in better health. In fact, the cost of replacing a component before its end of life is generally much lower than having the machine to stop for a period of time, waiting for that component to be replaced.

To realize the IoT vision, it needs a technology that glues together different networks together, like the TVWS, as IoT networks can be locally sparse. For example, smart crossing lights can broadcast their status to nearby cars, which in turn could dynamically update the travel time, to proactively activate smart objects at homes. What typically happen is that these communication links can be realized through different technologies. In the previous example, the crossing light might transmit locally using WiFi, the car could use the cellular connection, and the messages are delivered in the smart home through a DSL connection or fiber optic. In this heterogeneous scenario, TVWS is one of the candidate technologies to possibly offer a bridge to all these devices, since they could propagate the messages farther than other technologies and can thus deliver the messages to more devices. Crossing lights would then be able to talk between each other and disseminate locally not only their status but also on the crossroads, so cars can infer more information.

Even if this is only an example, it exemplifies the potential of M2M communication in TVWS. Longer range means potentially more devices to be connected and from which obtain and give data.

eHealth

Another interesting domain for TVWS M2M is that of eHealth services, in which patients are monitored through wireless devices, and enhanced medical services can be provided also in hard-to-reach areas. TVWS network can provide useful services such as patient monitoring and locate them in the premises of a hospital. In addition, TVWS networks can also realize the remote patient monitoring, so that people can still be monitored at home, by constantly communicating with the doctor, without the need for human intervention, which can be warned only when needed, by setting for instance appropriate triggers. Moreover, many medical devices rely on the wireless medical telemetry system (WMTS), which operates in similar bands of TVWS. Therefore, they would need little intervention to be adapted to also use a larger spectrum band. However, there are several interference issues that need to be carefully studied, since medical devices would compete also with other, consumer-like devices in the same bands.

It is worth noting the ongoing development of an eHealth test-bed in the Philippines [27], aimed at connecting 100 sites in rural areas of Philippines. The goal/plan initially is to use TVWS as a public service allowing for connectivity for education, eHealth, eGovernment services, environmental sensor networks such as those used by PAGASA and DOSTs Project NOAH, as well as Internet access in public places such as halls and town plazas. The trial is ongoing, proceeding slowly, mostly due to a precise policy missing.

A similar trial is ongoing also in Africa [29], where Microsoft is partnering with the Botswana-University of Pennsylvania, to provide medical services in poorly connected regions. They selected TVWS since it represents a less costly, faster, and farther-reaching Internet connection that is a promising option for connecting the previously unconnected populations of remote and underserved areas.” Thus, TVWS are again selected as a candidate technology to practically realize services where the communication range is an issue and also because they represent a cheaper technology, compared to other well-known wireless standards. In addition, it is worth to note that once TVWS will roll out, the economy of scale might decrease even more the price of the devices, as more TVWS chips will be produced. Also in Bhutan TVWS are currently investigated to provide last mile connectivity, particularly for eHealth services, as patients incur in high costs to reach healthcare centers.

Apart from human medicine, TVWS networks are also currently investigated for animal treatment and monitoring. In particular, recently the London Zoo partnered with Google [38] to monitor endangered species through the use of a TVWS network. Moreover, animals can also be viewed online through video streaming by people through the Internet. TVWS have been selected as the ideal technology as animals can be hidden behind trees or in caves, and therefore the carrier wireless technology in use should be resilient to environmental obstacles. In areas such as rainforests and deserts, where animals can travel long distances and where therefore it is difficult to monitor animals for long period of times, TVWS can definitely fill the technology gap by meeting these needs. The test-bed has been rolled out in the London ZOO, and it is planned to extend it also to wider areas. Apart from wild animals, there is also an interest in TVWS for livestock monitoring [51] through the use of smart geofences, which can monitor animals inside a huge area, thanks to the propagation characteristics of TV bands.

M2M Operational Characteristics

In this section we present the details of M2M characteristics and how these are tackled by cognitive networks.

Generally speaking, M2M communications are typically referred to small and battery-powered devices, although this is not always true. For instance, video surveillance cameras where it is also performed image recognition, either on the edge or on the cloud, can be still seen as M2M, as no human intervention is performed in the middle of the communication. However, it is also true to say that most of the M2M deployed scenario envision small, battery-powered devices which report different kind of data to a central entity, in charge of analyzing them and inform the users. Clearly, the number of intermediate levels can change, as intermediate computation may be performed to achieve network optimization.

Focusing only on the communication slice of the picture, we define the following operational characteristics of M2M communication:
  • Self-instantiated: no human intervention should be envisioned, like checking connectivity or acknowledging any sort of data in order to establish a new M2M connection.

  • Self-healing: in case of network disruption, such as nodes that depart the network either voluntarily or because they have run out of batteries, the network should be able to heal itself.

  • Battery efficient: although we mentioned that M2M communication can be in main powered devices, M2M communication should be designed to be energy efficient. Hence, M2M devices are typically sleeping and wake up only upon certain events. Hence, it should not be assumed that they can receive a message, and optional device configuration should be performed by the device station in reply to events on which M2M devices wake up.

  • Lightweight: M2M protocol should also be designed to carry the minimum amount of protocol overhead possible, as this would require additional time and computation to be achieved, such as CRC an alike.

  • Multiple transmissions: unlike human communication, M2M may perform many, simultaneous connections. Coupled with the battery efficiency requirement, this involves developing novel techniques to access the channel by M2M devices. Imagine the Industry 4.0 scenario, in which thousands of tiny devices installed report sensed data to a central entity. Clearly, these devices should access the channel fast, and the high number of them poses technical challenges on how this should be performed.

  • Heterogeneous kind of traffic: M2M traffic can be characterized in two different shapes: bursty or sporadic. Bursty traffic refers to devices which typically sleep and when they wake up send a large amount of information. This may be surveillance cameras which perform image recognition locally and, only when something which requires additional analysis has to performed, send that information to a central entity. Sporadic refers to anything which is not bursty and comprises periodic beaconing of information to a central station.

  • Security: although not a specific requirement of M2M communication, security can take a dual meaning in M2M communication. At first, there is the communication security, which may mean encryption. However, this should also be tailored to battery-powered devices; hence, it should not require excessive computation power. There is also the security of the device itself, since they can be stolen, and hence they should be able to sense unusual behaviors and events and report those to the central entity.

  • Latency: depending on the scenario, latency may be or not a constraint. For critical scenario, such as eHealth, clearly latency becomes a more severe constraint. However, it is also worth to mention the latency when data has to be synchronized through different M2M devices, for instance, for predictive analytics, when considering Industry 4.0 scenarios or of building structure monitoring. There, latency becomes of paramount importance, to determine the sequence of events reported by different devices.

It is also worth to say that cognitive networks should be carefully tailored to meet M2M communication. For instance, tasks such as spectrum sensing and spectrum decision, which are normally battery consuming, have to be reengineered to meet the constraints this scenario poses. For this reason, among others, dynamic spectrum access and cognitive networks have a narrow area of application for M2M. In fact, the overhead needed by cognitive networks is not particularly suitable for battery-powered devices, which may waste considerable amount of energy to perform network maintenance and access tasks. In fact, in the M2M domain, typically network operations are performed vastly on a remote central unit, and less consuming tasks are left to the edge of the network, formed by the end nodes.

A fundamental difference which has to be made between different M2M scenarios is about the operational range. M2M scenarios can be built either for short-range or long-range communication, and several technologies and protocols have been proposed for both. In the following, we outline some of the major differences between the two and protocols and standards which have been proposed.

M2M Range

Short-Range M2M

Short-range M2M technologies are historically used to build wireless sensor networks (WSN), due to their ability to form networks of devices in close to each other. Certainly, one of the most adopted technologies in this field is IEEE 802.15.4, although more recently also IEEE 802.15.1 (Bluetooth Low Energy) has gained interest.

Among the IEEE 802.15.4 family, it is interesting to discuss the IEEE 802.15.4m amendment, which is the use of IEEE 802.15.4 on the TV white space (TVWS) bands.

TVWS are frequency bands which are progressively becoming available for secondary user operations, thanks to the digital TV switchover. Basically, thanks to the transition from analog to digital TV, several channels are not used anymore and can possibly be exploited for other use cases. TVWS have already shown interesting performance in scenario such as communication in rural areas, thanks to the low-frequency band used which helps in achieving a large communication range. Many standards have been proposed to work in TVWS, like IEEE 802.22 for regional networks, IEEE 802.11af for indoor, WiFi like deployments, and IEEE 802.15.4m for M2M communication and more oriented to IoT and WSN.

At the moment of writing, not all the countries in the world have regulations to transmit in TVWS. It is worth mentioning the US case, in which the FCC already published the rules to access TVWS, and the UK case, where Ofcom followed a similar path. Although they differ in details, they share the fundamental building block of the TVWS architecture, which is the remote spectrum database. Basically, it contains the positions of the TV transmitter, and through path loss models, it determines in which areas a given channel is free or occupied. Devices which need to access the TVWS should firstly query the remote spectrum database and ask for free channels, by providing their position. The remote spectrum database will eventually reply with a list of available channels on which the device can transmit, since they are free at that moment.

IEEE 802.15.4m has been proposed in 2014 [7, 28], and the major differences over the standard IEEE 802.15.4 standard are devoted to the physical layer, with the IEEE 802.15.4m defining three novel PHY layers, an FSK PHY, an OFDM PHY, and a narrowband OFDM PHY (NB-OFDM). They are designed to operate on TVWS frequency bands, up to 862 MHz.

The OFDM layer, which is the one which offers the highest data rate among the available ones, offers six different modulation and coding schemes (MCS), ranging from around 390 kb/s with a BPSK modulation to a maximum of 1562.5 kb/s, when using a 16-QAM. Note that these values can be increased up to four times when using channel bonding. The study in [7] summarizes the practical use of IEEE 802.15.4m, finding that in practical scenarios free of noise, the highest throughput can reach around 30 m, with a transmitting power of 20 dBm. Clearly, this can be increased if higher transmission powers are allowed or when using stronger MCS which require lower link budget.

Concerning dynamic spectrum access, IEEE 802.15.4m should ask to the remote spectrum database the available channels. However, [32] finds that when using bursty transmissions, which are typical of M2M scenarios, TV receivers better compensate for interference. Hence, [7] shows that it is possible to transmit small packets even on occupied channels, without causing interference to the primary user of the channel.

Long-Range M2M

While short-range M2M achieves a maximum transmission range of tenths of meters, long-range M2M foresees to connect devices which are located several kilometers away. This is generally done either by increasing the transmitting power or by reducing the communicating band. Recently, a lot of standards have been proposed and are now used in different scenarios. These include LoRa, LTE-M, and Weightless.

LoRa is a technology developed by the LoRa alliance and is rapidly becoming one of the most used standards in the M2M and IoT world. Its network topology is a star of stars, which means that each device is directly connected to a LoRa gateway, which can communicate to the central entity, called the LoRa NetServer, which also handles the authentication and association of the devices. Basically, the LoRa gateway is simple, acting mostly as a relay, since the complexity is moved to the LoRa NetServer. Hence, features such as mobility and information redundancy elimination are inherently supported, as the NetServer manages them all. Concerning the data rate, LoRa devices can reach a peak of 50 kbps, transmitting at a maximum power of 24 dBm.

LTE-M is an LTE version which enables machine type communication, particularly for constrained devices. This technology has the clear advantage of relying on a vastly deployed technology such as LTE, and transmitting in smaller bands enables longer range and battery savings when reducing the transmission power. This would also enable hybrid devices, which might leverage on the full LTE stack when needing bandwidth hungry data transfer and switch to the more efficient LTE-M for smaller communication. The 3GPP studied LTE-M in the area of Washington [1], achieving a maximum of more than 10 km. NB-LTE-M further improves the capabilities of LTE-M, by reducing the bandwidth to 200 kHz and allowing a maximum transmitting power of 23 dBm instead of 20 dBm like in LTE-M.

Finally, Weightless is a set of technologies originally proposed by Neul and now developed by the Weightless Special Interest Group (SIG). It is composed basically of three different technologies, namely, Weightless-N, which uses a narrowband transmission scheme, achieving a maximum range of 5 km. For bi-directional communication there is Weightless-P, although this reduces the maximum transmission range from 5 to 2 km. Finally, Weightless-W is based on TVWS, which tops the other two in terms of communication range, achieving a maximum of 10 km.

Dynamic Spectrum Access for M2M

The need for dynamic spectrum access (DSA) comes from the scarcity of the spectrum resources, which is also the basis for the birth of cognitive networks in general. Basically, we are living in a world where almost all the spectrum are statically allocated to different services. In [21] the FCC shows the current allocation of frequency spectrum up to 3000 GHz, and it is possible to note how little space is left for new services. Apart from the ISM bands such as the 868/915 MHz, 2.4 GHz, and 5.8 GHz where different services take place, it is challenging to license new bands for other purposes. However, a lot of studies have shown how little licensed space is used. We mention [39, 44, 55] for the European-related measurement studies, [14, 41, 50] for the USA, [30, 56, 57], for Asia, and [40] for Africa. Clearly, measurement bands differ, and also the purpose of the study varies. However, there is a global consensus that some licensed bands are heavily underutilized. As new services are deployed, which build upon new technologies, spectrum is a scarce hence valuable resource. With these premises, cognitive networks are born, and more specifically the building block which takes into account the so-called spectrum holes is named DSA [36].

DSA can be mainly implemented through two different architectures: centralized DSA or distributed DSA. In the centralized DSA architecture, it is foreseen as a server which allocates the radio resources to clients, while in distributed DSA the devices compete or cooperate together to access the spectrum. Contention can occur, for instance, through a control channel, where devices declare their willingness to transmit and reach agreement together.

Certainly, one of the major challenges to tackle is the fairness between clients. While such achievement might be easy to obtain in centralized DSA, as the remote server has a global view of all the clients and the air time allocated to them, achieving fairness in distributed DSA is more challenging. One possible solution would be to gain access to the spectrum proportionally to the sensing time [35]. If a client senses the spectrum more, hence gives to the neighboring devices more information about the spectrum state, it should be given more opportunities to communicate. However, in the domain of M2M communication, where devices are typically on batteries and with low duty cycle, this technique is clearly inefficient. This can be mitigated using solutions like the one presented in [26], where forced vacation periods for devices must take place; hence, sooner or later a device will have the opportunity to transmit. However, for M2M communication distributed DSA solution requires energy hungry operations such as sensing and requires to wait some time to understand the state of neighboring devices, hence which are not suitable for battery operated devices.

Cognitive networks are realized through different modules, in charge of performing tasks related to recognize, access, and efficiently use the spectrum. For the purpose of M2M, we will look at spectrum sensing, spectrum sharing, and spectrum management. Spectrum sensing is the ability for a device to sense the spectrum and identify potential spectrum holes, which are portion of the spectrum band at a given time and frequency which are not utilized by anyone. Spectrum sharing is performed across multiple devices, which share the resources to gain access to it. Finally, Spectrum management relates to all the operations needed to grant access to the spectrum and manage the resource between the clients. In other words, it is the overarching architecture composed of other submodules, in charge of executing different tasks.

Spectrum Sensing

As we already introduced, spectrum sensing is the ability for a device to sense the environment and acquire information about the spectrum occupancy, on a given frequency band. This is generally done by listening to the channel and analyzes the signals received. The spectrum detection, which is the task of determining whether a channel is occupied or not, can be done with different degrees of complexity, which clearly lead to different results.

In general, we can divide spectrum sensing in two broad categories, namely, energy detection spectrum sensing and cyclostationary spectrum sensing. Energy detection has the advantage of being simpler; hence, they require less resources to be run. Moreover, they do not need any information on the nature of the signal they are sensing. However, they are generally less accurate than other techniques, and they cannot distinguish between primary signals and secondary signals. More accurate techniques, such as the cyclostationary one, analyze the signal to determine whether a primary user is present or not. Generally speaking, they are more accurate, at the cost of a higher computation power required, and knowing the nature of the primary signal. At the same time, they may also distinguish between primary signal and secondary signal, opposite to the energy detection systems which only sense an occupied channel.

In [45] the authors proposed a compressive spectrum sensing algorithm, designed specifically for M2M communication. Specifically, the design is for UHF TV, thanks to TVWS technology. TVWS have been targeted as highly interesting for M2M [58], as they have favorable characteristics that bind well with M2M. The main idea behind the proposed solution is to sense only a part of the spectrum, instead of a wider band. Instead of sampling the whole band with fast update rates, the spectrum is sampled at sub-Nyquist rates, and only specific portions of the spectrum are further evaluated, thus reducing the overall complexity of the algorithm. Clearly, knowing which part of the spectrum needs further analysis has multiple benefits:
  • If information is shared among M2M devices, it could reduce the complexity also for other devices, at the cost of adding further overhead both on the management part and on the battery consumed to transmit the information.

  • If the device can make a decision without involving complex techniques, it saves battery. Clearly, this cannot come at the cost of potentially interfering with the primary user.

  • Through cooperative sensing, devices may share the computation costs, so that the energy required for each one is lower.

Energy detection, although fascinating for M2M communication, thanks to its simplicity, presents other drawbacks that need to be tackled [12]. For instance, it is difficult to set an appropriate threshold for the sensing, as setting a too low threshold may lead to missed communication opportunities due to environmental, and a high threshold may result in being too conservative. In fact, carefully setting the threshold is certainly the most challenging part in designing an efficient spectrum sensing algorithm, and this has been the main focus of several works which can be found in literature [19, 52]. While [52] provides general considerations and boundaries for energy detectors, [19] proposes a novel spectrum sensing algorithm, which involves cooperation between the nodes. In particular, they communicate to reach a consensus on the optimal energy threshold to be used. Even though this comes at the cost of involving additional communication, [19] shows that the gains are in favor of their solution.

Cooperative spectrum sensing uses shared knowledge among the nodes to reach a decision about the spectrum occupancy. A major challenge which arises with other techniques is related to the well-known hidden node problem, that is, nodes which might be sensed by some devices and not by others. In cooperative spectrum sensing, the hidden node problem is tackled by sharing the channel occupancy by nodes placed in different positions, hence reducing the probability that the hidden node is not sensed by anyone.

In cooperative spectrum sensing, we may have three different architectures:
  • Centralized Cooperative Sensing, in which nodes report spectrum measurements to a central entity, usually main powered, which analyzes the results and reports either hard or soft decisions to the clients.

  • Distributed Cooperative Sensing, in which there is not a central entity, but spectrum measurements are shared among nodes in a distributed way. Clearly, synchronization challenges have to be tackled.

  • Relay Cooperative Sensing, in which nodes send their measurements to other devices, which relay them to a central entity. The closer to the central entity, the higher the energy consumption, since a higher number of measurements should be reported.

Regardless of the specific architecture used, cooperative spectrum sensing also carries the problem of user selection, which is the subset of devices chosen for the next round of sensing. These devices should be representatives (i.e., cover the area of interests) and possibly be independent (i.e., in different positions, to limit the hidden node problem) [47].

Spectrum Sharing

Spectrum sharing refers to the techniques needed in order to share the spectrum resource between devices. Techniques vary, but the underlying idea is that devices could either compete for the spectrum, such as through auctions [20, 54], or through cooperative schemes [11, 23, 25, 59]. In particular, [25] is focused specifically on M2M communication, and by exploiting some of the M2M constraints, such as a small packet length, it shows how it is possible to transmit leveraging WiFi whitetimes, which is the inter-time between subsequent WiFi transmissions. In other words, in [25] vision M2M communication, thanks to its bursty nature and small packet lengths, can coexist with other transmissions by using period of non-transmissions in normally occupied channels.

Moreover, we can also distinguish between two different architectures of spectrum sharing:
  • Centralized Spectrum Sharing, in which the access to the spectrum is managed by a central entity, which is often named as Spectrum Coordinator. It can lease the spectrum in a competitive way, sending the same information to all the devices which eventually compete to access to it, or controlled, in which it decides which portion of spectrum should be leased to a given device.

  • Distributed Spectrum Sharing, in which devices take decision locally, based on the information they have, which could also be obtained through cooperation with other nodes. Again, cooperation improves the overall performances of the spectrum sharing, at the cost of an increased overhead and energy consumption.

Much work has been done on spectrum sharing architectures. Many works explore the possibilities offered by game theory, modeling games in which the objective of the client is accessing the spectrum [24, 31, 43].

Southwell et al. [48] focuses on quality of service games for spectrum sharing. Their solutions are totally distributed, and they also show the overarching benefits of a distributed solution over a centralized one.

A novel paradigm of spectrum sharing also exploits the height of the devices, as a countermeasure to reduce interference between devices accessing the spectrum, named 3D spectrum sharing, proposed in [5, 8]. The idea is that generally spectrum is considered free or occupied at a certain location, without considering the height of the buildings which may have been constructed there. For specific M2M use cases such as smart metering [4], this may significantly help in finding spectrum opportunities, as spectrum sharing is considered on a per-floor basis.

More formally, we can define the spectrum occupancy at a given location \(\mathscr {L}\) as
$$\displaystyle \begin{aligned} S_{\mathscr{L}} = \mathscr{O} \end{aligned} $$
(1)
where \(S_{\mathscr {L}}\) is the spectrum information regarding location \(\mathscr {L}\) and \(\mathscr {O}\) is the result obtained through a given technique of spectrum sharing. Since \(\mathscr {L}\) is a location, it is generally defined as
$$\displaystyle \begin{aligned} \mathscr{L} = < Lat, Lng > \end{aligned}$$
where Lat and Lng represent the latitude and the longitude, respectively. Most notably, this is the definition used by the TVWS spectrum database [9], which has been shown to offer less communication opportunities than those which may be used without interfering with the primary users. 3D spectrum sharing modifies the definition of \(\mathscr {L}\), changing it into
$$\displaystyle \begin{aligned} \mathscr{L}_{3D} = < Lat, Lng, \mathscr{H} > \end{aligned}$$
where \(\mathscr {H}\) is defined as the height from ground floor. In tall buildings, or in the case of specific networks such as TVWS, 3D spectrum sharing has shown good results in finding additional spectrum sharing opportunities, compared to other techniques. Of course, this comes at an additional cost and with additional constraints. At first, devices should know which are their \(\mathscr {H}\). This can be either manually defined when the device is installed, or it can be determined by the device during its operations. If manual configuration is used, the spectrum occupancy estimation may become inaccurate, due to the fact that devices may be moved or substituted. In case of the device determining its \(\mathscr {H}\), additional costs for building it should be accounted for, due to the need of a barometric sensor. Moreover, pressure value changes during the day, due to natural events such as thunderstorms or clouds, which should be again detected and properly managed [6].

An additional aspect crucial to M2M communication is related to the energy consumption. Switching from a 2D to a 3D space may involve additional computation costs, as more complex spectrum propagation and detection techniques should be used. However, this may be solved by using a spectrum manager which is in charge of handling all the spectrum information of a building. Again, this is not straightforward to implement, due to devices which may not be in the range of the spectrum manager and should then use relays to reach it in order to report measurements or to obtain information about the free channels.

Spectrum Management

Spectrum management comprises different tasks to be achieved, to efficiently utilize the available spectrum. This does not involve only spectrum sensing or spectrum sharing, as they may be relative to a specific frequency band. In fact, in spectrum management it is also foreseen that clients may have access to multiple bands, which may be exploited for specific needs. This may include multiband devices, able to utilize both WiFi channels and others, depending on the scenario and on the instant needs, like [22], where the authors design a system which leverages both cellular technologies such as LTE-A, to other opportunistic technologies like TVWS. The underlying idea is to utilize LTE-A to perform spectrum management tasks and also to query the remote spectrum database for TVWS availability.

Other solutions leverage TVWS to overcome other technology limitations in specific scenarios, such as video streaming [7]. Here, the scenario of interest is different from M2M, but the general architecture presents many similarities, like the use of a technology to query the database like WiFi and occasionally TVWS if WiFi performance is not enough to support the video streaming.

We can characterize multiband solutions, based on the spectrum availability and on the technologies characteristics overlapping, into the following categories:
  • Two technologies, both always available, similar characteristics: this is the case, for instance, of technologies such as WiFi and Bluetooth, which do not have to rely on other entities to assess the spectrum availability, as they use the ISM bands. They share similar characteristics; hence, their use together is limited.

  • Two technologies, both always available, different characteristics: this is the case, for instance, of WiFi and LTE, assuming coverage. In this scenario, typically technologies complement each other: for instance, modern smartphones have the possibility to switch to LTE should the WiFi connection presents low performance.

  • Two technologies, one always available: this is the case of TVWS, in which another technology such as LTE or WiFi is used to gain access to the TVWS network, by asking to the remote spectrum database the list of channels available at a given location.

Concerning M2M communication, most notably here is the work presented in [13], which compares LTE-Direct and WiFi-Direct over classical LTE communication, in the specific scenario of M2M. Cellular communication has been extensively looked at as one of the most prominent candidates for M2M, as they have a large coverage and present characteristics favorable for this kind of communication, like a longer range of communication over other technologies. However, they should be optimized to fully unleash the potential of M2M, like optimizing them for bursty transmissions or a better energy efficiency for battery-powered devices.

In [13] the authors show that both LTE-Direct and WiFi-Direct outperform classical LTE, mostly thanks to a specific design targeted at M2M communication. However, no clear winner arises, as LTE offers the best energy efficiency, particularly with a high number of devices, while WiFi has an advantage when considering small packets.

Other proposals do not focus on different access technologies but rather use M2M to enhance the main network basically by providing enhanced coverage, by using M2M devices which act as relay [33]. Even though this kind of communication has to be considered M2M, as no human intervention is foreseen in between, it is hard to think at those devices as battery powered and only communication to other devices. In fact, they can be devices which offer different services, one of which is M2M communication to extend the transmission range.

Spectrum management can thus be considered both a solution which enhances and optimizes the performance of M2M device. At the same time, it requires additional computation resources which eventually lead to increased battery consumption and reduced device life. Hence, there is a trade-off between using specific technologies, which may not be well suited for a specific scenario, even though they are simple to be managed, and to a more complex spectrum management which involves different bands and requires algorithms to successfully exploit additional technologies. Needless to say, also the cost of the device increases, and possibly the size of it, as different technologies may require different antennas and transceivers to operate.

Case Study: Cooperative M2M for Smart Metering

In this section we describe in detail a specific use case for M2M, related to smart metering applications. The scenario we model is depicted in Fig. 1, where multiple devices need to report their data to a central aggregator. Devices can have different capabilities and different residual energy; hence, algorithms which make them cooperate can significantly extend their battery life.
Fig. 1

Scenario for M2M smart metering communication. Multiple heterogeneous devices need all to report their data to a central aggregator, which is in charge of merging them. Devices can either report the measurements alone, hence consuming a considerable amount of battery, or they can form clusters, in which only one of them sends the measurements to the aggregator, and information about the other measurements is shared within the cluster. In this picture, we picture in red smart meters which report information of all the smart meters in the cluster, and in white those smart meters which only transmit intra-cluster. Clearly, the device which reports the data to the utility aggregator may change in time

System Model

We model a generic network scenario S, composed by \(\mathscr {N}\) secondary devices (SUs), operating in the same area. In other words, we assume they are able to sense the same signals. Each SU operates a DSA-enabled device, able to sense signals over all the frequencies of interest. We study the use case for the specific frequency band of TV white space (TVWS), although it can be easily extended to other frequency bands as well.

In TVWS, devices should query the remote spectrum database to obtain access to the list of channels available on a given area. This can be done by a device to obtain the list for its own purposes or to share the list to other devices which may not have access to the remote spectrum database. Hence, we also assume that devices also have an interface such as WiFi, through which they can access the remote spectrum database.

As the devices are close to each other, we can also assume that they are at most at one hop from the device which executes the query to the remote spectrum database. Hence, we call the querying device master device (MD) and all the other \(\mathscr {N} {-} 1\) as slave devices (SDs). Energy smart meters may be main powered, but for other kind of smart meters, such as those measuring gas, it is not possible to assume main power, but battery power is viable, and eventually energy efficiency has to be accounted for.

The set of activities \(\mathscr {S}\) MD i has to perform through the day can be summarized as:
  • Sense: based on the measure and on regulations, the device has to acquire the measure of interest and store it locally.

  • Query: this task has to be done to acquire access to the spectrum. Depending on the network, it can vary.

  • Share: in this task, the MD and SD share information about the available channels.

  • Rx: the MD, which should reports all the measurements of the SDs, receives measurement updates by them, which transmit their data to it.

  • Tx: the MD eventually sends its measurements and those received by the SDs to the utility aggregator.

  • Maintenance: based on its residual energy, the MD decides whether to drop its role to another device, which in this case should be informed.

Clearly, an SD performs a subset \(\int \in \mathscr {S}\) of activities, where
$$\displaystyle \begin{aligned} \int = < Sense, Share, Rx, Maintenance > \end{aligned}$$
We depict the typical day routine of smart meters in Fig. 2.
Fig. 2

The daily schedule of an MD. Below the lines there are the Sense operations, while on the upper line, we show the communication operation, starting from Q, which is relative to the Query operation. After getting the channel information, the MD can Share it through the S operation. Then the MD is ready to receive data from the SDs through the Rx operation and transmit it to the utility aggregator in the Tx phase. Finally, Maintenance operation can be performed through the M phase

To model the energy consumption of the MD, we define the following Equation:
$$\displaystyle \begin{aligned} E_{MD} = E_{\mathrm{sense}} + E_q + E_{\mathrm{share}}^{MD} + E_{rx}^{MD} + E_{tx} + E_m^{MD}, \end{aligned} $$
(2)
where Esense is the energy consumed by the sensing operation, common to both the MD and SDs, Eq is the energy needed to query the remote spectrum database, \(E_{\mathrm {share}}^{MD}\) is the energy consumed by sharing the measurements from the MD perspective, \(E_{rx}^{MD}\) accounts for the energy spent receiving measurements from SDs, Etx is relative to the energy needed to transmit the measurements to the utility aggregator, and \(E_m^{MD}\) accounts for the maintenance costs, in terms of energy, from the MD perspective.
Clearly, each SD has a similar behavior, more precisely as follows:
$$\displaystyle \begin{aligned} E_{SD} = E_{\mathrm{sense}} + E_{\mathrm{share}}^{SD} + E_{rx}^{SD} + E_m^{SD}, \end{aligned}$$
where \(E_{\mathrm {share}}^SD\) accounts for the energy spent to acquire the list of channels available, \(E_{rx}^{SD}\) is relative to SDs sending their measurements to the MD, and finally \(E_m^{SD}\) is relative to management operations.

The general model we just presented can be tailored to meet specific needs, depending on the scenario and on the target objective. For instance, if all the devices have to query the spectrum database, then Eq has to be counted also for SD. Another case would be the use of a fixed MD: this case can be modeled by putting \(E_m^{MD} = E_m^{SD} = 0\).

To compute each term, we refer to Table 1, where we report all the symbols used to model the scenario.
Table 1

Symbols table

Symbol

Description

Er

Energy per reading

\(\mathscr {F}\)

Number of readings per day

γ

Energy needed to transmit locally

𝜖

Sleep energy

α

Energy while in Rx

\(t_{\alpha }^{MD}\)

Time in which the MD listens for SD data

\(t_{\beta }^{SD}\)

Time in which the SD listens for MD data

\(\mathscr {N}\)

Number of nodes

\(E_i^{\mathrm {start}}\)

Starting energy

In detail, we obtain
$$\displaystyle \begin{gathered} E_{\mathrm{sense}} = E_r \cdot \mathscr{F} \end{gathered} $$
(3)
$$\displaystyle \begin{gathered} E_{\mathrm{share}}^{MD} = \gamma \cdot (\mathscr{N} - 1) \end{gathered} $$
(4)
$$\displaystyle \begin{gathered} E_{rx}^{MD} = \alpha \cdot t_{\alpha}^{MD} \end{gathered} $$
(5)
$$\displaystyle \begin{gathered} E_{m}^{MD} = \mathscr{H} \cdot \gamma \cdot (N - 1) \end{gathered} $$
(6)
where \(\mathscr {H}\) is a 0−1 variable indicating whether management has to be performed or not, depending on the policy used.
Concerning the SDs, the following equations are then defined:
$$\displaystyle \begin{gathered} E_{\mathrm{share}}^{SD} = \gamma + t_{\beta}^{SD} \cdot \alpha \end{gathered} $$
(7)
$$\displaystyle \begin{gathered} E_{rx}^{SD} = \gamma \end{gathered} $$
(8)
$$\displaystyle \begin{gathered} E_m^{SD} = t_{\beta}^{SD} \cdot \alpha \end{gathered} $$
(9)

Based on the above model, and having all the data about the devices, it is then possible to compute the devices lifetime. Moreover, it is also possible to design specific algorithms which manage the cluster, such as the one defined in [3], where the authors also define the cluster goal, which is the minimum lifetime the cluster should be able to achieve. Since devices are heterogeneous, achieving a higher goal requires that some devices never become an MD, in order to save battery, while others will have a higher probability of being the MD.

Here we present three different algorithms, derived from [3], which are summarized as follows:
  • Greedy: a device remains as MD as much as it can, before reaching a point where it must save battery to reach the final lifetime goal.

  • Highest: at the end of each day, the MD drops its role to the device having the highest residual energy. This can be realized with different techniques, and here we leverage on the dynamic backoff. Basically, when the MD advertises, it wants to change its role to SD; each SD computes a backoff to reply to the election proportional to its residual energy. The higher the residual energy, the shorter the backoff time. If a device hears a reply from another SD, it cancels its transmission, since a new MD has been found. This solution has also been used in other scenarios, such as vehicular communication [18].

  • Cost Aware: since devices can be heterogeneous, this algorithm updates periodically the actual costs of being an SD or an MD, and during the Maintenance slot, it selects the client which has done less for the cluster. In other words, the device which has been less days an MD compared to others will be elected.

In this domain, several metrics of interest exist. Within the scope of this analysis, we consider the following:
  • Lifetime: defined as the time from the start till the first device runs out of battery.

  • Elections: this is the number of times an MD changes its role. It gives an idea of the overhead involved in each protocol, hence the additional energy used to perform the Maintenance.

  • Fairness: the fairness of the system ∈ [0, 1]. A higher value means that clients shared the workload based on their capabilities, while a low value indicates that there is a considerable variation in the workload between them. More precisely, the fairness is computed as
    $$\displaystyle \begin{aligned} F = 1 - \left(\mathrm{max}_{i \in [0,n[}\left(\frac{D_i}{MD_i} \right) - \mathrm{min}_{i \in [0,n[}\left(\frac{D_i}{MD_i} \right) \right), \end{aligned} $$
    (10)
    where Di is the amount of time device i has been MD, and MDi is the maximum time device i can be an MD to keep its lifetime goal.
Figure 3 shows the overhead needed by the three protocols. In Fig. 3a we show the number of elections per day, when considering a cluster of 500 nodes. Setting a higher minimum goal to reach for the devices increases the performance of the system, and the number of elections per day decreases, as devices tend to keep the MD role for a longer time and save battery. This is also confirmed by Fig. 3b, where it is straightforward to note that the total number of elections remains stable after a certain point, which is clearly dependent on the initial configuration of the nodes.
Fig. 3

Analysis on the elections, which translates to the algorithm overhead. A higher number of elections translate into a higher consumption by the devices, which need to send several messages to reach a consensus on the MD. (a) Elections per day. (b) Total number of elections

Figure 4 shows the cluster lifetime, along with the fairness. A higher minimum goal for the cluster pushes the capabilities of the devices, and some may not be able to support such high demands. Hence, on the average the cluster lifetime may be reduced. Figure 4a shows the percentage of cluster which are able to reach the goal, which obviously decreases as the minimum goal is increased. These results should only be regarded as qualitative, as precise quantitative results can be obtained only by setting appropriate parameters on the framework, depending on the specific scenario and on the characteristics of the devices.
Fig. 4

Lifetime and fairness of the cluster. Setting a higher minimum goal may result in an unfeasible scenario, as the client cannot reach the goal, hence a lower average cluster lifetime. The fairness of the system is achieved only when using the Cost Aware protocol, as the other two do not share the load among the devices but only select those which better suit the MD role at the moment. (a) Lifetime of the cluster. (b) Fairness of the cluster

The fairness of the system is an important parameter, as it shows the workload share. This is of paramount importance when performing maintenance operations, as having only one device which runs out of battery, while all the others still have energy remaining, raises the maintenance costs for such device. On the other hand, achieving a discharge which goes at the same pace for each one, depending on the different capabilities, can end up in having all the cluster to have similar energy remaining, hence perform maintenance operations on all the devices at once, thus decreasing the maintenance cost per single device.

Conclusion and Future Directions

In this chapter we presented the benefits and the challenges of DSA for M2M communication. M2M communication should be performed by devices without or with little human intervention. Thus, they should encompass self-healing and self-configuring capabilities. Cognitive network, hence DSA, is a technology that could help to overcome some of the challenges that M2M networks face, such as the energy constraints and the possible long range of the communication. Several experiments, both theoretic as well as simulative and with test-beds, have demonstrated both the benefits and the limits of M2M communication when considering DSA. In particular, energy efficiency should be taken into account when deploying M2M networks, and it becomes even more constrained when using spectrum agile techniques such as spectrum sensing and DSA.

Clearly, there is a plethora of topics which can be explored and further analyzed, which we have identified and commented in this chapter, and which we summarize here:
  • Energy efficiency: even though we stated that M2M device may not be battery powered, it is also straightforward to understand that in a vast set of scenarios, they will be. Energy efficiency can be optimized in several parts of the device, either on the computation, on the transmission, or on the architecture level, where improvements are obtained through device collaboration.

  • Spectrum sensing: it can be expensive both in terms of computation cost and also simply on the energy needed to listen to the channels, even more depending if a large spectrum is considered. Different techniques can be applied, such as cooperative spectrum sensing, in which devices organize and divide the tasks so that the energy consumption per single device is reduced. Other solutions involve identifying smaller portions of spectrum so that the task of sensing becomes more feasible. Regardless of the technique used, spectrum sensing is vital for DSA, and it is also an operation which needs to be repeated multiple times, as spectrum utilization may change suddenly.

  • Test-beds: most of the presented results have been studied through simulations or with devices which implement only some of the tasks a cognitive M2M device should utilize. Hence, it is challenging to find wider analysis, which take into account M2M communication and DSA accounting for realistic scenarios. In particular, primary user activity on the spectrum may be challenging to understand in real scenarios, rather than simulations where it can be controlled.

References

  1. 1.
    3GPP. Study on provision of low-cost machine-type communications (MTC) user equipments (UEs) based on LTE. TSG GERAN R1Google Scholar
  2. 2.
    Altintas O et al (2011) Demonstration of vehicle to vehicle communications over TV white space. In: 2011 IEEE vehicular technology conference (VTC Fall), San Francisco, pp 1–3Google Scholar
  3. 3.
    Bedogni L, Trotta A, Di Felice M, Bononi L (2013) Machine-to-Machine communication over TV white spaces for smart metering applications. In: 2013 22nd international conference on computer communications and networks (ICCCN), Nassau, pp 1–7Google Scholar
  4. 4.
    Bedogni L, Achtzehn A, Petrova M, Mähönen P (2014) Smart meters with TV gray spaces connectivity: a feasibility study for two reference network topologies. In: 2014 eleventh annual IEEE international conference on sensing, communication, and networking (SECON), Singapore, pp 537–545Google Scholar
  5. 5.
    Bedogni L, Trotta A, Di Felice M (2015) On 3-dimensional spectrum sharing for TV white and gray space networks. In: 2015 IEEE 16th international symposium on a world of wireless, mobile and multimedia networks (WoWMoM), Boston, pp 1–8Google Scholar
  6. 6.
    Bedogni L, Franzoso F, Bononi L (2016) A self-adapting algorithm based on atmospheric pressure to localize indoor devices. In: 2016 IEEE global communications conference (GLOBECOM), Washington, DC, pp 1–6Google Scholar
  7. 7.
    Bedogni L, Trotta A, Di Felice M, Gao Y, Zhang X, Zhang Q, Malabocchia F, Bononi L (2017) Dynamic adaptive video streaming on heterogeneous TVWS and Wi-Fi networks. IEEE/ACM Trans Netw (IEEE TON) 25:3253–3266CrossRefGoogle Scholar
  8. 8.
    Bedogni L, Achtzehn A, Petrova M, Mähönen P, Bononi L (2017) Performance Assessment and Feasibility Analysis of IEEE 802.15.4m Wireless Sensor Networks in TV Grayspaces. ACM Trans Sensor Netw 13(1), Article 8 (January 2017), 27 pagesCrossRefGoogle Scholar
  9. 9.
    Bedogni L, Malabocchia F, Di Felice M, Bononi L (2017) Indoor use of gray and white spaces: another look at wireless indoor communication. IEEE Veh Technol Mag 12(1):63–71CrossRefGoogle Scholar
  10. 10.
    Brew M, Darbari F, Crockett LH, Waddell MB, Fitch M, Weiss S, Stewart RW (2011) UHF white space network for rural smart grid communications. In: 2011 IEEE international conference on smart grid communications (SmartGridComm), Brussels, pp 138–142Google Scholar
  11. 11.
    Chang HB, Chen KC (2011) Cooperative spectrum sharing economy for heterogeneous wireless networks. In: 2011 IEEE GLOBECOM workshops (GC Wkshps), Houston, pp 458–463Google Scholar
  12. 12.
    Chatziantoniou E, Allen B, Velisavljevic V (2015) Threshold optimization for energy detection-based spectrum sensing over Hyper-Rayleigh fading channels. IEEE Commun Lett 19(6):1077–1080CrossRefGoogle Scholar
  13. 13.
    Condoluci M, Militano L, Orsino A, Alonso-Zarate J, Araniti G (2015) LTE-direct vs. WiFi-direct for machine-type communications over LTE-A systems. In: 2015 IEEE 26th annual international symposium on personal, indoor, and mobile radio communications (PIMRC), Hong Kong, pp 2298–2302Google Scholar
  14. 14.
    Cui P, Liu H, Rajan D, Camp J (2014) A measurement study of white spaces across diverse population densities. In: 2014 12th international symposium on modeling and optimization in mobile, ad hoc, and wireless networks (WiOpt), number WiNMeE, May 2014. IEEE, pp 30–36Google Scholar
  15. 15.
    Datta SK, Bonnet C (2014) Smart M2M gateway based architecture for M2M device and endpoint management. In: 2014 IEEE international conference on internet of things (iThings), and IEEE green computing and communications (GreenCom) and IEEE cyber, physical and social computing (CPSCom), Taipei, pp 61–68Google Scholar
  16. 16.
    Di Felice M, Chowdhury KR, Bononi L (2010) Analyzing the potential of cooperative cognitive radio technology on inter-vehicle communication. In: Wireless days (WD), 2010 IFIP, Venice, pp 1–6Google Scholar
  17. 17.
    Di Felice M, Bedogni L, Bononi L (2012) DySCO: a dynamic spectrum and contention controlframework for enhanced broadcast communication invehicular networks. In: Proceedings of the 10th ACM international symposium on mobility management and wireless access (MobiWac’12)Google Scholar
  18. 18.
    Di Felice M, Bedogni L, Bononi L (2013) Group communication on highways: an evaluation study of Geocast protocols and applications. Elsevier’s Ad Hoc Netw J 11: 818–832Google Scholar
  19. 19.
    Ebrahimzadeh A, Najimi M, Andargoli SMH, Fallahi A (2015) Sensor selection and optimal energy detection threshold for efficient cooperative spectrum sensing. IEEE Trans Veh Technol 64(4):1565–1577CrossRefGoogle Scholar
  20. 20.
    Fang YX, Xue G (2012) HERA: an optimal relay assignment scheme for cooperative networks, selected areas in communications. IEEE J 30(2):245–253Google Scholar
  21. 21.
    Federal Communications Commission. FCC online table of frequency allocations. Available online: https://transition.fcc.gov/oet/spectrum/table/fcctable.pdf
  22. 22.
    Gao Y, Qin Z, Feng Z, Zhang Q, Holland O, Dohler M (2016) Scalable and reliable IoT enabled by dynamic spectrum management for M2M in LTE-A. IEEE Internet Things J 3(6): 1135–1145CrossRefGoogle Scholar
  23. 23.
    Gao L, Duan L, Huang J (2017) Two-sided matching based cooperative spectrum sharing. IEEE Trans Mob Comput 16(2):538–551CrossRefGoogle Scholar
  24. 24.
    Gu Y, Saad W, Bennis M, Debbah M, Han Z (2015) Matching theory for future wireless networks: fundamentals and applications. IEEE Commun Mag 53(5):52–59CrossRefGoogle Scholar
  25. 25.
    Harris J, Beach M, Nix A, Thomas P (2016) Spectrum sharing for M2M applications through Whitetime exploitation in WiFi networks. In: 2016 IEEE wireless communications and networking conference, Doha, pp 1–6Google Scholar
  26. 26.
    Hsu LK, Chou CT (2013) A dynamic spectrum access (DSA)-based multichannel protocol for large scale machine-to-machine (M2M) networks. In: 2013 9th international wireless communications and mobile computing conference (IWCMC), Sardinia, pp 1217–1222.  https://doi.org/10.1109/IWCMC.2013.6583730
  27. 27.
    ICT-DOST. TV white space deployment in Philippines. http://icto.dost.gov.ph/tv-white-space-deployment-in-ph-the-largest-in-asia/
  28. 28.
    IEEE 802.15.4m standard. Available online: https://standards.ieee.org/findstds/standard/802.15.4m-2014.html
  29. 29.
    Ihara Y et al (2013) Distributed autonomous multi-hop vehicle-to-vehicle communications over TV white space. In: 2013 IEEE consumer communications and networking conference (CCNC), Las Vegas, pp 336–344Google Scholar
  30. 30.
    Islam MH, Koh CL, Oh SW, Qing X, Lai YY, Wang C, Liang YC, Toh BE, Chin F, Tan L, Toh W (2008) Spectrum survey in Singapore: occupancy measurements and analyses. In: Proceedings of the 3rd international conference on cognitive radio oriented wireless networks and communications (CrownCom), May 2008. IEEEGoogle Scholar
  31. 31.
    Ji Z, Liu KJR (2007) Cognitive radios for dynamic spectrum access – dynamic spectrum sharing: a game theoretical overview. IEEE Commun Mag 45(5):88–94CrossRefGoogle Scholar
  32. 32.
    Kerttula J, Jäntti R (2011) DVB-T receiver performance measurements under secondary system interference. In: Proceedings of COCORA 2011Google Scholar
  33. 33.
    Laya A, Wang K, Widaa AA, Alonso-Zarate J, Markendahl J, Alonso L (2014) Device-to-device communications and small cells: enabling spectrum reuse for dense networks. IEEE Wirel Commun 21(4):98–105CrossRefGoogle Scholar
  34. 34.
    Lee J, Bagheri B, Kao H-A (2015) A cyber-physical systems architecture for industry 4.0-based manufacturing systems. Manuf Lett 3:18–23. ISSN:2213–8463CrossRefGoogle Scholar
  35. 35.
    Li TL, Chou CT, Hsu LK (2015) Proportional sharing in distributed dynamic spectrum access-based networks. IEEE Trans Mob Comput 14(1):155–168Google Scholar
  36. 36.
    Liang YC, Chen KC, Li GY, Mahonen P (2011) Cognitive radio networking and communications: an overview. IEEE Trans Veh Technol 60(7):3386–3407.  https://doi.org/10.1109/TVT.2011.2158673 CrossRefGoogle Scholar
  37. 37.
    Lim JH, Kim W, Naito K, Yun JH, Cabric D, Gerla M (2014) Interplay between TVWS and DSRC: optimal strategy for safety message dissemination in VANET. IEEE J Sel Areas Commun 32(11):2117–2133CrossRefGoogle Scholar
  38. 38.
  39. 39.
    Marcu I, Marghescu I (2010) Evaluation of spectrum occupancy in an urban environment in a cognitive radio context. Adv Telecommun 3(3):172–181Google Scholar
  40. 40.
    Masonta MMT, Johnson DL, Mzyece M (2011) The white space opportunity in Southern Africa: measurements with Meraka cognitive radio platform. In: Proceedings of AFRICOMM. Lecture notes of the institute for computer sciences, social informatics and telecommunications engineering, vol 92. Springer, pp 64–73Google Scholar
  41. 41.
    McHenry MA, Tenhula PA, McCloskey D, Roberson DA, Hood CS (2006) Chicago spectrum occupancy measurements & analysis and a long-term studies proposal. In: Proceedings of the first international workshop on technology and policy for accessing spectrum (TAPAS)Google Scholar
  42. 42.
    Montori F, Contigiani R, Bedogni L (2017) Is WiFi suitable for energy efficient IoT deployments? In: A performance studyon proceedings of the 3rd IEEE internal forum on research and technologies for society and industry, technologies for smarter societies (IEEE RTSI), 11–13 Sept, ModenaGoogle Scholar
  43. 43.
    Niyato D, Hossain E (2008) Competitive spectrum sharing in cognitive radio networks: a dynamic game approach. IEEE Trans Wirel Commun 7(7):2651–2660CrossRefGoogle Scholar
  44. 44.
    Pérez-Romero J, Noguet D, López-Benítez M, Casadevall F (2011) Towards more-efficient spectrum usage: spectrum-sensing and cognitive-radio techniques. in URSI Radio Science Bulletin, 2011(336):59–74, March 2011Google Scholar
  45. 45.
    Qin Z, Gao Y, Plumbley MD, Parini CG, Cuthbert LG (2014) Efficient compressive spectrum sensing algorithm for M2M devices. In: 2014 IEEE global conference on signal and information processing (GlobalSIP), Atlanta, pp 1170–1174Google Scholar
  46. 46.
    Reed JH (2010) Enabling rural Virginia’s smart grid through white space communications, Virginia Tech, Technical reportGoogle Scholar
  47. 47.
    Selen Y, Tullberg H, Kronander J (2008) Sensor Selection for Cooperative Spectrum Sensing. 2008 3rd IEEE Symposium on New Frontiers in Dynamic Spectrum Access Networks, Chicago, pp. 1–11Google Scholar
  48. 48.
    Southwell R, Chen X, Huang J (2014) Quality of service games for spectrum sharing. IEEE J Sel Areas Commun 32(3):589–600CrossRefGoogle Scholar
  49. 49.
    Sum CS, Harada H, Kojima F, Lan Z, Funada R (2011) Smart utility networks in tv white space. IEEE Commun Mag 49(7):132–139CrossRefGoogle Scholar
  50. 50.
    Taher TM, Bacchus RB, Zdunek KJ, Roberson DA (2011) Long-term spectral occupancy findings in Chicago. In: Proceedings of IEEE international symposium on dynamic spectrum access networks (DySPAN), pp 100–107, May 2011Google Scholar
  51. 51.
    Tracking and monitoring of animals with combined wireless technology and geofencing, US Patent Application US20130340305Google Scholar
  52. 52.
    Umar R, Sheikh AUH, Deriche M (2014) Unveiling the hidden assumptions of energy detector based spectrum sensing for cognitive radios. IEEE Commun Surv Tutor 16(2):713–728. Second Quarter 2014CrossRefGoogle Scholar
  53. 53.
    Varghese A, Tandur D (2014) Wireless requirements and challenges in industry 4.0. In: 2014 international conference on contemporary computing and informatics (IC3I), Mysore, pp 634–638Google Scholar
  54. 54.
    Wang J, Liu A, Yan T et al. (2018) Peer-to-Peer Netw Appl. 11:(679)Google Scholar
  55. 55.
    Wellens M, Wu J, Mähönen P (2007) Evaluation of spectrum occupancy in indoor and outdoor scenario in the context of cognitive radio. In: 2007 2nd international conference on cognitive radio oriented wireless networks and communications (CROWNCOM), Aug 2007. IEEE, pp 420–427Google Scholar
  56. 56.
    Yin S, Chen D, Zhang Q, Liu M, Li S (2012) Mining spectrum usage data: a large-scale spectrum measurement study. IEEE Trans Mob Comput 11(6):1033–1046Google Scholar
  57. 57.
    Ying X, Zhang J, Yan L, Zhang G, Chen M, Chandra R (2013) Exploring indoor white spaces in metropolises. In: Proceedings of the 19th annual international conference on Mobile computing & networking – MobiCom’13, New York. ACM Press, p 255Google Scholar
  58. 58.
    Yuan Ma, Gao Y, Parini CG (2015) Sub-Nyquist rate wideband spectrum sensing over TV white space for M2M communications. In: 2015 IEEE 16th international symposium on a world of wireless, mobile and multimedia networks (WoWMoM), Boston, pp 1–6Google Scholar
  59. 59.
    Zhang Y, Yu R, Nekovee M, Liu Y, Xie S, Gjessing S (2012) Cognitive machine-to-machine communications: visions and potentials for the smart grid. IEEE Netw 26(3):6–13CrossRefGoogle Scholar
  60. 60.
    Zhou H et al (2015) Enabling efficient and wide-coverage vehicular content distribution over TV white spaces. In: 2015 international conference on wireless communications & signal processing (WCSP), Nanjing, pp 1–6Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2018

Authors and Affiliations

  1. 1.Department of Computer Science and EngineeringUniversity of BolognaBolognaItaly

Section editors and affiliations

  • Yue Gao
    • 1
  1. 1.School of Electronic Engineering and Computer ScienceQueen Mary University of LondonLondonUK

Personalised recommendations