QoS Strategies for Wireless Multimedia Sensor Networks in the Context of IoT at the MAC Layer, Application Layer, and Cross-Layer Algorithms

,


Introduction
Wireless sensor networks (WSNs) have a significantly large number of interconnected sensor nodes which can sense their environmental attributes like pressure, light, temperature, sound, humidity, and location. ey further cooperate among themselves over wireless transmission media across which they transmit their data as they monitor their environment [1]. For this to be done more accurately, there is a need for multimedia system support for better information gathering and environmental monitoring. WSNs have thus given birth to a new paradigm shift toward WMSNs. is is inspired by the recent advances in technology that have given rise to portable, cheap multimedia capture, transmission, and storage devices such as digital video cameras, microphones, low-cost smart phones, imaging sensors, memory cards, and hard disks. ese technologies are easy to integrate into a node and have made information gathering and monitoring of their environment easier and cheaper. Being low-cost smart devices, they have motivated many scholars to undertake research on WMSNs. ese can promptly transmit, store, compare, and combine data from heterogeneous sources.
WMSNs are an enhanced kind of WSNs that can sense and/or transmit both scalar and multimedia data including image, audio, and video streams in real-time or non-realtime transmission. ey are networks for wireless embedded devices that can permit the users to retrieve multimedia information from their environment [2]. ey have many applications in surveillance and environmental monitoring systems, traffic monitoring, target tracking, intrusion detection systems, telemedicine for advanced health care, etc. Iqbal et al., for instance, developed an efficient power allocation and personal wireless hub (PWH) placement strategy for maximizing the data rate under the cognitive radio interference constraint that has an efficacy with low complexity to facilitate paramedic staff in next-generation health care facilities using multimedia in smart hospitals [3].

Terrestrial WSNs.
ese can communicate with base stations efficiently and comprise 100 s to 1000 s of wireless sensor nodes deployed in the unstructured (ad hoc) or structured (preplanned) style. In the ad hoc mode, nodes are randomly distributed in the target region that is dropped from a fixed plane. e structured mode considers optimal placement, grid placement, and 2D and 3D placement models. Energy is conserved by use of low duty cycling, delay minimization, and optimal routing.

Underground WSNs.
ese are more expensive than terrestrial WSNs in terms of deployment, maintenance, equipment cost, and careful planning. e WSNs comprise many sensor nodes hidden in the ground to monitor underground conditions. To relay information from the sensor nodes to the base station, additional sink nodes are located above the ground. ese are highly affected by attenuation and signal loss and are very difficult to charge.

Underwater WSNs.
ese have multiple sensor nodes and vehicles deployed under water. Autonomous underwater vehicles are used to gather data from the sensor. Long propagation delay and bandwidth and sensor failures are a major challenge here.

Multimedia WSNs (WMSNs).
ese have been proposed to enable tracking and monitoring of events in the form of multimedia such as imaging, video, and audio. ese networks comprise low-cost sensor nodes equipped with microphones and cameras. e nodes are interconnected with each other over a wireless connection for data compression, retrieval, and correlation. e challenges with WMSNs include high energy consumption, data processing and compressing techniques, and high bandwidth requirements for proper and easy content delivery.

Mobile WSNs.
ey comprise a collection of sensor nodes which can move on their own and can interact with the physical environment. Mobile nodes have the ability to compute, sense, and communicate. e mobile WSNs are more versatile than static sensor networks. e advantages of WMSNs over the static WSNs include better and improved coverage, better energy efficiency, and superior channel capacity.

e Architecture of a WMSN.
e WMSN architecture has the following 3 subdivisions: (i) Single-tier flat architecture: this consists of homogeneous multimedia nodes which are able to execute any function to the sink via multihop routes. (ii) Single-tier clustered architecture: this consists of heterogeneous nodes passing sensed information to the cluster head for processing. (iii) Multitier architecture: this too consists of heterogeneous nodes and does object sensing and target capturing and tracking. Figure 1 shows a typical example of a WMSN architecture. Figure 1(a) shows a single-tier flat and homogeneous architecture, in which sensors having the same physical abilities are utilized. In Figure 1(b), we have a single-tier clustered and heterogeneous architecture, having nodes with different physical capabilities, e.g., multimedia nodes and scalar nodes. e multitier clustered and heterogeneous architecture (Figure 1(c)) has several layers of nodes having diverse types and processing tasks per layer [2]. WMSNs are widely deployed to offer infrastructural support and sensor accessibility rendering them suitable for Internet of multimedia things (IoMT) transmission [4]. Most applications (apps) in the IoMT, e.g., wearable devices, utilize the WMSN technology.

e Internet of ings (IoT).
In the IoT today, we visualize a situation whereby smart devices connect to a single network-the Internet. It is becoming more popular than it has been because of the multitude of connected devices available recently. A number of IoT-and IoMT-based applications are coming lately while attracting enormous attention. ese applications include smart cities, smart vehicles, homes, factories (Figure 2), and GPS tracking devices [5,6] and have attracted many inventions in the Americas, Asia, Europe, etc. A number of commercial and military applications come up because of introduction of multimedia objects in transmission of data such as remote patient monitoring in telehealth and telemedicine and traffic management systems enhanced by smart video cameras, among others [4]. is calls for an upgrade in functionality of IoT systems to the IoMT. Figure 2 shows an illustration of the IoT architecture which could be aided by RFID, optical tags, QR codes, Bluetooth low energy, Wi-Fi direct, and LTE-Advanced, among others. Most of the scholars emphasize improvement of efficiency for handling a lot of realtime information (info) but ignore multimedia transmission aspects [7,8]. e direction of research is shifting from the ordinary IoT to the multimedia-based IoT because of the need to enable smart devices to efficiently observe, sense, and understand their environment through multimedia data [9,10], hence resulting in the emergence of the newer field of Internet of multimedia things (IoMT).
IoMT is "the IoT-based paradigm that enables objects to connect and exchange structured and unstructured data with one another to enable multimedia-based services and applications" [3].
ere is a need for vast processing power, memory, and bandwidth for high QoS when transmitting multimedia data in the IoMT in comparison with scalar data in a traditional IoT. IoT system functionality should therefore be upgraded to the IoMT. We compare the two as discussed in [7]: (i) e IoT has standardized communication protocols, whereas the IoMT's protocols are nonstandardized. (ii) In terms of QoS, the IoT requires low bandwidth, whereas the IoMT requires higher bandwidth. (iii) e IoMT transmits heterogeneous multimedia data, whereas IoT data transmitted have limited heterogeneity. (iv) IoT sensor nodes consume less energy than IoMT sensor nodes.
(v) IoT devices are deployed in application-dependent RFID tags, but the IoMT is in video and audio sensors. (vi) In terms of service composition, the IoMT has no available specialized middleware, whereas the IoT has specialized service-oriented, architecture-based, and event-based middleware.
For improved QoS, best e ort services, and higher energy e ciency in IoMT networks and applications, enormous multimedia-supported routing is gaining ground in the research arena in the WMSN area in routing protocols, algorithms, and techniques based on network architectures and application requirements [11]. Nevertheless, according to Ahmad et al. [12], the enormous, resource-constrained, heterogeneous environment of the IoT challenges its expansion and deployment. is is because most existing IoT apps comprise overlaid deployments of wireless sensor and actuator networks in which apps cannot interact with each other or share and reuse the few available resources. In addition to that, e cient sensing and propagation of info  and swift response to changes in the physical world are challenging IoT requirements [12].

What Has Been Done?
In wireless communication systems, routing is one issue that is quite a challenge, yet there are very few surveys available especially regarding WMSN protocols [1] and more so the MAC layer, application layer, and cross-layer protocols although some surveys have been published on WSNs. To the best of our knowledge, no survey combines all these aspects. is makes our survey important to add to the existing literature. e authors in [1] discuss a number of WMSN routing techniques and the properties and shortcomings of these. However, they do not discuss any application layer techniques. Muzakkari et al. [13] survey some recent WSN contention-based, scheduling-based, and hybrid MAC protocols whereby they focus on the underlying principles, advantages, limitations, and their applications. But much as they only focus on MAC protocols, they still leave out a number of them in their survey. Yigitel et al. [14] carry out a survey on QoS-aware MAC protocols for WSNs. ey review the QoS challenges and views for WSNs, study the QoS mechanisms, categorize the state-of-the-art QoS-aware MAC protocols, and also talk about the advantages and drawbacks of the same. But being a 2011 release, a number of key protocols developed since then to date are not presented therein. Abbas and Kure [15] review various methods for QoS provisioning at the levels of routing, MAC layer, and cross-layer including the schemes for admission control and scheduling for QoS provisioning as well as the problems and challenges involved. But they leave out the protocols from 2010 to date. Shatnawi [16] and Karagiannis et al. [17] review the application layer protocols for the Internet of things. In their survey, they address a number of application layer protocols which are employed for IoT, for affirming a reliable tie among objects and things. ey evaluate the reviewed protocols in terms of architecture, communication model, security, and QoS as well as the weaknesses and strengths per reviewed protocol. However, they only concentrate on application layer protocols and still leave out some critical protocols. In a recent survey, Bernard et al. [18] discuss a number of cross-layer QoS strategies for IoT. But because of space constraint, we leave out some critical protocols, and besides, this paper does not talk about application layer and MAC layer protocols. AlAmri and Abdullah [19] carry out a survey on cross-layer QoS protocols for WMSNs in which they state that the cross-layer architecture is a novel idea that brings together a number of layers for enabling integration and exchange of information in between them with higher efficiency compared to the traditional layered model. ey discuss and compare the available cross-layer WMSN protocols that cross the uses of adjacent or nonadjacent layers. But they concentrate on cross-layer protocols and leave out the MAC layer and application layer protocols, and also some cross-layer protocols are left out. e surveys that have been reviewed are compared in Table 1. e major contributions that make our survey important are outlined below.

Importance of the Survey
(i) We discuss and draw a comparison between the IoT and the IoMT by discussing the differences between the two. (ii) We further discuss the newer paradigm of WMSNs and compare WSNs and WMSNs in our discussion. (iii) We make a comprehensive survey of the recent state-of-the-art routing protocols focusing on the MAC layer, the application layer, and the crosslayer paradigm. (iv) We discuss the WMSN protocols as well as the WMSN-enabled protocols that have been developed in the recent studies by several scholars. (v) In some of the reviewed protocols, we talk about the relative advantages and disadvantages of some of the routing approaches for readers to be in position to comprehend different techniques and so choose the most appropriate technique depending on the user requirements. (vi) In all these, we cover the challenges and opportunities existing for the discussed categories. (vii) We consider the possible future research trends before concluding this paper.

Challenges
(i) Since WSNs and WMSNs are usually distributed and ad hoc networks, their nodes are powered by batteries making energy optimization a challenge that needs to be addressed when designing their routing protocols. (ii) Considering WSNs, their routing protocols do not put into consideration multimedia applications which need a lot of bandwidth and processing energy and must be transmitted in real time with utmost fidelity. Extra challenges in routing are imposed in WMSNs in deployment of heterogeneous sensors since there could be audio, video, and still pictures in addition to scalar data during transmission. Besides, the heterogeneous data thereby handled all have different QoS requirements, and with many different business needs, different services with varying requirements give rise to a big challenge for the routing design. (iii) WMSNs have also got nonrechargeable batteries placing a limitation on the energy of the nodes, yet multimedia applications consume a lot of energy. Quite a number of complications and limiting factors exist practically despite the fact that sensors can be self-powering as enabled by energy harvesting. Predictive video encoding using MPEG-4 and other standards is also an energy-consuming process and so causes enormous degradation in routing efficiency complicating the design of WMSN protocols even more. (iv) e heterogeneous nature of the network is another challenge since there is a difference in the categories of nodes needed for communication so as to enable the facilitation of effective data collection and processing as well as efficient transmission. It is not easy to have a uniform communication protocol platform because of different functionalities unlike the case of the conventional WSNs compared to the WMSNs. (v) ere must be a trade-off between the energy efficiency and the multimedia QoS when deciding on the route taken. Data aggregation and compression routing protocols may be applicable for energysaving though they can lead to intolerable delays in WMSNs. Furthermore, there might be network congestion due to the many-to-many and mutual interference in wireless routes and the scarce WMSN resources. In case a given node gets so many high-rate streams, it lowers the performance of the network and raises the possibility of node failure due to energy drainage.
is paper mainly discusses QoS strategies for WSNs and WMSNs in the context of IoT from the MAC layer, application layer, and cross-layer perspective. is paper presents a state-of-the-art survey on routing in WSNs and WMSNs. e rest of this paper is organized as follows: In Section 2, we discuss the QoS strategies at the MAC layer. In Section 3, we discuss the QoS strategies at the application layer. Section 4 discusses the cross-layer QoS strategies, and in Section 5, we give some future research directions before concluding this paper in Section 6.

QoS Strategies at the MAC Layer
e medium access control (MAC) protocol is meant to regulate admittance to the shared medium as well as transmission reliability and efficiency through error correction in wireless transmissions among others. It is also responsible for performance of framing, addressing, and flow control. e MAC protocol design affects energy efficiency too. is is majorly categorized into contention-free protocols, contention-based protocols, and a hybrid of these.

MAC Protocols for
WSNs. Hybrid MAC protocols: these combine contention-free and contention-based MAC protocols on whose advantages they capitalize. Some of these are discussed below.

S-MAC (Sensor MAC).
is is a MAC protocol designed for WSNs. e main components of S-MAC include (i) periodic listen and sleep, (ii) collision and overhearing avoidance, and (iii) message passing. Energy Journal of Computer Networks and Communications consumption and self-configuration are controlled using these three techniques. Nodes periodically go into the sleep mode to control the energy consumed in listening to idle channels, and virtual clusters are formed by neighbor nodes for sleep schedule autosynchronization. e protocol employs in-channel signaling when the radio is in the sleep mode and message passing is applied to decrease contention latency for store-and-forward WSN applications during data transmission [22]. For adjacent nodes to solve their clock drift, S-MAC requires periodic synchronization. To reduce synchronization errors, exchanged timestamps are relative not absolute. Advantages: (i) S-MAC has a longer listening period than the clock error. (ii) It has more capacity to conserve energy than IEEE 802.11. (iii) It can do trade-off between energy and latency depending on traffic conditions [22].

B-MAC (Berkeley MAC).
is is a carrier-sense MAC protocol designed for WSNs featuring a simplistic, scalable implementation that does network variations and is effective in channel valuation [23]. e protocol has an interface with some flexibility leading to very low-power usage and controls collisions through collision avoidance mechanisms leading to effective use of the frequency band. It uses a sampling scheme that is adjustable to minimize duty cycling and idle listening so as to reduce power wastage. Nodes are scheduled with their duty cycle each within which they send the preamble to the channel in case of availability of data for transmission. In different duty-cycle schedules, channel usage is checked by the destination nodes; that is, if the preamble found is long, it will remain powered on until the destination address is obtained, and in case it is the target, it will either await sender's data or go to sleep. e protocol has lower latency than synchronous MAC protocols but has a problem of overhearing where the neighbor node gets lengthy preambles though not the intended receiver, hence wasting a lot of power in doing so [24]. It outperforms a number of protocols via reconfiguration, feedback, and bidirectional interfaces for upper layer services. It can run at low duty cycles, and its apps cannot experience synchronization overhead [23]. Advantages: (i) e protocol features a simplistic, scalable implementation that does network variations and is effective in channel valuation. (ii) It has an interface with flexibility leading to very low-power usage. (iii) It controls collisions through collision avoidance mechanisms leading to effective use of the frequency band. (iv) It uses an adjustable sampling scheme to minimize duty cycling and idle listening, reducing power wastage. Disadvantages: (i) It has a problem of overhearing where the neighbor node gets lengthy preambles though not the intended receiver, hence wasting a lot of power in doing so.

T-MAC.
is is a contention-based MAC protocol for WSNs that dynamically with little difficulty via fine-grained timeouts adapts a listen/sleep duty cycle. It uses this duty cycle by ending its active part so as to vary load in time and position, thereby decreasing the energy wastage on idle listening [25]. When starting an active period, it uses a short listening window, thereby improving the use of energy by S-MAC.
rough adaptive duty cycling, T-MAC saves power but at a cost of less throughput and enhanced delay. Both S-MAC and T-MAC under similar conditions perform in the same way though in variable workloads S-MAC uses 5 times more power than T-MAC. But T-MAC is complex and not scalable. If its active window is decreased, its snooping capability on nearby traffic and adaptation to variable conditions of the network is reduced [23]. All nodes in this protocol are configured to wake up periodically, listen to adjacent nodes, and go back into the sleep mode till the next frame as more messages come into the queue. Collision avoidance and reliable communication are ensured through Request-to-Send (RTS), Clear-to-Send (CTS), and Acknowledgment (ACK) messages among nodes [25].

Author(s)
Year Characteristics Bhandary et al. [1] 2016 Routing strategies in WMSNs, properties, and limitations Muzakkari et al. [13] 2018 WSN contention-based, scheduling-based, and hybrid MAC protocols Ahmad et al. [12] 2011 QoS-aware MAC protocols for WSNs Muzakkari et al. [13] 2010 QoS provisioning at the levels of routing, MAC layer, and cross-layer Shatnawi [14] 2016 IoT application layer protocols Karagiannis et al. [15] 2015 Application layer protocols for the Internet of things Bernard et al. [16] 2019 Cross-layer QoS strategies for the IoT in WMSNs AlAmri and Abdullah [17] 2017 Cross-layer QoS protocols for WMSNs Guo et al. [20] 2012 Cross-layer and multipath-based video transmission for WMSNs Ksentini et al. [21] 2006 Toward an improvement of H.264 video transmission over IEEE 802.11e through a cross-layer architecture for WMSNs T-MAC design: the major forms of energy consumption in T-MAC are through (i) idle listening, (ii) collisions, (iii) protocol overhead, and (iv) overhearing. In contrast, in the active period, the node remains listening and sending data. If no activation event occurs for a certain time t, then the active period ends. e activation event might be data reception on a radio, communication sensing on a radio, and end of transmission for the data packet of the node or ACK, knowing that the neighbor has finished transmitting.
Clustering and synchronization: on waking up, a node begins by listening to the channel. If after a certain time there is nothing heard, it chooses a schedule for transmitting Synchronization (SYNC) packets appended with the time of the next frame startup. If during this time it hears another node's SYNC packet, it follows its schedule subsequently transmitting its own SYNC. Following both schedules enables it to have an activation incident at the beginning of both frames.
RTS and t choice: if a node does not get a CTS message after sending an RTS, then either (i) there is a collision preventing the destination node from hearing the RTS, (ii) RTS or CTS is overheard and destination node barred from responding, or (iii) the destination node is in the sleep mode. So, if after time, t, there is no answer, the source node switches to the sleep mode. It should only do so if none of its neighbors is still communicating. A new interval, t, can be triggered by a neighbor's RTS or CTS reception. A node may be out of range to hear the initiating RTS. e length t of the interval should be long enough to get the CTS packet's beginning [25], and t is thus given by the comparison in the following equation: where C � contention interval length, R � RTS packet length, and T � turn-around time. Advantages: (i) It saves power through adaptive duty cycling. Disadvantages: (i) It is unscalable and more complex.
(ii) It offers less throughput and enhanced delay.

CU-MAC.
is protocol supports IoT standards requiring request-response transmissions. It employs a multichannel mechanism for constant bidirectional transmission of packets at low duty cycle to resolve hidden-terminal issues. According to Danmanee et al. [24], CU-MAC is an SI-MAC protocol that is aimed at improving bandwidth of the channel by initiating connections and transmitting data through different channels using the multichannel approach.
Protocol design: the protocol has the following three properties that make it perform more efficiently with WSNs in the IoT: is is enabled depending on how the buffer overflow protection policy is set in the buffer stack.
is stack has 2 parts: (a) bidirectional buffer slots for enabling sensor nodes to support linked data transmission and (b) normal buffer slots for normal data queuing. Buffer stack policies are set as follows: (a) A packet is only dropped in the buffer when it expires. (b) Unless the packet is from a target node to which a receiver is sending the data, the receiver denies packet reception when buffer slots fill up. (c) e bidirectional buffer slot is used for bidirectional buffer transfer.
CU-MAC can reduce power wastage by choosing to employ only a single channel for transmitting both control packets and data in case there is only one sender in the same region of coverage. It can further enable multiple transmissions through another usable channel if there are other sending nodes in this area [24]. e protocol has two phases: (i) advertising phase, in which a source node announces to neighbors before transmission, and (ii) data transfer phase which commences when the source node receives a Readyto-Receive Information packet from the destination node. Advantages: (i) It has a higher packet delivery ratio compared to other MAC protocols. (ii) It performs well in high traffic situations since its duty-cycle is lower. (iii) It reduces power wastage by employing only a single channel for transmitting both control packets and data. (iv) It can enable multiple transmissions through another usable channel if there are other sending nodes.

R-MAC.
is is a reservation-based MAC protocol designed majorly with an aim to be fair and efficient in terms of energy use. To eliminate collision of data packets, R-MAC [26] schedules control and data packet transmissions at the source and destination nodes rather than using RTS/CTS message exchanges. Channel utilization in R-MAC is enhanced by an ARQ method called burst-based acknowledgment which together with other scheduling algorithms addresses the exposed terminal problem besides enhancing the network throughput.
R-MAC design: all nodes in this protocol intermittently function in listen and sleep states so as to avoid wasting energy in overhearing and idle modes. e time taken to listen/sleep is the same for all nodes each of which Journal of Computer Networks and Communications chooses a schedule for itself rendering central scheduling and synchronization irrelevant. So, in the absence of traffic in the locality, the node simply does periodic listening and sleeping, and to eliminate collisions, transmissions are distributively synchronized using a reservation-based model in case there is a sender to transmit data. R-MAC differs from S-MAC in that it transmits data to a number of nodes using the sleep mode [24]. ere are 3 stages in this protocol: (i) latency detection, (ii) period announcement, and (iii) periodic operation, of which (i) and (ii) are for node synchronization and (iii) is for listen/sleep procedures.
(i) Latency detection phase: here, all nodes broadcast a Neighbor Discovery control packet (ND) at an arbitrarily chosen time since they are powered on. When the ND is received from the neighbor node, the receiver notes its arrival time and chooses a random time to send an acknowledgment (ACK-ND)-the same as the received ND in size-in which the duration from ND arrival to ACK-ND transmission I 2 is specified and the time from ND packet transmission to ACK-ND arrival I 1 is computed. (a) Propagation latency between the 2 nodes L � (I 1 − I 2 )/2. (b) us, propagation latency can be defined as "the interval from the time the first node sends the first bit of a packet to the time the second node receives the last bit of the packet" [24]. (ii) Period announcement phase: here, nodes choose their own random listen/sleep time and start time and broadcast these, and when packets are received, the node turns the received schedule into its own. All nodes will record their neighbors' schedules in relation to theirs. ese two phases make a number of rounds to ensure all nodes are informed about their neighboring nodes. (iii) Periodic operation phase: here, there is data transmission [26], and nodes intermittently go into the sleep mode and wake up. One period is a listen/ sleep cycle, and nodes possess similar periods. Two parts make a period T 0 : listen window T L and sleep window T S , as shown in the following equation: In this protocol, the nodes make use of control packets for communication. ese include REV (reservation packet), ACKREV (acknowledgment packet for REV), and DATA/ ACK-DATA (acknowledgment packet for data) message exchanges all of which are similar in size. Advantages: (i) It is fair and efficient in terms of energy use. (ii) It eliminates collision of data packets.

AR-MAC (Adaptive-Reliable Media Access Control).
is is a TDMA-based MAC protocol for wireless body area networks (WBANs) designed to reduce energy consumption as proposed by Rahim et al. [27]. e protocol allocates a guaranteed time slot (GTS) to all nodes for transmission depending on their needs. It utilizes periodic sleep and wakeup to minimize overhearing and idle listening based on the needs of the node. It has a central node (CN) having big batteries with higher computational power and single or double transceivers. For double transceivers, the sum of the time frame, T Frame , is assigned for node transmission. But in AR-MAC, a single transceiver CN is considered with T Frame having 3 subdivisions: contention-free period (CFP) for sensor connection, contention access period (CAP) for emergencies, and time T MS to communicate with the monitoring station (MS).
Channel selection: the CN first scans for redundant channels from which it chooses one for transmission and broadcasts its address and other information to the nodes. e destination nodes search for channels by scanning the radio frequency (RF) channels. If free, it switches to the next, and if busy, it will wait for time T CP listening to packets and again switch to the subsequent channel if it does not get the channel packets. When the node finally succeeds in getting a channel packet, transmission commences and an ACK packet is sent to the CN. Time slot assignment: a Time Slot Request (TSR) packet is sent to the CN by the sensor node after it has selected the RF channel. e TSR packet has data rate and time slot info for the node. Static-size time slots with static guard-band time, T GB , are proposed in [28]. is protocol employs the adaptive algorithm of time slot (TS) and GB time, and the CN allocates TS while sending Time Slot Request Reply (TSRR) based on the node traffic pattern. e size of the time slots varies depending on the node requirements. Based on the transmission model, the data packet transmission, ACK packet reception, and some amount of delay can be tolerated by allocated TS. To prevent interference between adjacent time slots resulting from the node and CN clock drift, T GB is inserted between them, and its value is given as follows: where F � GB factor, depending on the mean drift value. Nodes go to sleep after slots have been successfully assigned and wake up to transmit in assigned slots, thus saving energy wasted in idle listening, since even allocated slots are collision free. Synchronization: for periodic synchronization, TDMA needs additional energy [29]. It needs a lot of energy to synchronize nodes after several cycles. To reduce energy wastage and control collisions, AR-MAC employs a new synchronization scheme. When a data packet arrives, the current and expected packet arrival times with allowable delay (D) are compared by the CN, and a drift value (DV) is computed based on the difference, ΔT. e DV is sent to the node in the ACK packet for changing the slot in subsequent transmissions though the value is based on allowable delay and F. In case ΔT > D, the CN will send the DV in the SYNC-ACK packet for subsequent SYNC to nodes, or else a mere ACK packet is sent by the CN for the received packet. For any forthcoming data transmissions, the node will change its wake-up time plan based on the DV. With such a sync approach, a sleeping node will not lose sync for all cycles, N. Acceptable delay (D) is given as where F is the guard-band factor.
ΔT � expected arrival time − current arrival time, Frame format: the AR-MAC protocol employs two forms of packets, namely, (i) data packets in which a node transmits in its assigned time slot and employs CAP for emergency data and (ii) control packets that include channel packet, Time Slot Request (TSR) packet, Time Slot Request Reply (TSRR) packet, Synchronization-Acknowledgment (SYNC-ACK) packet, Data Request (DR) packet, and Acknowledgment (ACK) packet [27]. Energy consumption: if N is the number of cycles, energy consumption for these is measured as follows: where I Sleep � current from the voltage source V when the node is asleep. If E sw is the switching energy, E trans the transmission energy, E rec the receiving energy, and E Tout the time-out energy, then Advantages: (i) It reduces energy consumption.
(ii) It minimizes overhearing and idle listening based on the needs of the node via periodic sleep and wakeup.

ER-MAC.
It is a combination of TDMA and CSMA methods making it flexible and adaptable to topological and traffic changes and hence applicable in rapid response situations in WSNs. Collision-free slots can be planned using TDMA [30]: ER-MAC is designed in such a way that the following are met: (a) It has a higher ratio of packet delivery and low delay under lower energy consumption levels outperforming Z-MAC. (b) It keeps two priority queues in order to distinguish high from least priority packets. (c) It permits contention in TDMA slots so as to handle enormous traffic volumes. (d) It has a harmonized slot shape, and the nodes are able to locally change their plans enabling them connect to or exit the network with ease. Protocol design: it discovers the topology in such a way that it uses the simple flooding mechanism to build the tree as initiated by the base station (BS). e topology discovery aims at neighbor discovery and change tracking in addition to setting up a routing tree. A TOPOLOGY_DISCOVERY message comprising hop_count, new_parent_id, and old_-parent_id is generated by the BS. A node broadcasts it to discover if it has any potential children and a response to the parent. In case of the desire to change the parent, the former parent is notified. For all nodes, the hop count to the BS, parent ID, children, and onehop neighbor are registered with the BS at this stage. TDMA slot assignment: when a BS sends a SYNC message, it switches to the TDMA mode which when received by the child uses the broadcast slot for its children's synchronization. Neighboring nodes cannot share a slot since slots are assigned and schedules are exchanged accordingly. Slot assignment begins with a leaf node in a bottom-up mode such that the flow of messages to the BS may be smooth. Before getting the child-node schedules, a nonleaf node will not transmit its data via a unicast slot, descendant's data via many unicast slots, and child synchronization via a broadcast slot. Child nodes send SCHEDULE_NOTIFICATION messages to the BS to end this phase.
Local time synchronization is done via parent-child broadcast sync since they both need to operate the same clock so that when one is transmitting, the other is ready to receive. Packet prioritization happens in such a way that a high priority queue is first emptied before a lower priority queue is allowed to transmit. In case a queue is filled up, the shortest slack packet is dropped and then fairness over source is considered as queue modification is done to enable the BS balance info from different nodes. Energy conservation is by sender turning off the radio in absence of any info to transmit, and the receiver will switch to the sleep mode in case no packets are received after a certain timeout [30]. Advantages: (i) It is flexible and adaptable to topological and traffic changes.

Journal of Computer Networks and Communications
(ii) It is applicable in rapid response situations in WSNs. (iii) It has a higher ratio of packet delivery and low delay under lower energy consumption levels.

RI-MAC (Receiver-Initiated MAC).
is is an asynchronous duty-cycle protocol that tries to reduce time occupied by the communicating nodes on the medium to agree on time for data exchange. e sending node waits for the receiving node to signify the beginning of the communication session through a short beacon frame that it sends. e channel is only occupied by the beacon and data transmission, reducing channel occupancy and thus enabling data exchange for other nodes too. is implies an improvement in channel utilization that subsequently leads to increment in throughput, packet delivery ratio, and power efficiency over a wide-ranging traffic load [31]. Being a receiver-initiated model, it ensures substantial reduction in overhearing and lowers the probability of colliding and the cost of recovery compared to B-MAC [23] or X-MAC [32]. RI-MAC is further capable of using a beacon with ACK and Ready-to-Receive packets though it has a problem if a sending node with Ready packets needs to have the radio on till a ready-to-receive node is awake [13] which then sends it a beacon and transmission begins immediately and later is acknowledged by another beacon. Nodes intermittently wake up to listen to any possible frames meant for them depending on their schedules. A node announces its readiness to receive a frame through a broadcast beacon on waking up when it switches on the radio. When a data frame has been received, the receiver will spend some more time, known as "dwell time," in the active state so as to receive queued packets, which is the time in RI-MAC that depends on the number of contending senders. If there are several contending senders that might not have been predicted, there is a challenge in trying to reduce the receiver's active time to optimize power efficiency and reduce the cost of collision detection and lost data recovery even in the case of hidden senders. For this protocol, this can be addressed by a beacon frame from the receiving node for coordinating data frame transmissions of contending sending nodes. e protocol also beats B-MAC and X-MAC in minimizing the cost of collision detection and data recovery through coordination of the receivers' data frames. is is because the receiving node is aware of the maximum delay prior to the arrival of the frame. e ability to broadcast is supported by RI-MAC when it sends data as unicast to each sender's neighboring node or continuously transmits the frame in a back-to-back mode for as long as the sleep interval [31]. Advantages: (i) It improves channel utilization leading to increment in throughput, packet delivery ratio, and power efficiency over a wide-ranging traffic load. (ii) It ensures substantial reduction in overhearing and lowers the probability of colliding and the cost of recovery.

SRI-MAC (Synchronous Receiver-Initiated MAC).
is is a synchronous duty-cycle protocol that implements the method of receiver-initiated data transmission with an aim of enhancing network lifetime through energy wastage avoidance via collision, overhearing, and idle listening. To do this and also minimize duty cycle, it uses a sequence of adaptive beacons and RTS/CTS packets. In this protocol, sending nodes do not go to sleep till they have got CTS packets from receiving nodes to begin transmitting except for nonsending nodes that return into the sleep mode right away as if the channel was idle [33]. Since sending nodes have separate schedules, it reduces the contention probability leading to more energy conservation. A time frame is divided into the information period, the allocation period, and the communication period which are discussed as follows: Information period: at frame initiation, a beacon is sent by the receiving node declaring its return from sleep and readiness to accept data packets in case the channel is not busy. e beacon is a small packet alerting other listening nodes of the transmitter's readiness to get a request for channel. It consists of the receiver's ID and the duration allocation period (DAP) that relies on number of the receiving node's neighbors [13]. Allocation period: here, the receiving node stays awake for some little time after broadcasting its beacon to confirm if there is any node having an RTS packet to transmit. When the beacon has been received, an RTS packet (with 3 fields-sender ID, receiver ID, and data size) is sent by the sender prompting the receiver to send a CTS for SYNC as there is neighbor information exchange. Communication period: SRI-MAC is capable of organizing and scheduling transmissions by means of sensor node IDs. e receiver assigns order and transmitting time for every sender such that they switch off their radios till their time to transmit is due. However, the receiving node remains in the wake-up mode to get all sending node's data. SRI-MAC can enable sensors have lengthy sleeping times during transmission such that energy can be saved as long as it is neither sending nor receiving data. e protocol's communication period is premised on the TDMA technique which conserves energy since the radio's duty cycle is lowered and collision is eliminated [33].

H-MAC.
is protocol is built on the power saving mechanism (PSM) of IEEE 802.11 and slotted Aloha protocols to enhance performance using multiple slots. Time here comprises large frames each having an active and a sleeping part [34]. If any node has got data to send, it will ask for slots from the receiver node during active time and later transmit the packets in the prenegotiated slots. Nodes go into the sleep mode during sleep-time slots if they do not have data to transmit.
Latency in IEEE 802.11: if N is the number of hops, n the value for the current hop, t cs,n the back-off time, and t tx the transmission delay, the total latency is calculated using the following equation: and the average latency is given by the following equation: roughput in H-MAC is given by the following equation: where n p is the number of transmitted packets, n m is the maximum number of nodes connected in a time frame, C is the number of slots with the same length, t act is the listening time, and t s is the sleep delay. Advantages: (i) It enhances performance using multiple slots.

Z-MAC (Zebra MAC).
is rides on the strengths and offsets weaknesses of TDMA and CSMA. It yields very good channel usage and low delay under low contention and lowers collision among two-hop neighbors at a reduced cost. Its performance is strong via error synchronization, failure in assigning slots, and channel conditions that change with time.
In the set-up phase, the protocol operates through certain steps including neighbor discovery, slot assignment, local frame exchange, and global time synchronization, which unless the topology of the network changes significantly do not run again. When the data are being transmitted, there is improvement in network throughput and energy efficiency which covers up for the initial upfront, operational costs [35]. Disadvantages: (i) Initial slot assignment is quite a cumbersome task here since it is done at the beginning, yet this may not be at the same time in reality. It further assumes static links until setup happens again which too is impractical. (ii) It also suffers from the hidden-terminal problem since slot owners have a small contention window and priority for slots allocated to them due to the ECN messages.
(iii) Synchronizing time also poses a challenge in this protocol due to a large clock drift between nodes. (iv) Source and destination nodes too are not well coordinated in Z-MAC [36].

Q-MAC (QoS-Aware Media Access Control).
is protocol offers high QoS with improved energy efficiency and accesses the wireless channel using MACAW as the underlying protocol [37]. Traffic from different nodes is prioritized for QoS satisfaction depending on how critical the data are. It mainly has two scheduling algorithms in WSNs: (i) intranode and (ii) internode scheduling. e intranode scheduling scheme adopts a first-in first-out-(FIFO-) based queuing algorithm based on application and MAC layer abstraction for data packet classification. Internode scheduling reduces idle listening and collision on the channel for improved energy consumption. Advantages: (i) It gives a high QoS with improved energy efficiency.

WiseMAC (Wireless Sensor MAC).
WiseMAC is based on the Aloha protocol and employs preamble sampling in achieving low-power transmissions in infrastructure sensor networks by occasionally sampling the channel to check for activity [32,38]. e protocol employs the same method as B-MAC except that the sender shortens the extended preamble's length by scheduling its transmissions accordingly after learning the receiver's awake schedules.
is is done by the receiver node putting its subsequent awake time on the ACK packet such that when the sender needs to send to it again, it starts the preamble shortly before it wakes up so as to avoid energy wastage in transmitting the preamble [32]. e data frame is repeated instead of the extended preamble for less traffic transmissions that have a lengthier preamble than the data frame in WiseMAC. After processing the frame, a node that is not the target recipient will go back into the sleep mode, while the recipient stays awake till the session ends thereby sending an ACK packet. For an idle channel, the power consumption is rather low with this method. Nevertheless, there is a problem of long wake-up preamble limiting throughput and causing an enormous power consumption overhead in reception [38]. All nodes overhearing a transmission bear the overhead together with the recipient node. e protocol can address issues concerning low-power transmission but has no mechanism of nodes adapting to varying traffic patterns [32]. To compensate for drift between clock at the access point and on the sensor node, there is a need to compute the wake-up preamble duration. e drift varies directly with time since previous resynchronization and is given by the following equation: where θ is the time-base quartz frequency tolerance and l is the interval between two transmissions. Increase in traffic volume naturally reduces overhearing.

Journal of Computer Networks and Communications
Radio model: for protocols with lower power, transition delays between transceiver states and their power consumption are modeled. Some of the states include DOZE where the transceiver neither transmits nor receives but is prepared to power on to the receive/ transmit state very fast, RX where the transceiver listens to the channel, receives data, or demodulates data from the noisy/idle channel, and TX which is the transmission state by the transceiver. Let T S be the set-up time needed to turn the transceiver from the DOZE to the RX or TX state, T T the turn-around time needed to switch between RX and TX states, and B the transceiver's bit rate. P Z , P R , and P T are the values of power consumed in the DOZE, RX, and TX states, respectively: where P R and P T are the changes in power as a result of being in RX and TX states, respectively. Traffic model: let N be the number of sensor nodes connected to an access point (AP) and λ be the global rate at which the downlink Poisson traffic arrives at the AP from the fixed network. Traffic toward each sensor node is the same and is given by λ/N. Average packet interarrival time L with which data packets arrive at a node is given by L � N/λ. If T C and T D are the control packets and data packets, respectively, assuming a small traffic volume, the global interarrival 1/λ far exceeds the summation of durations for the data packet and control packet and the turn-around time. Mathematically, it is given as e average power consumed by WiseMAC and transmission delay incurred are given by equations (14) and (15) [38]: where where P W and D W are the power consumed by WiseMAC and the transmission delay for WiseMAC, respectively. Advantages: (i) It achieves low-power transmissions in infrastructure sensor networks through preamble sampling while checking for activity. (ii) It reduces overhearing due to increased traffic volume. Disadvantages: (i) It has a problem of long wake-up preamble limiting throughput and causing enormous power consumption overhead.
2.1.14. X-MAC. It is a simple, minimal transmission power MAC protocol that uses a short preamble with decoupling for the sending and receiving node sleep plans in WSNs. is leads to a significant reduction in energy usage at sending and receiving nodes and decrease in per-hop latency and gives flexibility in adapting to bursty and periodic sensor data sources [32].
Protocol design: X-MAC's design is aimed at increasing energy efficiency, reducing data latency, and increasing throughput. is enables application in different kinds of packetized and bit stream digitized radios as well as simplified, low-overhead, distributed application. e design is as such in order to address the low-power listening problems including overhearing, excessive preamble, and incompatibility with packetizing radios as in [32]. Asynchronous duty cycling: a node with information to transmit will send the packets after sending a preamble, whereas the remaining nodes will preserve their sleep schedules that are not synchronized. e receiver will sample the link on waking up and stays awake if it detects a preamble and if it is the target; it will return to sleep if it is not the target after getting the whole preamble.
Embedding the target ID in the preamble to avoid overhearing: the lengthy preamble is divided into many shortened packets embedded with the receiver node's ID so as to eliminate overhearing. One lengthy preamble is constituted by the shortened preamble packet stream. Any node on receiving the short preamble after waking up will examine the target ID on the packet. It goes back into the sleep mode right away resuming duty cycling if it is not the targeted recipient or stay awake for successive packets if it is the one. is approach checks energy wastage due to network density. It can also be used on several types of radios.
Reducing excessive preamble using strobing: it is possible to enable low-power transmission and also save energy by reducing time for preamble transmission through the use of enhanced preambles in addition to preamble sampling. X-MAC does not send an endless preamble packet stream but puts slight pauses in between packets to allow the sending node to pause and listen to the channel and allow the receiving node to resend an early ACK packet during that pause. is makes the sending node begin sending data packets and the receiving node cut excessive preambles and thus decrease per-hop latency and energy wastage on unnecessary waiting and transmission. Packetizing radios: low-power listening is limited in capability for supporting packetizing radios but X-MAC's short strobed preambles can support all types of digital radios. Adaptation to traffic load: while a number of WSN applications are reliable in traffic production, they still must adapt to changing traffic. Nodes are going to have separate sleep schedules depending on traffic load. Considering systems that have time-fluctuating traffic, all predetermined static schedules are suboptimal. Optimality: Buettner et al. [32] model the expectation of energy as follows: where E s � expected energy to send a packet; P Tx and P Rx are the transmission and receiver power, respectively; S p , S al , and S d are the duration of the sender's preamble, ACK listen, and data transmission periods, respectively; and R l and R s are the receiver's listen and sleep periods, respectively. Expected energy to receive a packet is given as E r � (listen cycle energy + sleep cycle energy) * (expected iterations for preamble arrival) +(energy to send ACK) +(energy to receive packet), And the one-hop expected latency is given as L at � (preamble duration + ACK listen) * (expected number of iterations) + duration to send packet, Advantages: (i) It increases energy efficiency.
(ii) It reduces data latency.

A-MAC (Advanced MAC).
A-MAC is a TDMAbased protocol whose design is aimed at enhancing the lifetime of the network for low rate and reliability of data transmission [39]. To assign time slots, it does not depend on a central controller like most protocols, but it gathers information from its neighbors so as to allocate the slots through a distributed algorithm. Energy is controlled by scheduling power when there is not a single event. Its design comprises many frames with each having many time slots which at the beginning have got a beacon transmitted to synchronize and exchange information with neighbors. Nodes to participate in the subsequent session are selected by the controlled node, so they do not go into the sleep mode with others [13]. is protocol operates under four time slot divisions, namely, initial, wait, discover, and active states. A node in the initial state begins by listening to the medium for the beacon packet from other nodes for network synchronization, which when received makes it change the timer through subtraction of beacon transmission time from reception time. To avoid synchronization problems due to drift and allow for continuity, the node has to get the strongest beacon signal. It then enters the wait state after choosing a number of waiting frames so as to reduce the probability that nodes get into the discover state simultaneously. In case the synchronization beacon is lost while still in the wait state, it will return to the initial state for another one. On expiry of the waiting counter, it enters the discover state and collects information from neighbors by listening to their beacon messages. On successful selection of a time slot, it enters the active state where it endlessly transmits beacons at the initiation of the slot. A node goes to sleep when a beacon has been transmitted and there are no more data for transmission or in case the neighbor's beacon received indicates no data packets are coming in [39]. Advantages: (i) It can enable multicasting.
(ii) It enhances the lifetime of the network.
(iii) It offers reliable data transmission.

L-MAC (Lightweight MAC).
Based on TDMA, this protocol is aimed at minimizing the number of transceiver switches, adapting the nodal sleep interval to data traffic amount, and making implementation less complex [40]. A packet is transmitted comprising a control and a data part. Control packets have a fixed size and the time slot controller's ID. ey show the distance from the transmitting node for simplified routing to the network gateway as well as addressing target node and reporting data unit length and internode synchronization. Adjacent nodes are more interested in getting control packets from nearby nodes. e node's transceivers are switched off until the subsequent time slot unless the node is also addressed or the message is omnicast. If so, it listens to the data unit which may not fill the remaining time slot in totality. e transceivers on either side are turned off when the transfer of packets is successful. Energy wastage through idle listening in uncontrolled time slots is avoided via a short time-out interval. Network nodes Journal of Computer Networks and Communications 13 are able to communicate without colliding. is saves energy and increases the lifetime of the network. A node in L-MAC can only transmit one message per frame [13]. e network is set up in such a way that all nodes are not synchronized on being switched on. e gateway takes control of a timeslot so as to have them synchronized, and its immediate neighbor gets the control messages thereby synchronizing their timers with the gateway's. Neighboring nodes learn of multiple gateways' time slots in their vicinity just after one frame. e just synchronized nodes randomly choose time slots for controlling as long as they are not occupied. e nodes thus preserve a table for the neighboring nodes which makes a timeslot reusable after a minimum of 3 hops and prevents message collision. All nodes keep their slots before the battery is down or learn of an impending collision. All nodes can track their hop distances to a selected gateway node, and the information is communicated in control messages [40]. Advantages: (i) It minimizes the number of transceiver switches.

C-MAC.
is was designed for WSNs with 3 segments, namely, (i) aggressive RTS, having double channel check to assess the channel, (ii) anycast to initialize the flow, and (iii) convergent packet forwarding to stabilize the flow. It promises an improvement in energy efficiency, lower latency, and enhanced throughput while avoiding synchronization overhead though it operates only in low dutycycling apps especially where there is less traffic. But when there is more traffic, the protocol employs anycast-based packet forwarding for waking the nodes up or finding a forwarder quickly and converges from route-suboptimal to route-optimal unicast. After this, energy could be preserved by the nodes using a synchronized wake-up plan [13]. Disadvantages: (i) It operates only in low duty-cycling applications especially where there is less traffic.

Diff-MAC.
is is a CSMA/CA-based, QoS-aware, hybrid-priority-based MAC protocol for WMSNs. It targets enhancing the channel utilization while making use of efficient service differentiation to do coordination for medium access of each traffic class while giving fair, quick data transfer. It offers QoS using the following features [14]: e implementation of Diff-MAC on Imote2 indicates that resource requirements can be met on the sensor hardware. Results of extensive simulation runs showed that Diff-MAC outperforms its competitors.

Cluster-Based On-Demand Multichannel MAC Protocol for WMSN (COM-MAC).
is is an energy-efficient protocol in WMSNs with high throughput and reliable data transmission. It achieves energy efficiency by scheduling multichannel media access to reduce channel contention and remove collision, idle listening, and overhearing [41].
Design: it has a three-phased operation with the (i) request phase, (ii) scheduling phase, and (iii) data transmission phase. Request phase: here, all nodes have two protocols each for sending requests to the cluster head (CH), i.e., the contention-based and contention-free TDMA/FDMA protocols. e contention-based protocol allocates a channel per node for sending the request message, and free channels are useable as control channels. For contention-free protocols, a control slot is allocated when a network is deployed. Here, a slot per channel is allocated to a different node enabling requests to be sent without interfering with transmissions from other nodes. e contention-based protocol operates under two steps: (i) Control channel assignment phase: a channel is assigned per node for transmitting a request (REQ) message. A channel is assigned to multiple nodes, and nodes are evenly distributed to available channels for congestion control. (ii) Request transmission phase: nodes with data send a REQ message via the assigned control channel to notify the CH which responds with an ACK message. e contention-free TDMA/FDMA protocol too has two phases including the control slot assignment and request transmission phases. For unreliable channel conditions and low traffic load, the contention-based protocol is fit for use because of less collisions and congestion, whereas the contention-free protocol is preferable in heavy traffic load and reliable channel applications. Scheduling phase: from the above phase, a schedule is generated by the CH for coordination, and this enables nodes to transmit and broadcast their data via control channels. e schedule for broadcasting only comprises data for nodes assigned to a channel. e schedule is rebroadcast for transmission reliability enhancement, but for energy efficiency, the node's transceiver is switched off when all the schedule is completely delivered till it is the time for transmitting again. e time slot and data communication radio channel are all included in the schedule generated for each node. Data transmission phase: on getting the schedule, the sensor node sends its data as per the time slot and the channel assigned. e time slot is divided into data transmission and ACK sections. ACK supports link layer error control. Using implicit selective repeat ARQ, the CH sends an ACK message for all received packets. If there is no ACK, the packet is assumed lost, and hence, a retransmission during the subsequent interval is done.
is enhances reliability of transmissions though excess delay is possible here which is also intolerable for realtime critical apps like video or audio. is gives rise to a hybrid MAC protocol to exploit the unused spectrum during data transmission sessions [41]. Advantages: (i) It achieves energy efficiency by scheduling multichannel media access to reduce channel contention and remove collision, idle listening, and overhearing. (ii) It maximizes network throughput by dynamically allocating time slots and channels for nodes by execution of a traffic-adaptive, QoS-aware scheduling algorithm depending on needs for QoS and traffic data information.
(iii) It increases transmission reliability by incorporating spectrum-aware ARQ to avoid spectrum wastage.

EQoSA.
is protocol is developed for video and image transfer across WMSNs in a hybrid mode with provision of QoS. Concerning the active sensor nodes and their traffic loads, it employs dynamic session sizes and changes bitmap-assisted (BMA) MAC's fixed session size. Nodes declare any data available for transmission during contention time.
is forces the cluster head to make a schedule of how slots have been assigned. It communicates this schedule to the nodes giving each node a given number of slots per session and so enabling the busty traffic to be accommodated [14]. e protocol does not perform well with DiffServe mechanisms but does well with schemes supporting busty traffic loads, and strong cluster heads are necessary in WMSNs that can assign and broadcast slots.

QoS-Supported Energy-Efficient MAC (QE-MAC).
Based on the IEEE 802.11e standard, this protocol ensures fairness and has low latency and jitter and better efficiency on energy usage than other QoS-aware MAC protocols for WMSNs. It operates under a two-phased routine. e first phase introduces the innovating priority mechanism in IEEE 802.11e QoS, and several data types are assigned priorities. Its coordination function is a combination of distributed, contention-based channel allocation and centralized, polling-based channel access mechanisms. e second phase preserves the nodes' energy using dynamic duty cycling.
ere is an exchange of RTS/CTS packets to supplement CSMA/CA by muting nodes near the sender/receiver during packet duration [42]. Advantages: (i) It ensures fairness. (ii) It has low latency and jitter and better efficiency on energy usage than other QoS-aware MAC protocols for WMSNs.

Black-Burst (BB) Contention.
A contention-based MAC scheme for enhancing instantaneous QoS access to carrier-sense WSNs is proposed in [42]. In this scheme, nodes contend for the channel until it is free. Real-time packets are given higher priority over non-real-time packets, and if a node is transmitting real-time packets, it is prioritized over other nodes [15]. Voice and video are some of the real-time apps in consideration here which periodically access the common radio channel when transmitting. ese need constrained end-to-end delay of packets at the MAC layer. is protocol majorly does the following [42]: (i) Prioritization of real-time traffic (ii) Implementation of the round-robin mechanism in real-time nodes (iii) Bounded access delays among real-time packets (iv) Implementation of real-time communication with varying needs of bandwidth, good for multimedia traffic A mechanism for dynamically assigning packet priorities based on their deadlines and the traversed hops is proposed in [43].
e ReAllocative Priority (ReAP) mechanism is meant to provide QoS at the MAC layer to video traffic. is scheme introduces adaptive TXOP by modifying the TXOP dynamically based on the queue length [15]. Other protocols in this category include dynamic duty cycle and adaptive contention window-based MAC protocol, frame sharing (FRASH) MAC protocol, and real-time independent channels (RICH) MAC protocol. Advantages: (i) It can implement real-time communication with varying needs of bandwidth.

Challenges and Opportunities
(a) One of the challenges in satellite transmission is cross-talk leading to interference in different frequency bands. It is necessary to increase the rise and fall times as well as protect sensitive nodes as we circumvent floating ones if we are to regulate crosstalk as in satellite communication therefore. We further need to investigate the possibility of using differential signaling. is constitutes an opportunity for further research. (b) Another challenge is that the hidden and exposed terminal problems are still noticeable despite the use of RTS/CTS solutions in MAC since collisions are still possible from transmissions of separate nodes. At the MAC layer, ACKs might block efficient channel usage by prohibiting exposed terminals from reusing the channel and may be eliminated leaving no proof of sender data packet reception. Using a directional antenna is one of the proposed techniques of high-speed WSNs like IEEE 802.11. is constitutes an opportunity for further study into these issues for better solutions, though. (c) Another challenge is that, since nodes in contentionbased MAC protocols are mobile, there are many collisions by the packets and schedule inconsistencies in schedule-based protocols in the twohop neighborhood information when nodes are entering or leaving [44]. Deng et al. [45] propose a mobility-based clustering (MBC) protocol for mobile nodes to enhance the packet delivery rate. is protocol in comparison with CBR has 25% chances of decreasing collision-induced packet losses and 50% chances in comparison with low-energy adaptive clustering hierarchy (LEACH). (d) e process of analyzing and evaluating routing performance for WMSNs is quite complex because of the bursty multimedia traffic. Because of the way encoding scheme frames are structured as well as the changes in between the scenes that occur naturally, the compressed video frames display some substantial burstiness on several time scales. ere will also be frequent route recalculation that is as well energy intensive for nodes frequently capturing and transmitting energy. e solution here could as well possibly be the use of the MBC protocol. Route recalculation too should be dynamic to cater for the mobile nodes which might be in and out of the range at different instants of time. Recently, the use of smart dumpsters for effective waste management in smart cities has also been proposed as a possible solution [46]. (e) For WSNs, their routing protocols do not look at multimedia applications which require enormous bandwidth and processing energy and should be transmitted in real time with utmost fidelity. Extra challenges in routing are imposed in WMSNs in deployment of heterogeneous sensors since there could be audio, video, and still pictures in addition to scalar data during transmission. Heterogeneous data handled also have different QoS requirements, and with many different business needs, different services with varying requirements give rise to a big challenge for the routing design. ere is a need to design protocols that can support WMSNs such as EQoSA [14] for this purpose.

QoS Strategies at the Application Layer
We discuss many protocols used to solve different needs of communication between machines.

Constrained Application Protocol (CoAP).
It is a request/ response protocol designed by the Internet Engineering Task Force (IETF) for constrained devices such as those having insufficient RAM or CPU and WPANs with low-power constraints, resulting in poor transfer of packets and enormous overhead. e protocol was designed for machine-to-machine (M2M) apps and system automation to reduce bandwidth requirements, increase packet transfer, decrease the overhead, and make work simpler, leading to lightweight implementation. HTTP commands like GET, POST, PUT, and DELETE are used in a client-server architecture. Using the publish/subscribe architecture, CoAP can support multiple users and hence yield better performance via multicasting, enabling asynchronous message exchange. is protocol can support both multicasting and unicasting via UDP and hence is a reliable and simple protocol. is is done via the messaging sublayer (for reliability based on UDP) and the request/response sublayer (for communication and interaction) [16].
It uses the following messages: (i) Confirmable message: this guarantees reliability using ACK messages sent by the receiver node. ese could be synchronous or asynchronous if there is a need for more computational time. CoAP has no in-built security features. It is secured by the datagram transport layer security (DTLS) that runs on top of UDP though it was not designed for the IoT. is makes its suitability debatable since DTLS offers no support for multicasting. It also needs more packets for handshaking, thereby increasing the traffic volume and computational resources and so shortening the lifespan of mobile devices since they use batteries [17]. Advantages: (i) It can support multiple users and hence yield better performance via multicasting.

Message Queue Telemetry Transport (MQTT).
Developed by IBM and later improved by OASIS with an intention of decreasing bandwidth, MQTT is a simple, opensource publish/subscribe protocol analogous to the clientserver version. It is appropriate for lightweight IoT applications and M2M transmissions in constrained situations like limited power, computation capability, and memory, or bandwidth. e network bandwidth and the computational resource usage are decreasing since clients are not updated making publish/subscribe protocols outperform request/ response protocols via IoT requirements [16]. is protocol can run over TCP/IP and so has the following advantages: (i) Delivering of packets with reliable guarantee (ii) Enabling multicasting (iii) Remote device session establishment (iv) Reduction of transport overhead and protocol exchange to minimize network traffic (v) Provision of the notification mechanism when anomalies occur [47] MQTT's design targets economical use of bandwidth and battery, so it has a major application in Facebook Messenger. MQTT ensures QoS and reliability of message delivery by providing the following three QoS levels: (1) Fire and forget: a message is sent once depending on the network's best effort, and there is no need for ACK (lowest QoS level). (2) Delivered at least once: a message is sent at least once, and an ACK message is needed to avoid duplicates. (3) Delivered exactly once: by a four-way handshake, it ensures a message is sent exactly one time (highest QoS level) e architecture: it has three key components: (i) publishers-lightweight nodes sending data to the broker before returning to sleep; (ii) broker, which organizes sensory data in topics and sends them to interested subscribers; and (iii) subscribers-IoT apps interested in data sent by sensors through the broker, as shown in Figure 3.
MQTT is a simple, lightweight protocol that is better than CoAP via low and high packet loss probability in terms of throughput and exhibits lower latency in low sampling rates [48]. It supports a number of implementations for embedded devices and mobile applications and offers web-socket support too. Nevertheless, it has some shortcomings: (i) It is only employed in simple data types. (ii) It exhibits high latency in high sampling rates. (iii) It is unsuitable for real-time apps in such cases. (iv) It does not provide automatic discovery and/or sufficient security at the protocol level [49].

Secure Message Queue Telemetry Transport (SMQTT).
is is an extension of MQTT that is based on lightweight attribute-based encryption using elliptic curve cryptography. e protocol has three entities, as shown in Figure 4: (i) publisher that publishes information under a certain topic, (ii) subscriber that obtains the same topic information via a broker, and (iii) public key generator (PKG) or broker that acts as the trusted third party. e protocol has four phases, namely, (a) set-up phase, (b) encrypt phase, (c) publish phase, and (d) decrypt phase.
(a) Set-up phase: publisher and subscriber provide a unique ID with attributes for registration with the PKG. is generates a secret master key set and public parameters that it publishes with set U � {A 1 , A 2 , . . ., A n }. For ciphertext policy-attribute-based encryption (CP-ABE), the PKG gives a key set to devices with attributes [50]. (b) Encrypt phase: based on the access tree, the publisher creates an access policy. An access tree is sent by the publisher to the PKG for generating the key-generation policy in the key policy-(KP-) based ABE algorithm. e publishers know beforehand the topics and subscribers to access as well as the rules for access. Subscribers are issued with private keys for all communication sessions. Publishers in the CP-ABE algorithm come up with access trees and rules, where the payload is encrypted by the publisher who also issues more info needed for decryption and the rules. With the help of public key cryptography, data are encrypted giving rise to ciphertext on the basis of KP/CP-ABE and the receiver gets SUBACK from the broker [51]. (c) Publish phase: ciphertext is embedded in the SPublish command as payload, and topic name is defined in the header by the publisher. e broker receives the SPublish packet and replies with PUBACK prompting the publisher to respond with the PUBREL packet. e broker then broadcasts to topic subscribers and deletes the data, forwarding the PUBCOMP packet to the sender node. (d) Decrypt phase: the encrypted information is decrypted by the subscriber with the help of a secret key. Considering KP-ABE on verification that it satisfies the access policy, the subscriber invokes the PKG for a corresponding key which it sends after verifying the request. With CP-ABE, subscribers use the private key for decryption after authentication, thereby enabling offline interaction [51]. Advantages: (i) e information being encrypted makes it more secure.

MQTT for Sensor Networks (MQTT-SN).
is is a version of MQTT designed to operate under environments of low bandwidth, high link failures, and short message length. It is a publish/subscribe (pub/sub) protocol designed for UDP-based WSNs to visualize communication for lowpower devices [51,52]. It comprises a publisher device acting as the MQTT-SN client to forward communications to a gateway that also changes the MQTT-SN message to the MQTT message before delivering it to the broker that sends it to MQTT-SN subscribers [51].
It has a gateway for changing protocols from MQTT to MQTT-SN ( Figure 5) [52]. is protocol supports nodes in the sleeping mode using an offline keep-alive procedure that can be used by battery-operated devices to get into the sleeping mode. To support sleeping clients, MQTT-SN has a new offline keep-alive procedure with which battery-operated devices can go to the sleeping state in which their messages are buffered at the server until they are awake. In MQTT-SN, clients with no preconfigured gateway address can figure out the real address of their operational gateway within the network through a discovery procedure. Simultaneously, we may have several gateways in a single wireless network that can be cooperative via load sharing.
MQTT-SN architecture: it is composed of MQTT-SN clients, gateways, and forwarders ( Figure 5). rough the MQTT-SN gateway (GW), the MQTT-SN protocol is used by the MQTT-SN clients to connect to the MQTT server. Integration between the MQTT-SN GW and the MQTT-SN server might be possible or not. Translating between the MQTT-SN and the MQTT for a stand-alone GW is done using the MQTT-SN protocol. If a GW is detached from the network, it can be accessed by the MQTT-SN clients through a forwarder. is just encapsulates the received MQTT-SN frames that are on the wireless side before forwarding them to the GW unaltered. It then decapsulates the received frames too, resending them to the client as sent by the GW [52].
ere are 2 types of gateways: (a) transparent gateway and (b) aggregating gateway, as shown in Figure 6. (i) Transparent gateway: this launches and sustains a direct MQTT connection between the MQTT-SN client and the MQTT server which is entirely reserved for end-to-end, client-server messaging. e number of servers and that of MQTT-SN clients connected to the GW are the same. is transparent GW translates syntax between these protocols and is easier to implement compared to the aggregating GW. But the MQTT server must support di erent sessions for all connected clients. Some MQTT servers are con gured to enable few parallel sessions. (ii) Aggregating gateway: this maintains a single MQTT connection to the server on behalf of all clients. All exchanged messages from the MQTT-SN client to the aggregating GW are directed to the GW which sorts the information to forward to the server. e aggregating GW decreases the number of MQTT sessions simultaneously supported by the server in WSNs with many battery-operated sensors and actuators (SAs) having limited storage and processing capabilities. It is however more complex to implement compared to a transparent GW. e general format of the MQTT-SN protocol message has two or four message header octets and n octets of variable header that is optional depending on the message in question. MQTT-SN has 256 message types (MsgType) with those from 0X1E− 0XFD, 0X19, and 0XFF reserved for future usage [51]. Advantages: (i) e protocol can operate under environments of low bandwidth, high link failures, and short message length.  (ii) It is a pub/sub protocol designed for UDP-based WSNs to visualize communication for lowpower devices. (iii) It can be optimal for low-cost, battery-operated nodes having low processing power and less storage. (iv) It supports nodes in the sleeping mode using the o ine keep-alive procedure used by batteryoperated devices.

e Extensible Messaging and Presence Protocol (XMPP).
e standardization of this protocol by the IETF made it popular for usage over the Internet especially in instant messaging and chatting platforms. It is however susceptible to spamming attacks, and its failure to have worldwide support has made Google too withdraw its support of the standard [17]. e protocol is mainly useable in the clientserver architecture but can also be applicable in client-client XMPP communication [53]. Among the application layer protocols, XMPP is the only one that supports both asynchronous (publish/subscribe) and synchronous (request/ response) services while running over TCP but can still run over other protocols.
Given the simplicity of its implementation and ease of usability in XMPP, instant messaging (IM) is suitable for XMPP and is implemented using the message stanza <message/>. ese stanzas are usable in one-to-one or multiuser sessions or chatting or in error reporting and sending headlines. Attribute type of the stanza de nes the message type. e attribute type could be a chat, group chat, headline, normal, or error. An instant messaging network in XMPP can support many XMPP clients and servers with a gateway con gured to enable other IM network users to connect to other XMPP users. e XMPP server mainly controls sessions and XML streams to clients and servers in addition to routing XML stanzas in the streams [53]. e TCP ports 5269 and 5222 have been, respectively, registered for the XMPP server-to-server connections and XMPP client connections by the Internet Assigned Numbers Authority (IANA). XMPP enables speci cation of XMPP Extension Protocols (XEP) to enhance its usefulness [17].
XMPP security: clients connect to the server using transport layer security (TLS) and simple authentication and security layer (SASL) protocols which are at the same time used by the servers for interdomain communications. For data con dentiality and integrity, TLS will encrypt the streams of XML for all sessions. To con rm the identity of any client that needs to access the server, the SASL protocol does the authentication of the XML stream. For any communicating session with a server to start, the client needs to resolve the server's DNS hostname. For protection of the server of the client from third-party attacks, the server maintains the authentication details secret, i.e., IP address and access method, and only original connections of the server are required.
In server-to-server communications, the SASL protocol is employed for authentication and con dentiality. Serverto-server communications may as well be deactivated from a server by an administrator depending on the organizational policy. To guard against domain-spoo ng attacks, server dialback can be a good solution since servers support it securing more the XML stanzas. Nevertheless, it cannot be used to authenticate, secure, or encrypt in-server streams, and the server identi cation resulting therein is not strong enough to be relied on. It further cannot secure against DNS poisoning attacks and IP session hijacking for remote domains [53]. High-security domains can use TLS and SASL. Advantages: (i) It is popular for usage over the Internet especially in instant messaging and chatting platforms. Disadvantages: (i) It is susceptible to spamming attacks.
(ii) It has failed to have worldwide support making Google withdraw support of the standard. (iii) It is not secure against DNS poisoning attacks and IP session hijacking for remote domains.

Advanced Message Queuing Protocol (AMQP).
is protocol can be utilized by a number of platforms for message exchange since it is open source and written in many languages. It is a publish/subscribe model depending on a reliable and efficient messaging queue with capacity for instant messaging or e-mail for different applications especially in IoT environments. It differs from others in the sense that message types can be specified and have their source traceable and how to do trade-off between performance, security, and reliability. For reliability of message transmission, AMQP makes use of some delivery guarantees like at most once, at least once, and exactly once. It can also however utilize the TCP transport layer for guaranteeing reliability. e AMQP's publish/subscribe method has two components, namely, (i) exchange queue meant for routing of messages to an appropriate order in the queue and (ii) message queue that acts as storage pending their delivery to receivers [16]. In this protocol model, the broker can make routing decisions contrary to other messaging systems where apps using the queue have in them embedded the decisionmaking logic for recipients or senders. e developer may need to check each of the affected apps so as to change the logic for routing and delivery of the messages [54]. Advantages: (i) It is convenient.
(ii) It has capacity for interoperability between different vendors.

Data Distribution Service (DDS).
A product of the Object Management Group (OMG), DDS, is a pub/sub protocol meant for M2M communications with two sublayers. ese are the data-centric pub/sub sublayer responsible for connecting anonymous data publishers to subscribers and the data-local reconstruction sublayer that is optional and integrates DDS into the application layer [56]. e DDS architecture connects system components, as shown in Figure 7. Publishers and subscribers are disconnected with respect to (i) time, (ii) space, (iii) flow and behavior, (iv) platforms, and (v) programming languages [57]. A DDS data publisher generates topics, and publishersubscriber coupling is represented as topic name, data type schema, and publishers and subscribers' QoS attributes. Delivering of data is done by the publisher layer. e data writer and publishers together make a decision about any necessary alterations for subsequent forwarding to the subscribers who need to send data to IoT applications. Data readers are responsible for delivering topics to the subscribers after reading through them [56].
Some of the major components of the DDS architecture include the following [57]: (i) Topic: this is a logical path specifying publication and subscription's data type between DataWriters and DataReaders. For a successful session to begin, topic names, types, and DataWriters and Data-Readers' QoS should correspond. (ii) Domain: this is a logical setting for transmission that can be utilized for separating and optimizing network transmission in a setup of similar applications. If DDS apps share a domain ID, then they can exchange data among themselves. (iii) DataWriter and DataReader: these are terminal objects for writing and reading messages, respectively, to and from a universal data space. (iv) Participant: this is an object that stands for a publisher or subscriber of the DDS app in a domain acting like a container for other objects. Advantages: (i) It is highly reliable and offers QoS support.
(ii) It is highly scalable.
(iii) It is appropriate for real-time apps.
(iv) It is fault tolerant.
(v) It has high level of interoperability.
(vi) It does not need brokers that may act as bottlenecks. (vii) It offers automatic discovery. (viii) It has a user-defined data structure for topic.
(ix) It supports publisher and subscriber decoupling. Disadvantages: (i) It is originally developed for only stand-alone LANs. (ii) A lot of memory is needed. (iii) It has no open-source libraries to support constrained devices. (iv) It is hard to design and configure compared to other protocols [49].

RESTful Services: REpresentational State Transfer (REST).
is is an architectural style based on a set of principles that define and address networked resources. It offers web services that enable transmission from the HTTP device to the device within the IoT architecture and enables clients to reach the server [16]. is protocol utilizes HTTP commands to backup the request/response messaging model. HTTP is a popular and secure protocol in the WWW using TLS/SSL [16]. A RESTful application has the following characteristics: (i) State and functionality are separated into distributed resources. (ii) All resources are distinctively named by the use of unique HTTP commands (GET, POST, PUT, or DELETE over the Internet).

Journal of Computer Networks and Communications
(iii) It is a client/server, stateless, layered protocol that supports caching. REST uses the same methods as HTTP for all requests or responses, with POST and GET for creating and retrieving resources, PUT for updating and changing the state of resources, and DELETE for removing resources. Advantages of REST over arbitrary web services (SOAP) in the IoT context: (i) It reduces overhead.
(ii) It reduces parsing complexity.
(iii) It is stateless.
(iv) It is easy to implement and learn. is together with its natural t over HTTP makes it a preferred method for Web 2.0 applications to showcase their data and secure M2M sessions in the IoT. (v) Applications that support RESTful web services exhibit better performance than the resource-constrained WSN nodes [59]. (vi) REST is important for the IoT since it supports commercial M2M cloud platforms. (vii) Its implementation in smart phones and tablet apps is also easy since it needs only an HTTP library that is accessible to all operating system platforms [56]. Disadvantages: (i) Although it enjoys wide usage in commercial M2M platforms, it is not likely to dominate because it is hard to implement. (ii) Its HTTP usage makes it incompatible with constrained M2M devices leaving its usage to end applications.

WebSockets.
It is a web-based protocol operating on a single TCP channel and providing full-duplex transmissions.
It permits secure duplex communication between a client executing unreliable code in a restricted environment and a distant node communicating from that code while using the origin-based security model employed by web browsers over TCP [60]. e protocol is not a request/response model and is not a publish/subscribe model. So, unlike other protocols, a web-socket session is established by the client initiating a handshake to the server to launch the session. is, once established, a full-duplex client-to-server connection begins asynchronously and runs until terminated by either party [56]. It gives a model for browser-based apps utilizing bidirectional communication with no need of many HTTP connections to servers. WebSocket comprises two parts, i.e., (i) the handshake having a client's message and the server's handshake response and (ii) the data transfer. HTTP headers should be repeated by applications in all client requests and server responses arising from the endless polling which might increase the communication overhead subject to the application [61]. WebSocket can be used to come up with scalable, real-time web applications since it can enable full-duplex transmission via one socket across the Web. Its security can be ensured with the help of TLS/SSL. is protocol is suitable for resource-constrained devices, and IoTapplications will not have a good user experience with it since it operates a client/server architecture. Nevertheless, it can outcompete all TCP-based protocols because of the following advantages: (i) It is t for real-time transmission.
(ii) It has improved security. (iii) It minimizes transmission overhead. (iv) It can provide e cient messaging systems using WebSocket Application Messaging Protocol (WAMP) [17].

Streaming (Simple) Text-Oriented Messaging Protocol (STOMP).
is is a TCP-based protocol designed for text message-oriented middleware (MOM) with the use of Advantages: (i) It is lightweight for constrained networks.
(ii) It has flexibility to choose quality of services with the given functionality. (iii) It is standardized by OASIS Technical Committee. (iv) It is easy and quick to implement.

Challenges and opportunities:
Here, security threats are imminent because of the openendedness of the application layer [63]. ese could include the following: (i) ere is poor or lack of a security design for an application's function. Coming up with a good design to address this could be a good idea/ opportunity. (ii) Some of the programs might have backdoors evading otherwise secure controls to allow unauthorized access to network resources. We require mechanisms to patch up such loopholes. (iii) Applications that lack or have weak authentication mechanisms can be easily targeted by unlicensed users to abuse the network. Such mechanisms need to be strengthened further. (iv) Excessively complicated access control rules may because of their misunderstanding or poor writing give unauthorized access allowing dubious activity rather than protecting the network they purport to protect. ese should be simple and clear to understand by all users. (v) In booting workstations without disks and managing network devices, we usually use the TFTP protocol although access does not require authenticating the user by username/password. is makes an intruder able to gain easy access to config/access info as long as he can guess the filenames. is needs to be fixed.
A few steps can be used to secure the application layer. To protect network data transmitted across the network, applications need to strengthen their encryption and authentication mechanisms. Apps too need to enable strict controls of privileges for data access using good mechanisms to balance between usability and effectiveness [63]. Applications handling sensitive data need detailed logging and audit capability as well as testing and reviewing. Table 3 summarizes the application layer protocols discussed with their weaknesses and differences among them.

Service-Differentiated Real-Time Communication Scheme (SDRCS).
is is an event-driven routing protocol that routes real-time traffic by featuring a cross-layer packet-forwarding design in which it integrates the real-time routing functionality with a new priority-based MAC scheme. e protocol approximates distributed packet traversal speed for traffic classification and admission control on the basis of this design. It further localizes decision-making by prioritizing packet forwarding to maximize the speed of packets [64]. Advantages: (i) It improves bandwidth utilization: bandwidth degradation due to unschedulable packets can be avoided. (ii) It offers a lightweight packet schedulability estimation mechanism. Here, it uses received signal strength and admission control, and then early missed-deadline packet-dropping policies are made. (iii) It delivers high end-to-end throughput via higher source data rates alongside strict end-to-end latency needs. (iv) Nodes which meet real-time requirements of less traffic load and better channel quality are given highest priority when it comes to packet forwarding. e SDRCS operates via 5 mechanisms, namely, (1) RSS-(Rich Site Summary/Really Simple Syndication-) based grouping, (2) admission control, (3) prioritized queuing, (4) real-time MAC, and (5) dynamic forwarding. is protocol has a queuing policy that employs per-hop deadline-based queues, and the nodes schedule their packets using FIFO priority-based queues [1]. e principal features of this protocol are mentioned as follows: (i) It is event driven and is easily adaptable to any network changes. (ii) It does not need additional hardware for localization or multichannel transmission. (iii) e cross-layer packet-forwarding strategy can be used when transmitting multimedia traffic. (iv) It is able to efficiently circumvent any voids therein.
Early deadline miss (EDM) policy drops all unscheduled packets. e SDRCS circumvents wrong packet drops due to EDM since it adjusts to network changes easily.

Sensor Fuzzy-Based Image Transport (SUIT).
is protocol regulates congestion using a scheme based on fuzzy logic. It transmits packets with reduced quality to allowable levels if there is congestion, enabling it to transmit lowerquality packets/frames without being dropped. e number of packets delivered per second thus rises, and while this enhances video streaming apps, it is more appropriately made for those that quickly transmit non-real-time video such as in video surveillance. SUIT can stream JPEG pictures other than the traditional predictive video coding methods that are not suitable for sensor devices because of their constraints of energy, memory, and computational speed. Progressive JPEG-PJPEG performs better via packet transmission rate and delay though it degrades the image quality because of congestion [65]. is protocol employs cross-layer methods for interlayer data exchange, though it still employs some layering, making it still possible to preserve all advantages of the layered design since it has application and transport layers but has no MAC or routing layer.
For congestion detection and control, SUIT bases on the percentage number of in-bound and out-bound frames per window, number of contenders, and buffer occupancy of the next-hop node. Since buffer management has an effect on effectiveness and QoS in WMSNs, it is useful in the transport layer, so the protocol prioritizes packets for better QoS. SUIT reorders packets by tracking the sequence numbers of the delivered packets since they all have a source ID, frame number, and sequence number that it uses to detect any missing packets. e protocol however does not provide packet reliability [65]. Advantages: (i) e protocol prioritizes packets for better QoS.
(ii) It can transmit lower-quality packets/frames without being dropped. Disadvantages: (i) It does not provide packet reliability.

Network Layer QoS Support Enforced by a Cross-Layer Controller (NLQS)
. is scheme permits packet-level service differentiation as a function of throughput, end-to-end packet error rate, and delay [66] (see Figure 8). is enhances network layer QoS. It comprises a cross-layer control unit (XLCU) for configuring and controlling networking functionalities at physical, MAC, and network layers [18].
is depends upon the unified logic which affects choices for application layer needs and status of functional blocks which do implementation of networking functions [66]. Advantages: (i) Cross-layer interactions can be controlled without weakening the upgradability, modularity, and simplicity of designing the system.

Cross-Layer Signaling Shortcuts (CLASS).
CLASS has a high level of effectiveness, flexibility, and comprehensiveness. According to [67], it has the following features: (i) Direct signaling among remote layers leading to faster transmission. (ii) Lightweighted internal message format: standardized protocols are not suitable for use in internal signaling since they are not usually lightweight, for instance, transmitting against faults in the network. CLASS requires only three fields, namely, destination address, event type, and event contents. (iii) e setup of its exterior messages is standard: considering external signaling, ICMP and TCP/IP are suitable for general messages and short notifications, respectively. (iv) Message control protocol should ensure optimized and organized exchange of dense simultaneous messages across layers for high efficiency and avoidance of possible conflicts. e protocol applies to many cross-layer signaling situations. Advantages: (i) e protocol is scalable and has a very low propagation delay. (ii) It is very e cient and exible since it uses direct signaling between any two layers. (iii) It is very comprehensive and e ective.

Cross-Layer and Multipath-Based Video Transmission (CMVT).
CMVT is a hybrid of application and network layers. At the application layer, it carries out the encoding of video streams to video data frames (I-, P-, and B-frames) by means of the MPEG-4 encoding format. At the network layer, route discovery and data transfer take place. Two algorithms, greedy forwarding and rollback, are used to discover a number of paths from the source to the sink node via the route discovery method [20]. A speci ed node i calculates the evaluation of its adjacent node j using the following equation: where f ij evaluation value for the node i to j; d 2 (j, D) is the distance from the node j to the destination node D; d 2 min (i) and d 2 max (i) are the minimum and maximum distances of neighbors of the node i to D, respectively; e init (j) is the initial energy of the node j; e res (j) is the current residual energy of the node j; and α is the energy coe cient given as follows: where e max (i) and e min (i) are, respectively, the maximum and minimum energy left of all node i's neighbors. e network layer further transmits video streams with CMVT doing status evaluation choosing an appropriate communication path for any packets, and the QoS guarantee level for the path i is calculated as where f i is the evaluation value for the path i, h i are the hops for the path i, n i is the summation of packets sent via i, n i is the summation of packets sent by sources, and ω is the energy consumption factor [18]. Advantages: (i) e CMVT protocol is very superior in media transmission, particularly large-scale WMSNs.
ere is a reduction in interference of neighboring cells leading to uniform coverage in densely deployed networks [68]. A connection is established by two-way authentication between the sender and the receiver. When a Clear-to-Send (CTS) packet and short interframe space (SIFS) are obtained by the sender, packets are transmitted right to the receiver with no valuable cooperation. If available, the two nodes find out whether they can exchange data. is is done by the use of a helper identification signal to establish how feasible cooperative communication is. It is established in the presence of a signal, and if it is not available, direct communication is activated. For distributed systems, helperinitiated cooperation is preferable because of the presence of the RTS/CTS packets [69]. Advantages: (i) CoopMAC has an advantage over others due to the spatial diversity amongst the three nodes [70]. (ii) It has the potential to achieve significant throughput and delay with no complexity in the design of the system.

Cooperative MAC Protocol for Multihop Networks (M-CMAC).
Just as CoopMAC, M-CMAC is made in such a way that high data rate stations help low data rate stations to forward traffic for broadcasting [71]. ere are also helpers chosen in such a way that 2 fast-hop transmissions replace a slow hop transmission. e helper via whom there exists least delay from the sender to the receiver is considered best because of the minimum two-hop transmission rate and so is the chosen neighbor to the duo. Here, the working assumption is that all nodes have their position coordinates known, and thus, the Euclidean distances between all pairs can be calculated and converted to the data rate for that link. All nodes have cooperative tables (CT) of potential helpers with destination and helper MAC address, Euclidean distance, and total distance through the helper. Any sender with data to transmit checks out for the existence of any helper in the CT for the destination and if available forwards RTS for channel reservation for single-hop duration. M-CMAC has got an RTS format having a shape with five fields similar to the one given as follows: Frame control Duration Source address Destination address Helper address e destination node's address is kept by the source in the part for the helper address that is in turn saved as the destination address. Nodes inspect helper and destination address fields on receipt of RTS, whereby the node acts as the other node's helper in case the helper address field is not the same and the helper transmits CTS back to the source if at all it wants to forward data. On delivery of the CTS packet, the source transmits to the helper which in turn sends the packets to the destination that also sends ACK to the helper on receipt of the packet. In M-CMAC, there exists a higher level of channel reuse (parallel transmissions) due to the increased number of nodes therein that subsequently raises availability of helpers for data forwarding.
is results in increased throughput in comparison with CoopMAC and IEEE 802.11 DCF [71]. Advantages: (i) It promises a higher throughput compared to CoopMAC and IEEE 802.11 DCF.

Cluster-Based Cooperative Routing (CBCR) Protocol.
e CBCR protocol has a multihop data-forwarding function realized at the link layer with cooperative links that use M-CMAC. is protocol encompasses two stages, namely, the routing relay selection phase and data forwarding phase.
(i) Routing relay selection phase: all nodes announce their presence to their neighbors by broadcasting periodical beacon messages that carry the MAC address of the node. Each of these constructs a relay table that contains all neighboring nodes with which it is able to communicate. e node further broadcasts its neighbor list in case of any change to its entries since the previous broadcast. e MAC addresses of nodes next to the node X are contained in its relay table's column one, and the neighbor node's row has MAC addresses for neighbors of the adjacent node. Based on its relay table, each node independently chooses routing relays. Choosing a node as a relay node depends on the number of nodes it connects-it should be the highest number [71]. (ii) Data forwarding phase: a node with packets to transmit needs to first verify if the receiver is in the same cluster and if that receiver has a helper to which the packet is thence forwarded but in whose absence the packet is delivered right to the receiver. In case the intended receiver is in a different cluster, the relay table will be checked to see whether that receiver is reachable via other routing relays. If it is reachable, packets will be sent to the routing relay directly or via the helper if at all the relay has one. In case of destination unreachable via relays, packets are multicast to all relays by the node [71]. Advantages: (i) It enables multicasting.

MAC-PHY Cross-Layer Protocol.
A cooperative crosslayer standard for cooperation at the physical layer in nextgeneration WMSNs is developed in [70]. It provides a complete MAC layer algorithm that gives a supportive shell for the PHY-MAC layer. Like CoopMAC, its scheme depends on the middle node for communication between nodes. e MAC layer protocol has been changed to regulate information exchanges at the physical layer. e receiver node receives duplicate packets from source and helper nodes to decode the data [70].

MAC-Centric Approach.
is cross-layer protocol targets multimedia applications by using the MPEG-4 scheme [21]. It is characterized by 4 access categories (AC3-AC0) according to their priority in transmission. It is meant for supporting different QoS needs in upcoming video applications and enabling MAC layer differentiation for H.264 partitioning [18]. Because of low bandwidth, delay, and other QoS challenges that cause inefficiency in transmitting multimedia data in WMSNs, a number of algorithms depending on IEEE 802.11e have been proposed to support transmission of quality videos [72]. Selecting an AC depends on the measures for QoS such as the loss rate and delay.
us, AC3 having the highest priority is mapped to the parameter set concept since the stream is sensitive to bit loss especially in video transmission [21].

Adaptive Cross-Layer Forward Error Correction (ACFEC).
In the ACFEC [73] model, data packets are exchanged among nodes through the access point (AP) which operates the infrastructure mode in which it adds FEC to video data [72]. ese data are dealt by encapsulation using a streaming server to the receiver through the wireless AP as RTP packets. An adaptive FEC controller senses the category of packets out of the RTP header and recovers the header of packets from UDP. e encoder generates some error-correcting packets whose number is determined by the source packet number of the block. Multimedia transmissions are monitored by the controller using MAC failure data, and its counter is increased by one if there is a failed transmission. e controller uses the failure counter to change the packet number produced after transmitting a block [18]. In case there are lost packets, it changes the redundancy rates and produces extra packets to replace the lost ones and satisfy the requirements of the receiving node [72]. e FEC packet number is enhanced or reduced to satisfy the receiver's requirements and stop packet losses. is is done through accurate detection of packet losses and adjustment of redundancy rates. If all video data packets are well received, there will not be generation of FEC packets [73]. Advantages: (i) It promises a good QoS through packet loss reduction and redundancy rate adjustment.

Balanced Cross-Layer Fuzzy Logic (BCFL) Design Routing Algorithm in WSNs.
A new fuzzy logic-based routing algorithm (BCFL) [74] was designed using dispersion of the cross-layer parameter as the fuzzy logic inference system input. Each cross-layer parameter has a dynamic weight depending on the value of dispersion. e design comes with some innovations as follows: (i) for fuzzy logic inference system input, the absolute parameter value is substituted by parameter dispersion, thereby significantly reducing algorithmic complexity; (ii) dispersion does not change with the order of magnitude according to the dispersion formula; and (iii) the weight of the parameter depends on the size of its dispersion, and the two are inversely proportional to each other. is enables BCFL to have some nobility unlike other algorithms as per the following distinguishing properties making it advantageous: (i) It has got simple if-then rules which remain constant even when the constraints increase. (ii) It is capable of dealing with many constraints with no increment in complexity. (iii) It has capacity to yield a more balanced solution than other algorithms. (iv) It is easily adaptable to changes in network conditions and topology even when the changes are frequent like in underwater WSNs. e algorithm is useable in choosing the CH in cluster-based routing protocols [74].

Minimum Hop Disjoint Multipath Routing Algorithm with Time Slice Load-Balancing Congestion Control Scheme.
MHDMwTS is a two-phased routing protocol comprising path build-up and path acknowledgment phases. It consists of multiple sources each of which has three build-up disjoint paths, namely, primary, alternate, and back-up paths. On activation, the source node requests to build up a route to the nearest hop neighbor, which is the path build-up phase [19]. In this phase, step one has source activation in which the requested node for path building adds its number and timestamp and sends to the least hop-count neighboring node. is goes on until the least time latency sink having the information needed to construct the primary route is reached [75]. In step two, there is path extraction when the new package from another path arrives whereby the extracted path is compared with the primary path. If the node is shared, the package will be rejected or an alternative path will be searched for to get a backup through comparison of the preceding two paths. Phase two is path acknowledgment where we have the third step in which the sink returns the ACK packet to the sender with path information having nodes and their time information after it has computed it using the timestamp [18,19]. Advantages: (i) e protocol minimizes end-to-end latency.
(ii) It enhances congestion control. [76] come up with a CLOD for scheduling at the data link layer, routing at the network layer, and controlling congestion at the transport layer with an assumption of fixed link capacity. rough congestion control, energy efficiency is improved. Transport layer congestion at the nodes is minimized by compressed sensing (CS) in which transmitted bits are reduced, whereas optimally allocating resources decreases congestion on the links at the data link layer. CLOD promises minimized computational complexity and better performance in light traffic scenarios [77]. Generating and storing CLOD is more efficient compared to Gaussian random matrices. It prolongs network lifetime and saves energy.

Cross-Layer Optimal Design (CLOD). Authors in
Advantages: (i) It does congestion control subsequently increasing energy efficiency.

Challenges and Opportunities of the Cross-Layer Design (CLD)
(i) Different layers in a real network assume different functions and/or services. A layer only communicates with its neighboring layers. e layered model degrades system performance because of many features in wireless transmission. QoS parameters in the layers of WMSNs are shared by routing protocols to optimize the performance. Nevertheless, there is a need to design cross-layer models for increased efficiency in routing [1]. (ii) e physical layer plays a highly vital role in CLD.
Rate adaptation and channel allocation take place at the physical layer via signal processing to enhance QoS. End-to-end performance in wireless media is affected by the changes therein given the effect on the purposefulness of the protocols at the network layer. CLD offers solutions to conserving power, minimizing energy, and controlling both the flow and congestion in the network, as well as fault tolerance, so it is an opportunity for designers considering other layers. We further desire to develop CLDs that shall make an impact on network operation and get close attention [18]. (iii) CLD interface standardization: the architecture must deliver module functionality although there exist some queries amongst modules about possible interfaces predicted by the necessity to share information among remote protocol layers. Technical challenges include developing, designing, and standardizing of cross-layer interfaces and algorithms that satisfy the cross-layer optimization requirements among protocol layers. Table 4 summarizes these cross-layer models with a few comments.

At the MAC Layer.
In WMSNs, achieving duty cycling is tricky since video traffic is volatile. Listening to adjacent nodes wastes energy. We need to switch states from awake to sleep. is requires further research.
Self-organizing networks and green communication: although most CRWSN energy-saving algorithms are designed to reduce node transmission power, energy consumption is inevitable during the operational mode of BSs because of internal processing. Scholars are trying with policy-makers to encourage a shift to green WSNs to reduce operational costs and carbon footprints. With self-organizing CRWSNs for automatic switch-off when idle, energy is saved, which calls for cognitive switching algorithms. So integrating cognitive switching with radio resource allocation in CRWSNs is a promising research area [78].

At the Application Layer.
Considering a WMSN architecture, end users access the network via the BS to and from which they send their sensed data. We desire to get guaranteed QoS and energy efficiency by designing a new architecture in the form of hardware and software which needs to be investigated further for a fix of the same [1].
QoS and energy efficiency are very crucial in WMSNs and mostly real-time apps that require guaranteed bandwidth and throughput in their network lifetime. Most protocols ignore base station (BS) mobility and WMSN nodes. Traffic management, telemedicine, and battlefield surveillance apps require mobile nodes or BSs, so designing dynamic routing protocols to be adaptable in these circumstances is necessary.

Cross-Layer.
Security of data transmitted on WMSNs, e.g., in military surveillance and e-commerce, is key with QoS and energy efficiency. WMSNs are vulnerable to attacks like worms and sinkhole. Moreover, their computation power is not that high complicating implementation of strong, secure protocols. is needs further attention.
For fairness and priority issues in CRWSNs, nodes are assigned diverse priorities depending on importance and urgency of their data. For better network performance, there is a need to develop fair resource allocation mechanisms with priority-based fairness in sensors. ere is still potential in this research area [78].
We must develop multiobjective, adaptive protocols to optimize QoS metrics involved in routing for best transmission results to get a trade-off between different optimization metrics. We need to design standardized crosslayer algorithms which are able to meet cross-layer optimization standards.

Conclusion
We have reviewed the different QoS strategies for WSNs in the context of IoT from the MAC layer and application layer as well as the cross-layer paradigm. For the MAC layer, we have reviewed protocols for WSNs such as the hybrid of the contention-free and contention-based MAC protocols and those for WMSNs. We further reviewed a number of application layer protocols including many machine-to-machine request/response and publish/subscribe protocols. For system optimization, cross-layer QoS strategies are very important in wireless communication. We have reviewed a number of cross-layer strategies. For all categories reviewed, challenges and opportunities are discussed. Finally, some possible future directions are discussed for research and application showing a promising potential for future research in this relatively new area especially as more WMSN applications emerge recently.

Conflicts of Interest
e authors hereby declare no conflicts of interest regarding the publication of this work.