Data Concurrent Transmission MAC Protocol for Application Oriented Underwater Acoustic Sensor Networks

Xiaoxiao Zhuo,Hong Yang,Meiyan Liu,Yan Wei,Guanding Yu,Fengzhong Qu,*

1 The Key Laboratory of Ocean Observation-Imaging Testbed of Zhejiang Province,Zhejiang University,Zhoushan 316021,China

2 The Engineering Research Center of Oceanic Sensing Technology and Equipment,Ministry of Education,Zhejiang University,Zhoushan 316021,China

3 The College of Information Science and Electronic Engineering,Zhejiang University,Hangzhou 310027,China

*The corresponding author,email:jimqufz@zju.edu.cn

Abstract: With the increasing demand for marine exploration,underwater acoustic sensor networks(UASNs)are prone to have the characteristics of largescale,long term monitoring,and high data traffic load.Underwater media access control(MAC)protocols,which allow multiple users to share the common medium fairly and efficiently,are essential for the performance of UASNs.However,the design of MAC protocols is confronted with the challenges of spatial unfairness,data eruption,and low energy efficiency.In this paper,we propose a novel data concurrent transmission(DCT)MAC protocol,which is able to exploit long propagation delay and conduct concurrent transmission.Specifically,we present the theoretical performance analysis of the proposed MAC protocol in detail and give an analytical solution of the success concurrent transmission probability between nodes.In addition,simulation results are provided to demonstrate that our proposed protocol is appropriate for UASNs and can significantly improve the performance in terms of network throughput and energy consumption.Finally,we give some typical future applications of UASNs and discuss the demands on MAC protocol design.

Keywords:underwater acoustic sensor networks;MAC protocols;data concurrent transmission

I.INTRODUCTION

Over the past few years,humans have never stopped trying to explore the mysterious marine world,which leads to the increasing demand for advanced underwater communication networks.Since electromagnetic waves and light waves are greatly attenuated under sea,they might only be suitable for short-distance underwater information transmission[1–3].Sound waves have been envisioned as the most suitable paradigm for reliable long-distance communications under sea benefiting from its less propagation attenuation and longer transmission distance[4].

Different from terrestrial radio wireless communications,underwater acoustic communications have some unique characteristics,such as low spread speed,high bit error rate,and narrow frequency band[5,6].With the development of acoustic modems over the past few years,current underwater acoustic communications can provide reliable links between smart underwater objects,such as underwater sensors,autonomous underwater helicopters(AUH),autonomous underwater vehicles(AUV),unmanned underwater vehicles(UUV),and surface buoys.Furthermore,these developments encourage the application of underwater acoustic sensor networks(UASNs),which can be widely used in marine environmental monitoring and disaster warning,submarine resource exploration,real-time control of underwater nodes,and military operational command[7–9].However,due to the sophisticated application demand onlarge-scale,longterm monitoring,and high data traffic load,establishing real-time Internet of underwater things(IoUT)based on UASNs is urgently needed[3,10,11].Media access control(MAC)protocols,which allow multiple users to share the common medium fairly and efficiently,are generally acknowledged as significant schemes for UASNs.As such,the notably expanded network poses new difficulties in the design of underwater MAC protocols,such as spatial unfairness,data eruption,and low energy efficiency.

1)Spatial unfairness.To monitor large-scale underwater networks,UASNs are deemed to be sparse and partitioned,leading to different spatial distances among nodes.Propagation delays are proportional to the spatial distances.In some cases,the source node which is closer to the destination node has more opportunities to access the channel,resulting in spatial unfairness.Thus,it means that not only long propagation delays could not be omitted,but also the discrepancy of propagation delays should be taken into account when designing MAC protocols.Novel methods should be proposed to consider the effect of spatial unfairness.

2)Data eruption.As UASNs expand,interoperability between nodes becomes ubiquitous,leading to a dramatic increase in data traffic load.Accordingly,the probability of packet collisions would increase.Thus,it is important to develop appropriate schemes to avoid packet collisions and increase network throughput,especially in the networks with high data traffic load.

3)Low energy efficiency.Due to limited battery capacity and difficulty of battery refilling under sea,the energy consumption of underwater facilities are expected to be minimal.Especially,for large-scale underwater networks,nodes are far away from each other leading to excessive attenuation.These characteristics increase the transmission power burden on large scale underwater acoustic communications.Thus,when designing MAC protocol,reducing redundant packet transmissions is an imperative strategy to improve energy efficiency.

Due to these unprecedented challenges and difficulties mentioned above,most existing MAC protocols are far from satisfactory to meet the performance requirement of UASNs with the characteristics of largescale,long term monitoring,and high data traffic load.Some existing protocols have poor performance due to low channel utilization or high packet collision rate leading to low throughput.Especially,the strategy of frequently sending and receiving control packets in some protocols is a major factor that cause low channel utilization and large energy consumption.

To enhance the performance of UASNs,this paper proposes a MAC protocol named data concurrent transmission(DCT)protocol.This protocol provides an effective solution to solve these challenges mentioned above.It exploits inherent long propagation delay to carry out the concurrent transmission,which improves the channel efficiency and the network throughput.And it also allows different source nodes to transmit data packet concurrently with only once handshaking,which improves the energy efficiency.In addition,since nodes with different propagation delays could transmit packets concurrently,nodes with different distance could have more equal chances to access the channel.As a result,the spatial unfairness could be alleviated.Our proposed protocol has the following three advantages:1)reducing energy consumption by decreasing the number of handshakes;2)alleviating packet collisions and improving network throughput by making use of long propagation delay to transmit data packets concurrently;3)giving the analytical solution of the success probability of concurrent transmission between nodes,which has not been found in the existing literature.

The rest of this paper is organized as follows.In Section II,we provide a comprehensive overview of existing underwater MAC protocols,with a special focus on dealing with challenges of spatial unfairness,data eruption,and low energy efficiency.We then elaborate the proposed DCT protocol in Section III and analyze the theoretical performance in detail in Section IV.Simulation results are presented and analyzed in Section V to show the performance improvement on network throughput,energy consumption,and end-to-end delay.Furthermore,we give some future applications for UASNs and discuss demands on the MAC protocol design in Section VI.Finally,we conclude the whole article in Section VII.

II.OVERVIEW OF UNDERWATER MAC PROTOCOLS

In the past years,various MAC protocols have been developed to improve the performance of UASNs.In this section,we provide an overview of MAC protocols from the perspective of their applications and techniques for handling the challenges mentioned above.Due to the limited bandwidth in underwater acoustic communications,code-division multiple access(CDMA)and frequency-division multiple access(FDMA)protocols are not suitable for large-scale and high data traffic load networks.Thus,here we mainly introduce time-division multiple access(TDMA)protocols.As listed in Table 1,we classify TDMA protocols into two categories,the direct access protocols and the reservation-based access protocols.

2.1 Direct Access Protocols

UW-Aloha protocol[12],a well-known direct access MAC protocol,enabled nodes to get access to the channel randomly.However,it could not solve spatial unfairness because it did not consider the difference in propagation delays.Moreover,the network throughput would significantly decrease in the scenario of high data traffic load due to increased collision probability.To reduce the collision probability,Slotted-Aloha protocol[13]took the strategy of deferring packet transmissions to the beginning of each time slot.In this way,this protocol can successfully avoid ongoing packets being interfered by a new attempt and therefore improve network throughput to some extent.Later,delay tolerant MAC(DTMAC)protocol[14]was put forward to overcome long propagation delays especially in high traffic scenario.It established a probability-based model for throughput to calculate the transmission probability and repeated transmission time.Without channel reservation and acknowledgment,the throughput of DTMAC would not be influenced by propagation delay.Furthermore,other protocols,such as delay-aware probability based MAC(DAP-MAC)protocol[15]and traffic-adaptive receiver-synchronized(TARS)underwater MAC protocol[16],suggested to conduct concurrent transmission by leveraging the long propagation delay.However,when the network traffic became heavy,they would have low channel utilization and high collision probability due to packet collisions caused by spatial uncertainty.In addition,to avoid packets collisions at receivers,the work in[17]derived collision probabilities of ALOHA-based protocols to calculate the annulus interference region.Then it proposed an optimal scheduling protocol for space-time coupling based periodical transmissions.This work provided a potential solution to avoid collisions in random access MAC protocols.

Although direct access protocols have the advantage of low energy consumption because there are no extra reservation packets required,they generally suffer from severe packet collisions when the network traffic becomes enormously heavy.In addition,they have not considered the spatial unfairness resulting in packet collisions and unfair communication efficiency.

2.2 Reservation-Based Access Protocols

The reservation-based access mechanisms,such as handshaking and scheduling,were devised to address the problem of hidden terminal,exposed terminal,and spatial unfairness.Multiple access with collision avoidance(MACA)protocol[18]required a series of handshaking before each data transmission and only allowed one pair of nodes to transmit data packet after a successful handshaking.Although this protocol worked well in terrestrial radio networks,it still suffered from high collision probability due to long propagation delay in UASNs and high energy consumption resulting from frequent handshaking.To improve the handshaking efficiency and success probability,the MACA-like protocols such as MACA-MCP[19],MACA-MN[20],MACA-U[21],FI-MACA[22]were proposed.Besides,to eliminate packet collisions,Slotted-FAMA protocol[23]based on require-to-send(RTS)/clear-to-send(CTS)handshaking mechanism was developed,in which time was divided into several slots and packets could only be transmitted at the beginning of each slot.Although this strategy could avoid data packet collisions,its frequent handshaking between nodes not only greatly reduced network throughput,but also caused high energy consumption.Compared with Slotted-FAMA,multi-session FAMA(M-FAMA)[24]and Slotted-FAMA based protocol with data train called SFAMADT[25]allowed multiple sessions for different pairs,which could increase network throughput.

Moreover,to compensate for long propagation delay,a series of scheduling mechanisms have been proposed in several literatures.The famous protocol named staggered TDMA underwater MAC protocol(STUMP)[26]collected the information of nodes’po-sitions and scheduled the neighbors’transmission to avoid packet collisions.To match the performance of an ideal STUMP,a series of transmit delay allocation MAC(TDA-MAC)[27–29]protocols were designed to improve the network throughput but without the need for clock synchronization.Besides,the delayaware opportunistic transmission scheduling(DOTS)MAC protocol[30]was implemented to reduce the collision probability as well.But if a node detected any conflict,it would resume its transmission after a random period of time.This incurred energy waste.Similarly,a dynamic channel negotiation MAC mechanism(DCN-MAC)[31]firstly applied channel negotiation by exchanging network information and then scheduled packets transmission based on dynamic single node wake-up.In the handshake-based ordered scheduling(HOSM)protocol[32],packet transmissions were well ordered and scheduled by receivers to reduce the number of control packets.Multiple source nodes were allowed to send data packets after only one reservation,which can effectively improve channel utilization and decrease the energy consumption.Furthermore,the DCO-MAC[33]proposed a hybrid MAC protocol with a receiver-based MAC protocol in heavy traffic load and contention-based MAC protocol in light traffic load.This strategy was implemented to improve the network throughput and decrease the end-to-end delay.The delay and queue aware adaptive scheduling-based medium access control(DQA-MAC)[34]protocol took three actions,adaptive scheduling transmission,reduction of handshaking packets,and concurrent transmission,to improve the network performance.In addition,some protocols[35–37]have been proposed to reduce the handshake overhead and allocate time slots ahead of time.They formulated collisions among nodes as the vertex coloring problem in a spatial-temporal conflict graph,which could highly improve the spatial reuse efficiency.

Table 1.The categories and properties of MAC protocols in UASNs.

In summary,the reservation access protocols can remarkably alleviate packet collisions.However,most protocols would take long time for reservation and thereby consume considerably more energy.

Our proposed DCT protocol combines the advantage of handshaking protocols and scheduling protocols.It effectively takes advantage of long propagation delay and at the same time implements collision-free data concurrent transmission.In addition,with only once handshaking,two or more nodes could transmit data packets concurrently.Therefore network throughput of the proposed protocol will be increased and the energy consumption will be decreased.

Figure 1.The operation of DCT protocol.

III.DATA CONCURRENT TRANSMISSION(DCT)MAC PROTOCOL

In this section,we first discuss the condition of data concurrent transmission to ensure that data packets from different nodes can be transmitted concurrently without packet collisions.Then the structure of DCT protocol is outlined followed by the elaboration of its two phases.

3.1 Condition of Data Concurrent Transmission

As shown in figure 1,the condition of data concurrent transmission is illustrated in DCT protocol.Source node 1 and source node 2 send DATA packets simultaneously at the beginning of time slot.It can be perceived that the propagation delay difference between source node 1 and source node 2 should be at least larger than the transmission time of a DATA packet.Under this condition,their DATA packets could be received one after another at the receiver without collisions.Therefore,the expression of data concurrent transmission condition is yielded,i.e.

wheretiandtjmeans the propagation delay of source nodeiandjrespectively,tDATAstands for the transmission delay of a DATA packet,andtgis the guard time.

When the propagation delays difference between two nodes satisfy the concurrent transmission condition,they can transmit data packets concurrently to the same destination node in the same time slot without collisions.In this way,not only can the propagation delay be effectively utilized to increase network throughput,but also the number of RTS/CTS handshaking packets can be reduced to further save energy.

3.2 Overview of DCT MAC Protocol

In our proposed DCT MAC protocol,the transmission process is divided into two phases.The first phase is to establish a delay table and a concurrent transmission table to meet the requirement of concurrent communications.And the second phase is to conduct collisionfree concurrent transmission based on RTS/CTS handshaking mechanism.The detailed procedures of DCT protocol are shown in figure 1,which will be explained in the following.

It is worth noting that although DCT protocol is a handshaking based MAC protocol that divides time into multiple time slots,it is able to permit different nodes to send data packets concurrently at the same slot without any collisions.To achieve this goal,we should first find the corresponding concurrent transmission nodes for each node based on the propagation delay.Thus,a delay table is established to store the propagation delay between the source node and thedestination node,and then a concurrent transmission table is built for nodes satisfying the concurrent transmission condition based on the information stored in the delay table.

Table 2.An example of delay table and concurrent transmission table.

After the completion of these procedures,data transmission based on RTS/CTS handshaking starts.When the destination node receives RTS packets from different source nodes,it looks up the concurrent transmission table to see whether the other source nodes of RTS packets are in the concurrent transmission table of the earliest arriving source node(also called the handshaking node).If yes,the node will be specified as the concurrent transmission node of the handshaking node.Otherwise,the destination node would select a corresponding concurrent transmission node for the handshaking node from the concurrent transmission table.In the next time slot,the destination node will send CTS packet to inform both the handshaking node and concurrent transmission node to transmit DATA packet simultaneously.After the data packets are received,the destination node will send an acknowledgment(ACK)packet to notice the source node if they are successfully received.

3.3 First Phase of DCT MAC Protocol

3.3.1 Generate Delay Table

In the first phase of the proposed DCT MAC protocol as depicted in figure 1,all nodes in the network send a short frame(SF),which includes its sending time,to the destination node.After receiving the SF,the destination node can calculate the propagation delaytibetween itself and each source node according to the sending timetsiand receiving timetri,i.e.ti=tri-tsi.The MAC address of the source node and its propagation delay is stored in the delay table by the destination node.An example of a delay table with 5 source nodes is presented in Table 2.Since the delay table only needs to be updated when the position of one node is changed,the first phase could be conducted once in a long time duration while the second phase should be implemented continuously.So for a long time duration in the network,the impact of time overhead of the first phase could be ignored.Furthermore,in the first phase,only sensor nodes need to transmit SF packets,and the central nodes do not need to respond with any packets.So the energy consumption is too low to make impact on the total energy consumption of the network.

3.3.2 Generate Concurrent Transmission Table

After getting the delay table,the destination node will compare the propagation delays of all source nodes.If the discrepancy between any two propagation delays is greater than the transmission delay of the data packet,the concurrent transmission condition is satisfied.In this case,the corresponding two nodes can transmit concurrently,and one of them will be stored in the table as the concurrent transmission node of the other.As an example depicted in figure 1,the source node 2 is the concurrent transmission node of the source node 1.The algorithm of generating concurrent transmission table is shown in Algorithm 1.The generated concurrent transmission table is displayed in Table 2.A source node and any of its concurrent transmission nodes stored in the Table 2 can send data packets concurrently in the same time slot.At this point,the first phase of communications is completed.

3.4 Second Phase of DCT MAC Protocol

3.4.1 Send RTS Packets and Select Concurrent Transmission Node

Algorithm 1.Generate concurrent transmission table.1:if the SF packet is for this node then 2:calculate the propagation delay between this node and its source node 3:if this node receives more than one SF packet from different source node then 4: for node i∈the source nodes of the SF packets do 5: for node j∈the source nodes of the SF packets do 6: compare the propagation delays with each other 7: if satisfy the data concurrent transmission condition then 8: the two source nodes are the concurrent transmission node with each other,and store the MAC address of the concurrent transmission node in the concurrent transmission table 9: else 10: the two nodes are not the concurrent transmission node with each other 11: end if 12: end for 13: end for 14:else 15: the node has no concurrent transmission node 16:end if 17:else 18:drop the SF packet 19:end if

When a node has data packets to send,it first sends an RTS packet to the destination node at the beginning of a time slot as shown in the second phase of figure 1.The node whose RTS packet first arriving at the destination node is considered as the handshaking node.In the same time slot,the destination node may receive other RTS packets from the remaining source nodes in the network.It will look up the concurrent transmission table to determine whether the other RTS sending nodes are the concurrent transmission nodes of the handshaking node.If the concurrent transmission nodes do exist,among them the one whose RTS packet arriving earlier would be selected as the corresponding concurrent transmission node of the handshaking node.Otherwise,the destination node will randomly select a concurrent transmission node from the concurrent transmission table,as displayed in the third column of the Table 2(b).The algorithm of selecting concurrent transmission node is given in Algorithm 2.Next,the destination node embeds the MAC address of the concurrent transmission node and handshaking node into the CTS packet and sends it at the beginning of the next time slot.The CTS frame is transmitted as probe to:1)inform the handshaking node to send data packet;2)inform the concurrent transmission node to send data packet;and 3)notify the other nodes in the network to back off for avoiding collisions.

Algorithm 2.Select a concurrent transmission node.1:if the RTS packet arrives earliest then 2:its corresponding source node is the handshaking node and its MAC address is embedded in the CTS packet 3:else 4:for node i∈the source nodes of the RTS packets do 5: if node i is the concurrent transmission node of the handshaking node then 6: embed its MAC address in the CTS packet 7: else 8: if the concurrent transmission table of the source node is not empty then 9: find a concurrent transmission node for this handshaking node from the table and embed its address in the CTS packet 10: else 11: no concurrent transmission node 12: end if 13: end if 14:end for 15:end if

3.4.2 Send Concurrent Transmission DATA Packets

After receiving the CTS packet,the handshaking node sends DATA packet at the beginning of the next time slot.When the other nodes in the network receive this CTS packet,they will first judge whether their own MAC addresses match the MAC address of the concurrent transmission node embedded in the CTS packet.If the MAC address matches,the node will send DATA packet at the beginning of the next time slot.Note that the DATA packet sent by the concurrent transmission node are named as OTHERDATA packet in order to distinguish it from the DATA packet sent by the handshaking node to the same destination node.In this case shown in figure 1,since the propagation delay difference of source node 1 and source node 2 is larger than the DATA packet transmission delay,the OTHERDATA sent by the concurrent transmission node will be received by the destination node before the DATA packet arrives and they would not conflict with each other.However,if the node in the network is neither the destination node of the CTS packet nor the sending node,it will back off until ongoing transmission round is completed,like the source node 3 shown in figure 1.The algorithm of sending concurrent transmission DATA packets is shown in Algorithm 3.

3.4.3 DATA and OTHERDATA Packets Receiving Confirmation

After receiving OTHERDATA and DATA packets,the destination node sends an ACK or NACK packet at the beginning of the next time slot to inform the handshaking node whether the DATA packet has been successfully received.However,the concurrent transmission node cannot discover whether its transmitted OTHERDATA packet is successfully received.Therefore,in this protocol,we embed a flag in the ACK/NACK packet to make a judgment for the OTHERDATA packet.If the OTHERDATA packet has been successfully received,the concurrent transmission node will prepare for the next transmission task.Otherwise,it will keep the OTHERDATA packet and wait for the next retransmission.

Figure 2.Network model.

IV.APPROXIMATE PERFORMANCE ANALYSIS

4.1 Analysis of the Network Throughput

In this section,network throughput of the proposed DCT protocol is elaborated.To derive and evaluate the performance of network throughput,we consider a specific sensor network withNcentral networks in a set ofC={C1,...Cn,...,CN}as depicted in figure 2.Note that the DCT protocol could be applied not only in central networks,here we just take this type of network as an example.In this network,each central networkCncontains a central nodecn,Mnsensor nodesMn={c1n,...,cMnn}andIninterference nodesIn={c1n,...,cInn}.Sensor nodes collect data and then transmit data tocn,while interference nodes are in the communication range ofcnbut their central nodes are notcn.In the network model,we make some assumptions.First,the communication distance of all nodes is limited withind0.Second,the distance between any two central nodes is assumed to be larger than the communication range limitd0,which is reasonable in large-scale networks.In addition,the sensor nodes are randomly deployed in a circle according to the uniform distribution.The center of the circle is assumed to be nodecnand the radius isd0.Finally,the generation of data packets follows a Poisson distribution with averageλpackets per second.

According to the flow model[38],the average channel utilization could be calculated as the ratio of data transmission time and total transmission time,which is

Figure 3.Markov chain model for state of data transmission.

To calculate above three time durations respectively,we define three data transmission statesSincluding the fail statef,the success states,and the idle statei.For any data transmission round,its current data transmission state only depends on its previous data transmission state,thus the process could be defined as a Markov chain depicted in figure 3.According to the property of Poisson distribution,we assume that the probability that one node has no data packet to transmit in a time durationtis

Then in a central network containingMnsensor nodes,the number of nodes with no data packets to send follows binomial distribution,i.e.X～B(Mn,p(t)).Thus,the probability thatknodes have no data packets to send is

In the idle state,all theMnnodes in one central network have no packets to send during the time duration of the last state.Thus,the probability of idle state in a time duration could be calculated as

wheretimeans the time duration of the last stateS∈{i,s,f}.

Specifically,the idle time duration contains only an idle time slot to wait for possible RTS packets,which is

whereTis the time slot length,i.e.T=tCTS+d0/v+tg,tCTSis the CTS packet transmission time,vis the sound wave propagation speed,andtgis the guard time.Note that since the first phase of the protocol is only implemented when the position of one node is changed,so it only needs to be conducted once in a long time duration.In addition,in the first phase,only sensor nodes need to transmit SF packets,and the central nodes do not need to response with any packets.The time duration for one node to transmit a SF packet istSF+d0/v,which is roughly as long as a time slotT.The more the number of nodes is,the longer the time duration of the first phase is.For the worst case,all nodes send SF packets at different time slots.Then the total time duration of the first phase could be roughly calculated byMn(tSF+d0/v)=MnT,which is the same as several time duration of the time slot.However,for a long time duration in the network,the first phase is only implemented once after amount of rounds of the second phase.As a result,the time duration of the first phase could be roughly ignored in the whole communication process.

In the success state,the second phase of the protocol could be processed successfully if there are packets to be sent during the time duration of the last state as well as RTS packets could be transmitted successfully.Then two cases would happen.Case(i)is that only one node transmits RTS packet and case(ii)is that two or more concurrent nodes transmit RTS packets.Then the successful probability of one central network containingMnsensor nodes andIninterference nodes could be expressed as

wheretsis the time duration of the last state,Pcon(d)is the concurrent transmission probability of RTS packets,dis the transmission time of RTS packets,i.e.d=tRTSv,PIis the concurrent transmission probability of RTS packets for interference nodes,i.e.

Lemma 1.The probability of concurrent transmission for two nodes randomly deployed in a circle could be given by

where d is the distance that satisfies the concurrent condition,specifically here d is the distance between two nodes that achieve RTS packets concurrent transmission,i.e.d=tRTSv;d0is the maximum communication distance as well as the radius of the circle.

Proof.Please refer to Appendix A.

Then the success time duration contains two time slots to transmit RTS packet and CTS packet,and some time slots to transmit DATA packet and ACK packet,which could be calculated as

whereTdatais the successful data packets transmission time containing the time from the start of a DATA packet to the successful reception of the ACK packet,soTdatacan be written as

wheretdatais the number of slots occupied by the transmission of a DATA packet and could be calculated bytdata=「(tDATA+d0/v),tDATAis the DATA packet transmission time delay,tACKis the ACK packet transmission time slot,i.e.tACK=T,and thePeis the packet error rate withLbits per packet and bit error rateBER,and it can be calculated asPe=1-(1-BER)L≈L×BER[23][39].

In the fail state,the central node fails to receive the RTS packet and needs to wait a RTS packet and a CTS packet transmission time slot before the new protocol process starts.The probability of data transmission in fail state is given by

Then the fail time duration contains two time slots for the RTS packet transmission and the CTS packet transmission,which could be expressed as

Based on the above formulation and the Markov chain depicted in figure 3,the state transition probability of the Markov chain could be expressed as

According to(13)-(21),the steady-state probability matrix of the Markov chain could be calculated by(22)-(23),wherePi,Ps,andPfis the steady-state probability of idle state,success state,and fail state,respectively.

Based on above,the busy time duration is the total time when data transmission state is in success state or fail state,which is

Similarly,the idle time duration is

In addition,the time duration for DATA packet transmission and OTHERDATA packet transmission could be expressed as

Therefore,substituting(24),(25),and(26)into(2),the average channel utilization can be calculated as:

Based on the above,the theoretical network throughput of DCT protocol,which means the number of bits of data packets that are successfully received per second,is given by

whereRis the bit rate.From this equation,we can find that the theoretical network throughput of DCT protocol is determined byλ,N,Mn,d0,BER,L,R,tDATA,tCTS,andtRTS.

4.2 Analysis of the Computational Complexity

The computational complexity of the proposed DCT protocol consists of the generation of the delay table and the concurrent transmission table,the selection of the concurrent transmission node,and sending SF/RTS/CTS/DATA/ACK packets.Then in the proposed network model shown in figure 2,the worst case of computational complexity could be calculated as Θ(Nmax(Mn)+Nmax(M2n)/2+Nmax(Mn)+cN),i.e.the computational complexity isO(Nmax(M2n)).For the Slotted-FAMA protocol,the computational complexity isO(1).For the DOTS protocol,the computational complexity isO(Nmax(M2n))since the transmission scheduling process in the DOTS protocol needs to calculate the interference among nodes based on the delay map.Therefore,the computational complexity of the proposed DCT protocol is the same as the DOTS protocol and worse than the Slotted-FAMA protocol.

V.PERFORMANCE EVALUATION

In this section,the performance of the proposed DCT protocol is evaluated and demonstrated in Aqua-sim simulator,an NS2 based simulator for UASNs[40].The simulation results of DCT protocol are compared with those of the representative MAC protocols,i.e.Slotted-FAMA protocol[23]and DOTS protocol[30].

5.1 Simulation Setup

In the simulations,nodes are randomly distributed in a square area of 3 km by 3 km.Each node generates data packets following a Poisson process with the same rateλ.For comparison,λvaries from 0.01 pkt/s to 0.1 pkt/s with step 0.01 pkt/s and the size of the DATA packetpDATAis assumed to be the same,i.e.pDATA=tDATAR.Note that the offered traffic is defined as the number of packets generated by each node per second.Moreover,power settings are 1 W for transmission,0.75 W for reception,and 0.008 W in idle state[41].The sound wave propagates at the speed of 1500 m/s.The bit rateRis 10 kbps.

The following performances are evaluated with different number of nodes and different sizes of DATA packets:

1.Throughput,denoted as the number of bits successfully received per second,i.e.

wherenDATAis the number of received DATA packets,Ttotalis the total simulation time.

2.Energy consumption,denoted as the average energy consumption of transmitting a DATA packet,is calculated by the total energy consumption divided by the total number of received DATA packets,i.e.

whereEtotalis the total consumed energy including the transmission energy,the reception energy,and the idle energy.Etis the transmission energy.It can be calculated by the transmission power times the transmission time,which is related to the number of transmitted packets,containing DATA packets,RTS packets,CTS packets,OTHERDATA packets,and ACK packets.Eris the reception energy.It can be given by the receiving power times the receiving time,which is related to the number of received packets.Eiis the idle energy associated with the idle time when there is no packet to be transmitted or be received.

3.Average end-to-end delay,denoted as average time duration from the time the packet is generated to the time it is successfully received,i.e.

whereDtotalis the total end-to-end delay,including the waiting delayDw,the transmission delayDt,and the propagation delayDp.

Figure 4.Network throughput with different traffic load and different size of data packet.

Figure 5.Network throughput with different number of nodes.

5.2 Simulation Results

5.2.1 Throughput

Figure 4 and figure 5 show network throughput with different size of DATA packets and different number of nodes.The number of nodes in the network varies from 10 to 60 with 50 bytes per DATA packet.For the DATA packet size,both 50 bytes and 100 bytes per DATA packet are simulated with 6 nodes in the network.In general,the throughput of DCT protocol is higher than that of Slotted-FAMA protocol and DOTS protocol with the same offered traffic.The throughputs of all protocols become stable after climbing up to given thresholds.

Figure 6.Theoretical and simulated normalized network throughput of DCT protocol.

Simulation results of data packet with different size are shown in figure 4.When the number of nodes is fixed,and the data packet sizes are either 50 bytes or 100 bytes,DCT protocol outperforms both Slotted-FAMA protocol and DOTS protocol.Figure 4 shows that the up-bound throughput of DCT protocol is higher than that of DOTS protocol by 26.40% with 50 bytes per packet and 9.18% with 100 bytes per packet while it exceeds that of Slotted-FAMA protocol by 38.40%and 15.00%respectively.These results demonstrate explicitly that the bigger the size of data packets is,the smaller the probability for nodes to satisfy the concurrent transmission condition becomes.Thereby,the performance gain of the network with 50 bytes per data packet is larger than that of the network with 100 bytes.Furthermore,for the same protocol in figure 4,when the data packet size changes from 50 bytes to 100 bytes,the up-bound throughput can increase by 34.94%,56.22%,62.41% for DCT,DOTS,and Slotted-FAMA protocols,respectively.Apparently,the throughput of DCT protocol increases the least.This is because the packet size could influence the probability of concurrent transmission in DCT protocol,which is in accordance with the protocol’s setting in(1).

In terms of the simulation results with different number of nodes shown in figure 5,when the number of nodes varies from 10 to 60 with fixed size of data packets,the throughput of DCT protocol is better than that of Slotted-FAMA protocol and DOTS protocol as well.Specifically,when the number of nodes in the network gets bigger,the throughput would change little.This is because whenλis set as 0.1 pkt/s,the network throughput has reached the upbound.So when the number of nodes increases,the throughput would not be improved.

Figure 7.Average energy consumption with different size of data packet.

In addition,based on figure 6,the theoretical throughput of the proposed DCT protocol agrees well with the trend of simulation results with different size of data packets.Besides,when the traffic load is low,the throughput of simulation results is a little bit higher than that of theoretical results.This is because when the traffic load is low,the collision probability among nodes in simulation tests may be lower than the statistical results leading to the higher throughput of simulation results.

In conclusion,compared with Slotted-FAMA protocol and DOTS protocol,the proposed DCT protocol has better throughput performance.The corresponding performance gain gets larger when data packet size becomes smaller or network becomes bigger.Thus,it is suitable to be applied in UASNs with large-scale and high data traffic load.

5.2.2 Average Energy Consumption

From the perspective of average energy consumption of each data packet,DCT protocol exhibits better performance than Slotted-FAMA protocol and DOTS protocol as well.Energy consumption is associated with the number of received data packets and total energy.In the simulation,we have set long enough time to run out the battery energy,so the average energy consumption of these protocols performs similarly with the throughput.Figure 7 shows that performance of average energy consumption with 50 bytes per packet and 100 bytes per packet with 6 nodes in the network.Figure 8 shows the performance of average energy consumption with 10 to 60 nodes in the network,respectively.The result indicates that the average energy consumption of DCT protocol is less than that of Slotted-FAMA protocol and DOTS protocol in these different scenarios.

Figure 8.Average energy consumption with different number of nodes.

Overall,at the same offered traffic load,figure 7 shows that the energy consumption becomes higher when the size of data packet increases from 50 bytes to 100 bytes.Furthermore,the energy consumption of DCT protocol is less than that of Slotted-FAMA protocol and DOTS protocol with these two different data packet sizes.The reason is that DCT protocol requires less control packets and thereby consumes less energy.

In terms of different number of nodes in the network,the energy consumption increases obviously.With the network getting larger,although more nodes send the RTS packets,only limited nodes can transmit data packets.Thus,the energy to send and receive RTS packets is wasted.

In conclusion,the simulation results demonstrate the superiority of DCT protocol in terms of energy consumption performance in various scenarios.Thus,the DCT protocol could be applied for the long term monitoring.

5.2.3 Average End-to-End Delay

Figure 9.Average end-to-end delay with different size of data packet.

From the results shown in figure 9,we can see that the average end-to-end delay of DCT is longer than that of the Slotted-FAMA protocol and the DOTS protocol when the offered traffic load is low.However,when the offered traffic load is higher,the DCT exhibits a shorter average end-to-end delay.This is because when the offered traffic load is low,unlike the Slotted-FAMA protocol and the DOTS protocol not requiring the first phase communications,the DCT protocol needs to perform the first phase communications to establish the delay table and the concurrent transmission table.When the offered traffic load increases,for the DCT protocol,the number of transmitted data packets during the same time is much more than that of both the Slotted-FAMA protocol and the DOTS protocol,thereby the average end-to-end delay would be reduced.From this figure,it can be also found that when the offered traffic load is large,the average endto-end delay of the DCT protocol is close to that of the DOTS protocol,and is much smaller than that of the Slotted-FAMA protocol.This is because with the network traffic load getting large,the nodes need to wait more time to have the chance to send data packets,though they may have attempted to send the RTS packets.

In terms of different data packet size,the end-toend delay of all these protocols changes only a little when the size of data packets increases from 50 bytes to 100 bytes.This is because when the size of data packets changes,only the transmission delay and waiting delay would change.The transmission delay would only changes from 0.005s to 0.01s.The waiting delay would change because the number of sending data packets would decrease due to sending less concurrent data packets when the size of data packets increases from 50 bytes to 100 bytes.However,these delay changes are all very short compared with the long propagation delay and long time slots.As a result,the end-to-end delay changes only a little in a long simulation time.In addition,when the offered traffic changes from 0.01 pkts/s to 0.03 pkts/s,the end-to-end delay changes little.This is because when the offered traffic is low,the nodes have little packets to be transmitted and at the same time packets generated in the nodes would be transmitted quickly.As a result,the end-toend delay is very short and changes little.For the DCT protocol,the end-to-end delay is long when the traffic load is low.This is because in the DCT protocol,nodes have to perform the first phase communications to establish the delay table and the concurrent transmission table.Packets generated in the nodes could be transmitted after the end of the first phase.As a result,the end-to-end delay is long although there are little packets to be transmitted.

VI.FUTURE WORK

The development of MAC protocol is closely related to oceanic exploration applications and it caters to the demands for the quality of service(QoS).We take four typical applications with different demands on network range and data traffic load as examples:marine ranching,disaster forecasting and warning,equipment monitoring and control,and Internet of underwater things,shown in figure 10.

1)Marine ranching.In recent years,the emerging application of UASNs to marine ranching has become an important trend.Thousands of sensors are laid in the farming area or carried by living creatures.The application range is usually not too far and the data traffic load is heavy.To reduce the cost of data collection and prolong the life time of nodes,marine ranching requires low energy consumption.Thus,MAC protocols have to be energy efficient and in the meantime can achieve high throughput.

2)Disaster forecasting and warning.A lot of cabled sea floor observatories have been applied in disaster forecasting and warning.Wireless communication and networks are also indispensable to support observatory demands on large scale area.Observation networks are generally large while nodes only need to transmit data periodically.Thus,MAC protocols have to orchestrate data transmission time and collaborate channel access of nodes with various distances and data traffic loads.

Figure 10.Future trends and typical applications for UASNs.

3)Equipment monitoring and control.The development of underwater vehicles encourages the promotion of communications between mobile nodes.To control the equipment,networks are required to have low latency.The MAC protocols are expected to achieve real-time information feedback and support the access of mobile nodes.

4)Internet of underwater things.In the future,the Internet of underwater things is envisioned to be in reality.At that time,smart underwater things are all combined and aggregated together to collect information of the oceanic world.The network range and data traffic load would be unprecedentedly large.Great efforts should be taken to tackle the challenges in applications.Thus,sophisticated MAC protocols should be designed to meet the demands on frequent data transmission and the access of large number of nodes.

VII.CONCLUSION

In this paper,we propose a MAC protocol called DCT protocol to enhance the network throughput and reduce the energy consumption in UASNs.The proposed protocol first establishes a delay table and a concurrent transmission table to find two concurrent transmission nodes,and then starts to transmit data concurrently.On the basis of the concurrent transmission condition,the packet transmission will not collide at the destination node.Performance analyses are given in detail,and the concurrent transmission probability is derived.The simulation results further verify the performance of DCT protocol.DCT protocol mainly provides a possible solution to cope with the challenges of spatial unfairness,data eruption,and low energy efficiency.There are still some constrains in DCT protocol.This protocol has not considered about the fast mobility of nodes.In addition,it is designed for star networks,and multi-hop networks have not been considered so far.As a future work,we could also broaden the network coverage by deploying multiple star networks.

APPENDIX

Proof of Lemma 1

To prove Lemma 1,we define two sensor nodes randomly and independently distributed following the uniform distribution in a circle with the central node as the center andd0as the radius.Then in the polar coordinate,the angle is uniformly distributed everywhere in a circle and the probability density function of the radiusrcould be expressed as

The distance between the central node and the sensor node is exactly the radius of the sensor node in the polar coordinate.So the distance differencerdbetween two sensor nodes and the central node isrd=|r1-r2|.Then based on(32),the probability distribution function ofrdcould be given by

To realize the concurrent transmission,the difference in the distance between the sensor nodes and the central nodes should larger than the length of RTS packets or DATA packets,which is

wheredisdRTS=tRTSvordDATA=tDATAvfor the concurrent transmission of RTS packets or DATA packets respectively.

ACKNOWLEDGEMENT

This work was supported in part by the National Natural Science Foundation of China under Grant 62171405 and in part by the National Science Fund for Distinguished Young Scholars under Grant 62225114.