Impact of Real Modem Characteristics on Practical Underwater MAC Design Lina Pu, Yu Luo, Yibo Zhu, Zheng Peng, Jun-Hong Cui Computer Science & Engineering Department University of Connecticut, Storrs, CT 06029 Email: {lina.pu, yu.luo, yibo.zhu, zhengpeng, jcui}@engr.uconn.edu

Shruti Khare, Lei Wang

Benyuan Liu

Electrical & Computer Engineering University of Connecticut Storrs, CT 06029 Email: {shruti.khare, leiwang}@engr.uconn.edu

Department of Computer Science University of Massachusetts Lowell, MA 01854 Email: [email protected]

Abstract—The grand challenges posed by the adverse acoustic channels have been extensively studied in medium access control (MAC) design for underwater acoustic networks (UWANs). In recent years a bunch of MAC protocols have been proposed and analyzed to address the long propagation delay and power constraint problems. However, the challenges from the real acoustic modem characteristics are not well investigated in the current research. Based on our experiments and sea tests with real acoustic modems - Teledyne Benthos modem, WHOI MicroModem and high speed UConn OFDM modem, we discover two important real system issues and discuss their significant effects on MAC design in UWANs. Experiment and simulation results reveal that the long preamble of acoustic modems harm most existing MAC protocols with short packets, such as acknowledgment, notification and reservation. We also find that reservation based methods are vulnerable to the long transmission time and busy terminal problem, while the random access based protocols can benefit from them. As a result, Aloha based protocols yield encouraging throughput performance in UWANs and are reconsidered as viable solution for underwater MAC. Based on the analysis, simulation and experiment results, we also provide discussion on the direction of future MAC designs.

I. I NTRODUCTION In the past several years underwater acoustic networks (UWANs) have gained significant attention from academic research [1], [2]. UWANs enable a wide range of aquatic applications, such as scientific exploration, commercial exploitation as well as coastline protection. The grand challenges faced in UWANs have also drastically attracted increasing endeavors from the research community. In underwater environments, radio signals suffer from significant attenuation and thus cannot propagate far. Alternatively, acoustic communication becomes the viable solution. However, this different communication method poses great challenges to underwater wireless networks. Compared with the radio signals, acoustic signals propagate much slower, with a speed about 1.5 × 103 m/s in water, five orders of magnitude lower than the radio (3 × 108 m/s). Therefore, this introduces long propagation delay issues in UWANs. Further, the bandwidth capacity of underwater acoustic channels is very limited and heavily depends on both the transmission range and the frequency [3]. According to [4], nearly no research or commercial modem can exceed 40 km×kbps as the maximum

attainable range×rate product. The adverse underwater environments pose grand challenges to underwater medium access control (MAC) protocol design for effective collision avoidance. Although the long propagation delay and the limited available communication bandwidth have been fully investigated in the recent literature on underwater water MAC protocol design [5]–[8]. Some problems caused by the characteristics of real acoustic modems, such as the long preamble sequence and data transmission time, are still overlooked in the protocols design and performance analysis. When taking these factors into account, the performance of most existing MAC protocols will degrade significantly. In this paper we first introduce several real acoustic modem characteristics revealed in our experiments and then perform both simulation and study the impact of these system features on the performance of selected MAC protocols with experiment and simulation results. The remainder of this paper is organized as follows. In Section II, we introduce the related work. In Section III, we first describe and analyze several unique real acoustic modem features. After that, in Section IV we study their impact on the performance of existing MAC protocols through experiment and simulation results. Section V provides possible directions for future underwater MAC protocol design in real systems. And finally our conclusions are drawn in Section VI. II. R ELATED W ORK In this section, we briefly review some existing underwater MAC protocols which can be classified into three categories: random access, handshaking and schedule based MAC protocols. Different categories of MAC protocols are affected by the modem features in different ways, which is going to be discussed in Section IV. A. Random Access MAC Protocols Since being proposed in 1970, Aloha [9] has been widely used in early satellite networks, which experience high latency due to the long propagation delay, similar to the non-negligible propagation delay characteristic in UWANs. However, limited bandwidth and energy capacity in acoustic modems pose new challenges to underwater MAC design. UWALOHA [10] takes

B. Handshaking based MAC Protocols In order to improve the energy efficiency in UWANs, the handshaking based protocols have been proposed by using small size of control packets to avoid collision among data packets. A typical handshaking MAC is Slotted FAMA (SFAMA) [11], using Request-to-Send (RTS) and Clear-to-Send (CTS) to reserve time slots for data packet transmission. However, the high latency harms channel utilization in SFAMA. Distance-Aware Collision Avoidance Protocol (DACAP) [6] allows the sending of CTS to be fast in order to reduce the waiting time at a sender node which can increase the throughput. Adaptive propagation-delay-tolerant collisionavoidance protocol (APCAP) [8] and Reservation-based MAC protocol (R-MAC) [12] utilize the interval time between handshaking signals to process other packets such that the channel efficiency is improved. COPE-MAC [13] further improve the efficiency by the introduction of parallel reservation and cyber carrier sensing. Noh et al. also proposed the Delay-aware Opportunistic Transmission Scheduling (DOTS) method [14] which uses topology information and handshaking to improve the throughput performance. However, all these existing handshaking based MAC protocols have overlooked the high collision probability among control packets caused by long preamble sequence in real acoustic modem which we will discuss later. C. Scheduling based MAC Protocols Scheduling based MAC protocols refer to the kind of MAC protocols with reassignment of time and/or frequency resources to sensor nodes in a network. Fixed TDMA, FDMA or CDMA have drawbacks of poor performance in terms of throughput and latency. In order to solve the spatial-temporal uncertainty problem [15] in scheduling transmission, a couple of effective spatial or time reuse MAC protocols have been proposed. UWAN-MAC [16] uses local synchronization to improve energy efficiency, while being limited to be only applicable to applications with ultra low traffic rates. Moreover, scheduling based MAC protocols are vulnerable to the long transmission time and busy terminal problem since the modem may not be able to send out a packet at the scheduled time when it is busy with other transmissions or receptions.

1

0.5 Signal Strength

advantage of the acknowledgment and backoff scheme to reduce the actual transmitted traffic rate, and thus lowers the collision probability. In ALOHA Collision Avoidance (ALOHA-CA) [5], a sensor node extracts packet destination information from the overheard packets to avoid collision in the single hop networks. Syed et al. [7] proposed TLohi employing a tone-based contention mechanism to detect collisions and help achieving good throughput. Even though random access MAC protocols are limited by their ability to avoid collisions and suffer from poor energy efficiency performance, Aloha based protocols are still widely employed in real networks, due to the protocol scalability and the low complexity of the control algorithm.

0

−0.5

−1

Fig. 1.

0

0.2

0.4 Time [second]

0.6

0.8

Preamble significantly elongates the transmission time of a packet

The details of this effect of acoustic modem features will be illustrated in the next section. III. C HARACTERISTICS OF ACOUSTIC M ODEMS Based on our experience from real system implementation and sea tests with real acoustic modems - Teledyne Benthos modem [17], WHOI Micro-Modem [18] and high speed UConn OFDM modem [19], we will describe two important practical issues that affect the MAC design in this section. A. Long Preamble in Acoustic Modems In acoustic modems, for the purpose of synchronization, signal detection, automatic gain control (AGC) control and channel estimation, a preamble must be designed as an initial part of a packet [17], [19], [20]. Take the Benthos modem [17] as an example. Even a packet with only several bytes useful information requires more than 1.5 seconds to transmit, shown in Table I, since any packet generated by the Benthos modem will have 1.5 seconds preamble sequence. This implies that “short” control packets are not short anymore and will equally suffer heavy collision as data packet. Another example is Aqua-fModem [19], the high speed OFDM modem developed by the UConn UWSN lab. In the Aqua-fModem the packet consists of two preamble blocks and a number of OFDM data blocks, and all data blocks have the same fixed length. No matter how few data bytes need to be transmitted, the packet is always padded to meet the minimum length of a data block. In this way, the minimum packet transmission time in Aqua-fModem is the transmission time of two preamble blocks plus a data block, whose waveform is illustrated in Fig. 1. As shown in Fig. 1, the preamble takes first two blocks for packet detection and synchronization, and the third block is the actual data part. From Fig. 1, we can observe that the minimum packet transmission time in AquafModem is about 0.66 seconds. Almost all existing MAC protocols suffer from the long preamble feature of actual acoustic modems, especially for protocols with small control packets. The impact is verified in Section IV-A with simulation and experiment results.

TABLE I T OTAL C ONTROL PACKET D URATION WITH D IFFERENT ACOUSTIC M ODEM Modem Type

Data Rate

Preamble Duration Time

Total Control Packet Duration Time ≈ 1.56s

800bps (Standard) ≈ 1.5s

Teledyne Benthos Modem

≈ 1.52s

2.4Kbps (Highest) UConn OFDM Modem

3.045Kbps

0.49s

0.66s

80bps (Standard) WHOI Micro Modem

1.47s 0.87s

300-5000bps (High PSK mode)

1.52s

B. Long Transmission Time

IV. I MPACT OF M ODEM FEATURES ON U NDERWATER MAC In Section III, we have discussed two unique characteristics of acoustic modems, the long preamble and the long transmission time. In this section, we will analyze the impact of these two features on the performance of existing MAC protocols.

Simulation without preamblem Simulation with preamblem

0.6 Normalized throughput

The high latency problem due to long propagation delays has been most concerned in recent underwater MAC protocol design. APCAP, a protocol utilizing the interval propagation time between handshaking signals for data packet deliveries, is designed for underwater networks when the propagation delay dominates the data packet transmission time. Simulation results in [8] have shown the effectiveness of APCAP on throughput improvement for propagation delay tolerant networks. However, in a real system the transmission time of data packet is typically longer than the propagation delay, especially for low speed acoustic modems. The long data packet transmission time degrades the functionality of MAC layer pipelining in protocols like APCAP. Busy terminal phenomenon is another problem occurs in acoustic modems which is caused by the half duplex feature and is aggravated by long transmission time. When the modem is overhearing packet, no matter whether the node is the intended receiver or not, it is busy processing the data and therefore cannot process other packets - this is called busy terminal phenomenon. This affects both incoming packets and outgoing packets. In radio networks, the impact of this phenomenon is negligible due to the short transmission time. However in UWANs, because of the long preamble and low data rate of acoustic modems, busy terminal will affect on underwater MAC performance significantly. In order to evaluate the impact of transmission time and propagation delay on MAC protocol perfromance, we define the transmission time duty cycle (TDC) coefficient as α = Ttrans /(Ttrans + Tprop ). TDC is an important factor that affects the channel utilization of underwater MAC protocols as well as collision avoidance probability in Aloha-like approaches. Assuming the sound speed in underwater is 1500 m/s, in Table II we list the data packet transmission time and TDC factor with different modems when the data size is 400 byte and node spacing is 1 km. In Section IV-B we will discuss the effect of TDC factor in detail.

0.5 0.4 0.3 0.2 0.1 0

0

0.02

0.04 0.06 Traffic rate

0.08

Fig. 2. Throughput of S-FAMA affected by preamble with Teledyne Benthos Modem Settings

A. Impact of Long Preamble Most handshaking and reservation based underwater MAC protocols need to send a small notification (NTF) or RTS packet for data transmission. RTS/NTF packets are generally with small size, e.g. several Bytes in DACAP [6] as well as ALOHA-AN [5]. In this case, the competitions among control packets are assumed to have a low collision probability, which is beneficial to achieve a collision free data transmission. However, this is under the assumption that smaller packets have small transmission time. So they have low collision probability. In real systems, however, this is not true in real underwater acoustic networks. With real acoustic modems we tested and evaluated, the assumption of short control packet (RTS/NTF) can hardly be satisfied in practical application due to the long preamble sequence. In Table I, we have listed the preamble length, transmission rate, as well as the total duration time of control packet with 6 bytes of useful information for three different acoustic modems. From Table I we can see, when taking the preamble sequence of acoustic modems into account, the duration time of the control packets increases significantly and the collision among these “short” packets can not be neglected any more, which implies the increased collision probability among these packets. In Fig. 2, the simulation results illustrate the negative effect of the long preamble on the throughput of S-FAMA, a typical handshaking based protocols. The high collision probability among RTS packets significantly reduces the successful reservation ratio of S-FAMA and degrades the throughput

TABLE II DATA PACKET L ENGTH AND TDC FACTOR IN D IFFERENT ACOUSTIC M ODEMS Modem Type

Data Rate

Total Packet Duration time

TDC Factor (α)

800bps (Standard)

≈ 5.5s

≈ 0.89

2.4Kbps (Highest)

≈ 2.83s

≈ 0.81

3.045Kbps

1.53s

0.70

Benthos ATM-88X Modem UConn OFDM Modem

80bps (Standard)

40.87s

0.98

300-5000bps (High PSK mode)

1.5-11.54s

0.70-0.95

WHOI Micro Modem

To summarize, the explicit control packets in underwater MAC introduce higher overheads than previously expected in underwater networks and significantly degrades the performance of most existing protocols.

0.5 simulation without preamble simulation with preamble experiment result

Normalized throughput

0.4

0.3

B. Impact of TDC coefficient

0.2

0.1

0

Fig. 3. Modem

0

0.05

0.1

0.15 0.2 Traffic rate

0.25

0.3

Throughput of Aloha affected by preamble in Teledyne Benthos

performance in real underwater system. S-FAMA suffers a 42% performance degradation observed in Fig. 2. Note that the slotted scheme in S-FAMA has the capability to mitigate the collision among RTS packets. On the other hand, the unslotted handshaking protocols, such as DACAP and APCAP, are more vulnerable to the long preamble sequence in real systems. In underwater MAC design, feedback/acknowledgement (ACK) is highly desirable for reliable network communications. Due to the constrained power supply and long propagation delays in UWANs, the end-to-end retransmission on the transport layer is costly. The hop-by-hop reliable data delivery becomes more essential than wireless networks. However, long preamble sequence on small ACK packets cause significant extra traffic load to the whole network. Collisions among data and ACK packets degrade the network performance in a remarkable way. The simulation and experiment results in Fig. 3 illustrate the impact of preamble on throughput performance of Aloha with ACK in underwater networks, which we call UWALOHA [10]. UWALOHA is employed with explicit acknowledgement to improve reliability performance. With the 1.5 second preamble sequence in Benthos modems, the throughput of UWALOHA has a 35% drop according to the simulation result without long preamble shown in Fig. 3. The consistency between simulation and experiment results justifies our analysis about the impact of long preamble on underwater MAC protocols. The details of experiment and simulation settings will be described in Section IV-C.

In scheduling based underwater MAC protocols, such as UWAN MAC and APCAP protocols, the packet transmission time is pre-decided and announced to neighbors for collisions avoidance. However, due to the busy terminal problem as well as long packet length, the unexpected arrival packets may block the pre-scheduled transmission scheme of these protocols, and therefore the performance of this kind of MAC may be degraded. When the packets fail to be sent out at the scheduled time, collisions appear and may even explode. Beside the increased collisions among control packets brought on by long preamble sequence problem, the busy terminal phenomenon further degrades the performance of reservation based protocols. Contrary to popular belief, random access based protocols performs very well in real systems because they can benefit from the busy terminal phenomenon. This is due to the fact that the actual traffic rate pushed into the channel is reduced by the busy terminal feature of modems. Collisions are essentially decreased and accordingly throughput and energy efficiency performance are improved. Moreover, when the packet transmission time dominates the propagation delay, the carrier sensing mechanism enabled by the busy terminal phenomenon further improves the performance of random access based protocols. Due to the benefit from busy terminal phenomenon, the collision window reduces from double data transmission time in Fig. 4 (a) to two times propagation delay in Fig. 4 (b) when the transmission time dominates. In this case, the advantage of random access protocols raises up for underwater networks. Fig. 5 compares the normalized throughput of UWALOHA and S-FAMA with different TDC coefficient. Benefitting from carrier sensing to reduce collision as well as increased channel utilization with large TDC, the normalized throughput of both UWALOHA and S-FAMA is improved a lot with the increased TDC coefficient. Compared with UWALOHA, the S-FAMA is more vulnerable to the high latency of underwater networks. Similar conclusion applies to other handshaking based methods. Nevertheless, the pipeline mechanism in APCAP MAC protocol starts to take effect when Ttrans < 2Tprop , i.e.,

A

DATA

DATA

Collision

B

DATA

Collision Window

Collision Window

C

Collision DATA

DATA

DATA

t

Collision Window

B

DATA

DATA

DATA

t

C

DATA

Collision t

Collision Window

DATA

DATA

t

t

(a)

FAMA with α=0.89 FAMA with α=0.70

0.5

t

(b)

Normalized throughput

A

0.4 0.3 0.2 0.1

Fig. 4. Collision window of random access protocols. (a) Propagation delay dominates. (b) Transmission time dominates.

0

0

Fig. 6.

0.2

0.4

0.6 0.8 Traffic rate

1

1.2

Throughput of FAMA affected by α

Normalized throughput

0.25 0.2 0.15 UWALOHA with α=0.70 UWALOHA with α=0.89 S−FAMA with α=0.70 S−FAMA with α=0.89

0.1 0.05 0 0

Fig. 5.

0.2

0.4

0.6 0.8 Traffic rate

1

1.2

Throughput of UWALOHA and S-FAMA affected by α

α < 0.67. However, from Table II we can see, the TDC coefficient α is generally greater than 0.7 even for the network with high speed OFDM modems. Other than handshaking based protocols losing effectiveness in underwater system, the random based method reveals encouraging throughput performance. Leveraging the enabled carrier sensing, the un-slotted handshaking method may outperform the slotted one under high TDC coefficient α. As illustrated in Fig. 4, the collision window for data packets reduces significantly with dominating transmission time, which is generally true for most underwater networks. Fig. 6 reveals the superior throughput performance of FAMA [21] than S-FAMA in Fig. 5 with Benthos modems. However, this advantage disappears under smaller α with the high speed OFDM modems. Under the impact from both the transmission time and propagation delays, the effectiveness of slot mechanism has to be reexamined in the future underwater MAC protocol design. C. Experiment and Simulation Settings In order to verify the impact of the real acoustic modem features on MAC protocols, we have tested the protocols in both field experiments and simulations. The results have been studied and evaluated in Section IV. In this section, we will introduce the experiment and simulation settings. The field test was conducted at Chesapeake Bay, Maryland, in May 2011. Our underwater network nodes were equipped with Benthos modems, which are known for their reliability and robustness as commercial products. The acoustic modems

have the same characteristics as we described in previous sections, namely long preamble sequence and busy terminal problem. The Benthos modems have up to several kilometers of communication range and about 1.5 seconds preamble. In the sea tests, eight nodes are deployed in a ring topology with about 1 km average distance between neighbor nodes, as shown in Fig. 7. The transmission rate of modem is 800 bps for the purpose of reliable transfer and the data packet size is 400 byte. The experiment results of UWALOHA are discussed in Section IV-A. To better understand the results from the field test, we later did simulations with the same parameters and settings as those in the Chesapeake Bay experiment. The adopted simulation tool is Aqua-Sim, a NS2 based network simulator. From Fig. 3 we can see that, when the practical issues are taken into account, the simulation outcome matches the experiment result well. This consistence justifies our analysis on the impact of modems features on MAC protocols. S-FAMA requires time synchronization which is costly in real systems. Thus in this paper, we only provide the simulation results of S-FAMA in Section IV-B and assume perfect time synchronization. All other simulation settings are the same as experiments. Both low speed Benthos modem and high speed OFDM modem are evaluated in the simulations. The length of preamble in the two modems are 1.5 and 0.49 seconds respectively. The TDC coefficient α in our simulations varies from 0.7 to 0.89, tested different situations with variant data packet duration times. V. R EAL S YSTEM F EATURES C ALL FOR I NNOVATIVE S OLUTIONS In real underwater acoustic modems, because of the long preamble sequence, the NTF/RTS/ACK type of control packets are not “short” anymore to alleviate collisions in channel competition. And the interferences among control packets degrade the throughput as well as energy efficiency of MAC protocols drastically. In addition, the modem may not be able to send out a packet at the scheduled time due to the busy terminal problem, which would impair the reservation-based MAC protocols, including handshaking-based and schedulingbased. The effectiveness of slotted mechanism may also need to be reexamined based on the ratio of the transmission time to

Aloha based method to the handshaking method in the real systems, which indicates that random access protocols have become promising solutions for UWANs. However, the low reliability and energy efficiency are still the major limitations to underwater applications and need to be addressed in future works. R EFERENCES

Fig. 7.

Node deployment in sea test

the propagation delay. As for random access based protocols, the enabled carrier sensing by the long transmission time and busy terminal phenomenon reduce the collisions and poses positive effect to some extent. In order to address the reliability of random access based protocols and the high collision probability caused by explicit acknowledgments, the way of adopting ACK packet in random access protocols should be designed carefully to minimize the adverse impact of long preamble problem. One viable solution for future work is to employ a specialized modulation method for small control packets, like the tone message in T-Lohi. But unfortunately the current tone message in T-Lohi is not able to carry all useful information. Another potential method is to implement the implicit control packet in underwater MAC protocols. The piggy-backed ACK packet for UWALOHA is one promising solution to mitigate the long preamble problem. Nevertheless, the piggy-backed acknowledgment leads to a series of new problems. For example, the acknowledgment will be delayed in low traffic rate scenarios because it has to wait for data packet to be piggy-backed with. This issue also occurs when the outgoing traffic is not balanced among nodes in the network. In both cases, the stop-and-wait mechanism becomes inefficient, which plays an important role in UWALOHA for the traffic rate controlling. Even through the piggy-backing idea is promising to address the long preamble problem in UWANs, the new problem poses new challenges to the future work on underwater MAC protocol design. VI. C ONCLUSION In this paper, we have studied two unique characteristics of acoustic modems and analyzed their impacts on the performance of existing MAC protocols as well as future MAC designs. Based on the experiment and simulation results, we demonstrate the significant performance degeneration of MAC protocols caused by the long preamble issue. Therefore, mitigating the long preamble in acoustic modems has become an practical and urgent issue for underwater network research. We also show that MAC protocols perform under the effect of both the propagation delay and transmission time in distinctive ways. The simulation results reveal the superiority of

[1] J. Partan, J. Kurose, and B. N. Levine, “A Survey of Practical Issues in Underwater Networks,” in Proceedings of ACM WUWNet’06, 2006. [2] L. Liu, S. Zhou, and J.-H. Cui, “Prospects and problems of wireless communication for underwater sensor networks,” Wireless Communications & Mobile Computing, vol. 8, pp. 977–994, October 2008. [3] M. Stojanovic, “On the Relationship Between Capacity and Distance in an Underwater Acoustic Communication Channel,” in Proceedings of ACM WUWNet’06, September 2006. [4] V. Capellano, “Performance Improvement of a 50km Acoustic Transmission through Adaptive Equalizationi and Spatial diversity,” in IEEE OCEANS’97, Nova Scotia, Canada, 1997. [5] N. Chirdchoo, W.-S. Soh, and K. C. Chua, “Aloha-based MAC Protocols with Collision Avoidance for Underwater Acoustic Networks,” in Proceedings of INFOCOM’07, April 2007, pp. 2271 – 2275. [6] B. Peleato and M. Stojanovic, “Distance Aware Collision Avoidance Protocol for Ad-Hoc Underwater Acoustic Sensor Networks,” in IEEE COMMUNICATIONS LETTERS, vol. 11, 12 2007. [7] A. Syed, W. Ye, and J. Heidemann, “T-Lohi: A New Class of MAC Protocol for Underwater Acoustic Sensor Networks,” in IEEE INFOCOM’08, 2008. [8] M. Xiaoxing Guo; Frater, M.R.; Ryan, “Design of a Propagation-DelayTolerant MAC Protocol for Underwater Acoustic Sensor Networks,” IEEE Journal of Oceanic Engineering, pp. 170–180, 4 2009. [9] N. Abramson, “The ALOHA System -Another Alternative for Computer Communications,” AFIP Press, vol. Vol 37, 1970. [10] Z. Peng, Z. Zhou, J.-H. Cui, and Z. J. Shi, “Aqua-Net: An Underwater Sensor Network Architecture: Design, Implementation, and Initial Testing,” in Proc. of MITS/IEEE OCEANS Conference, Biloxi, Mississippi, USA, October 2009. [11] M. Molins and M. Stojanovic, “Slotted FAMA: a MAC protocol for underwater acoustic networks,” in OCEANS 2006 - Asia Pacific, 2006. [12] P. Xie and J.-H. Cui, “R-MAC: An Energy-Efficient MAC Protocol for Underwater Sensor Networks,” in In Proceedings of International Conference on Wireless Algorithms, Systems, and Applications (WASA’07), 2007. [13] Z. Peng, Y. Zhu, Z. Zhou, Z. Guo, and J.-H. Cui, “COPEMAC: A Contention-based Medium Access Control Protocol with Parallel Reservation for Underwater Acoustic Networks,” in Proc. IEEE OCEANS, Sydney, Australia, 2010. [Online]. Available: http://ubinet.engr.uconn.edu/zhengpeng/publications/PID1202791.PDF [14] Y. Noh, P. Wang, U. Lee, D. Torres, and M. Gerla, “DOTS: A propagation Delay-aware Opportunistic MAC protocol for underwater sensor networks.” 2010 18th IEEE International Conference on Network Protocols (ICNP), 2010. [15] A. Syed, W. Ye, B. Krishnamachari, and J. Heidemann, “Understanding Spatio-Temporal Uncertainty in Medium Access with ALOHA Protocols,” in Proceedings of ACM WUWNet’07, 2007. [16] M. K. Park and V. Rodoplu, “UWAN-MAC: An Energy-Efficient MAC Protocol for Underwater Acoustic Wireless Sensor Networks,” IEEE Journal of Oceanic Engineering, vol. 32, no. 3, pp. 710–720, July 2007. [17] “Teledyne-Benthos Acostic Modems,” http://www.benthos.com. [18] L. Freitag, M. Grund, S. Singh, J. Partan, P. Koski, and K. Ball, “The WHOI Micro-Modem: An Acoustic Communications and Navigation System for Multiple Platforms,” in IEEE Oceans Conference, Washington DC, 2005. [19] Z. Peng, H. Mo, J. Liu, Z. Wang, X. Xu, S. Le, Y. Zhu, J.-H. Cui, Z. Shi, and S. Zhou, “NAMS: A Networked Acoustic Modem System for Underwater Applications,” in IEEE OCEANS, KONA, Hawaii, Sept 2011. [20] LinkQuest, “SoundLink High Speed Acoustic Modems,” http://www.link-quest.com/. [21] C. L. Fullmer and J. Garcia-Luna-Aceves, “Floor Acquisition Multiple Access (FAMA) for Packet-Radio Networks,” in ACM SIGCOMM’95, Cambridge, MA, USA, August 1995.