ROPA: A MAC Protocol for Underwater Acoustic Networks with Reverse Opportunistic Packet Appending Hai-Heng Ng, Wee-Seng Soh, Mehul Motani Department of Electrical and Computer Engineering National University of Singapore Email: {g0600158, elesohws, motani}@nus.edu.sg Abstract—In most existing sender-initiated handshaking based underwater Media Access Control (MAC) protocols, only the initiating sender is allowed to transmit data packets to its intended receiver after the channel has been reserved; none of the potentially backlogged neighbors of the sender can transmit in the duration after the current handshake. Therefore, each of those neighbors must initiate their own handshakes, which incur additional overheads and potentially result in poor channel utilization. In this paper, we present a novel approach to increase the channel utilization by allowing a sender to invite its one-hop neighbors (appenders) to opportunistically transmit (append) their data packets. After the sender finishes transmitting its packets to its own receiver, it can immediately switch its role to receive the incoming appended data packets, which arrive in a packet train manner. This greatly reduces the relative proportion of time spent on control signaling. We refer to this MAC protocol as ROPA – Reverse Opportunistic Packet Appending. From our extensive simulations and comparisons with existing protocols, we show that ROPA significantly increases the channel utilization and offers performance gains in terms of throughput and delay. I. I NTRODUCTION Underwater acoustic networking is deemed as the enabling technology for many underwater sensing applications [1]. How- ever, unlike terrestrial wireless communications, underwater Media Access Control (MAC) protocol design must address challenges posed by underwater acoustic communications, such as slow propagation speed and low bit rate-distance product. Among the existing underwater MAC protocols, there is a strong focus on handshaking based protocols as they provide multi-fold benefits such as the carrying of useful information in the control packets, and the ability to alleviate hidden node problem. In the following, we briefly describe some of these protocols. In [2], Sozer et al. propose a handshaking based MAC protocol with Stop-and-Wait Automatic Repeat Request (ARQ) mechanism. In [3], Molins and Stojanovic propose Slotted-FAMA, which uses a 4-way (Request-to- Send (RTS)/Clear-to-Send (CTS)/DATA/ACK) handshake with a time-slotting mechanism to avoid any data collision. However, it requires clock synchronization, and long slot length. In [4], Guo et al. introduce a MAC protocol called APCAP, which improves channel utilization by allowing a sender to take actions for other packets (e.g., starting another handshake, transmitting data packet) while waiting for a CTS frame to return. In another attempt to improve channel utilization, Chird- choo et al. [5] propose the MACA-MN, which sends multiple packets to multiple neighbors in a single 3-way handshake. This reduces the relative proportion of time spent on RTS/CTS signaling. However, due to the long duration of each handshake, the average waiting time can be very long before a node gains control of the channel to transmit. In [6], Chirdchoo et al. propose the RIPT protocol, which adopts a receiver-initiated approach to coordinate the packets from multiple neighbors to arrive in a packet train manner at the receiver. Although this approach also reduces the relative proportion of time spent on control signaling, the adoption of a receiver-initiated approach often demands a complex traffic prediction algorithm. In [7], we propose the MACA-U protocol; its key feature is the incorporation of several state transition rules that account for the long propagation delay in underwater. It aims to serve as a more appropriate benchmarking protocol, as previous works in underwater MAC protocols often rely on applying terrestrial MAC protocols in underwater for benchmarking. In existing sender-initiated handshaking based MAC proto- cols, only the sender is allowed to transmit single or multiple packets during a successful handshake; none of the other back- logged neighbors can transmit in this handshake, even though the channel has been reserved. In this paper, we introduce the novel concept of “reverse opportunistic packet appending”, for which we call the resulting MAC protocol “ROPA”. Similar to some of the aforementioned protocols, it seeks to improve channel utilization by reducing the proportion of time spent on control signaling. However, it achieves this differently by allowing the initiating sender to invite its one-hop neighbors (appenders) to opportunistically transmit (append) their packets. After the initiating sender finishes transmitting its data packets to its receiver (primary data transmissions in the forward path), it can immediately switch its role to receive the incoming appended data packets from multiple neighbors (secondary data transmissions in the reverse path), which arrive in a packet train manner. This is in contrast to the conventional approach, which requires each of those backlogged neighbors to initiate a separate handshake that incurs its own overheads. Similar to other handshaking based protocols, the ROPA uti- lizes information extracted from the control packets to alleviate the hidden node problem. The ROPA’s framework is also very versatile; when none of the sender’s neighbors has any packet to append, it can still perform its forward path’s transmissions, and this reduces to MACA with packet train. On the other hand, if the sender only receives its neighbors’ appending requests, but does not hear from its own intended receiver, it can still