Underwater sensor networks (UWSNs), the enabling technology for a broad range of aquatic applications, rely on acoustic communication for long-range transmissions. However, underwater acoustic communication suffers from several adverse features, such as long propagation delay and low bandwidth, making the networking design for UWSNs much more challenging than their terrestrial counterparts. Medium access control (MAC), a crucial component in networking design, coordinates among multiple agents that share the same channel resource and is responsible to schedule packet transmissions efficiently and fairly. However, due to the long propagation delay, the traditional handshaking-based random-access MAC protocols are not channel efficient for UWSNs.
In this dissertation, we address the inefficiency issues in random-access MAC for UWSNs, using a stochastic sending probability-based approach. We propose three handshaking-free underwater MAC solutions targeting the same goal: high network throughput, low packet end-to-end delay, robustness under dynamics and controlled implementation complexity. The three solutions are based on a common utility-based probability optimization framework, but with different design considerations and objective functions. We first leverage the feature of long propagation delay, often taken as negative, to improve the parallelism between multiple senders. Our proposed protocol, the Delay-Aware Probability-based underwater MAC protocol (DAP-MAC), characterizes the group compatibility relation, a proposed indicator for successful concurrent transmissions, and utilizes this relation in the stochastic optimization framework for the best transmission strategy. The drawback of DAP-MAC is it requires long time slots to accommodate concurrent transmissions. We reduce the slot size and explicitly resolve the unique spatial-temporal uncertainty issue in UWSNs in our proposed protocol, the Traffic-Adaptive Receiver-Synchronized underwater MAC protocol (TARS). In TARS, we consider the data queue status and design the throughput-optimal transmission strategy to be traffic adaptive, which is very suitable for mobile and traffic-varying UWSNs.
Simulation results show that TARS achieves the highest throughput and lowest packet delay among the typical representative MACs (about 13%~146% higher in throughput and 13%~21% lower in delay than others in a mobile ad hoc network). For further throughput improvement, we consider the capture effect and proactively create power capture at the receivers in our proposed protocol, the Stochastic MAC protocol with Randomized Power control (SMARP). We design a randomized power control scheme based on the non-negligible difference in acoustic propagation attenuation, and include the capture success probability in the stochastic optimization framework.
Simulation results show that SMARP yields better throughput and packet delay than TARS and other representative MACs (about 23%~240% higher in throughput and 27%~60% lower in delay than others in a mobile ad hoc network). Finally, toward the development of wireless networking for enabling marine observatory capabilities at the Northeastern University Marine Science Center, we implement the TARS protocol on our testbed. We conduct in-tank underwater experiments and evaluate the performance of the TARS protocol under a three-node single-hop network. Experimental results show that TARS achieves the highest packet delivery ratio with controlled low packet end-to-end delay than other representative handshaking-free underwater MAC protocols, and is very suitable for the actual implementation in UWSNs.
Advisor: Professor Yunsi Fei
Professor Edmund Yeh
Professor Stefano Basagni
Professor Adam Ding