Date of Completion

9-2-2014

Embargo Period

8-29-2014

Major Advisor

Jun-Hong Cui

Associate Advisor

Sanguthevar Rajasekaran

Associate Advisor

Song Han

Associate Advisor

Mohammad Khan

Associate Advisor

Zheng Peng

Field of Study

Computer Science and Engineering

Degree

Doctor of Philosophy

Open Access

Campus Access

Abstract

In the past decade, underwater acoustic networks (UANs) have drawn significant attention due to its potential benefits. However, unique characteristics of underwater acoustic communications, such as long propagation delays and low available bandwidth, pose grand challenges to underwater acoustic network design, including media access control (MAC). In this dissertation work, we investigate MAC related issues in underwater acoustic networks. First, we propose a cross layer solution to solve heavy collisions in geo-routing protocols. Geo-routing protocols, determined by its broadcast nature, can only work with broadcast MAC, and therefore it cannot effectively avoid collisions. To provide a better reliability for geo-routing protocols, we smoothly integrate handshake-based MAC and geo-routing protocol in a cross-layer approach. It incorporates the advantages of both a geo-routing protocol and a reservation-based MAC protocol. That is, it performs stateless routing and efficiently avoids collisions. Benefiting from these features, our proposed solution can provide a much higher end-to-end reliability and better energy efficiency. Additionally, we newly identify the busy terminal problem, which is caused by acoustic modem designs. Because of this problem, in practice an acoustic modem cannot be interrupted at will when it is receiving. However, an ability to interrupt receptions is generally assumed by existing MAC protocols, so the busy terminal problem would significantly alter packet sending patterns and affect collision behaviors in underwater MAC protocols. To better understand its impact, we develop a new theoretical model of successful transmission probability for a random access MAC protocol. Furthermore, we apply the proposed model in network optimization through a case study on nodal throughput maximization. Thirdly, in field tests we observe that existing MAC protocols do not perform as well as expected. To find out the reason, we develop collision models for representative random access and handshake based MAC protocols. By analyzing these models, we notice that two characteristics of acoustic modems, long preamble and extremely low transmission rates, significantly increase collisions and degrade the MAC performance. Also, we conclude that a collision-free solution is promising to achieve a high performance. Following this thought, we propose a cluster-based on-demand time-sharing based MAC considering the aforementioned modem characteristics. As a follow-up work, we further propose a distributed on-demand scheduling based MAC, namely, DOS. Through pure scheduling, DOS guarantees collision-free transmissions of not only data packets but also control packets. Compared with existing collision-free scheduling MAC protocols, DOS is distributed and on-demand, i.e., it locally schedules transmissions according to nodes' dynamic transmission requests. Further, DOS does not require CDMA or power adjustment for collision resolution, and therefore it is able to work with most types of acoustic modems.

COinS