Heterogeneous ad hoc networks: performance improvement through supplementary agent nodes

In an ad hoc network environment, it is not uncommon for nodes to have varying capabilities for attributes like battery life and mobility. Throughout this dissertation, we focus on such a heterogeneous structure and routing strategies that can differentiate between nodes and possibly exploit the unevenness of their characteristics. We start with a framework that categorizes nodes into two main classes as agents and clients. Agents, in the broadest sense, are defined as nodes not constrained by a finite battery life as they plug into AC power or have much longer lasting batteries with clients being the battery-operated mobile users. One possible interpretation of this framework is to populate a homogeneous ad hoc client network with agent nodes in order to increase the network's reliability and to reduce the amount of client transmissions by shifting the burden of packet relaying to agents. In fact, the positioning of agents in a stationary, regular mesh formation is a particularly interesting case and is referred to frequently throughout the dissertation. By simulating ad hoc networks in the presence and absence of an agent mesh, we demonstrate that agents can indeed provide better connectivity and lower client energy consumption, provided that an intelligent routing technique is used. For that purpose, we created the Agent-Supporting Ad Hoc On-demand Distance Vector Protocol, AS-AODV. AS-AODV is an AODV-based ad hoc routing algorithm that is capable of differentiating between agents and clients and integrating costs to the routes built, features lacked by AODV. As a result, AS-AODV is able to intelligently construct routes by forcing them to go through agent nodes as often as possible, hence minimizing the relays by clients. The performance of AS-AODV is established through a comprehensive set of simulations performed on an OPNET platform.