Implementation and real-world evaluation of routing protocols for wireless ad hoc networks

Sammanfattning: A wireless ad hoc network consists of a number of mobile nodes that temporarily form a dynamic infrastructure-less network. To enable communication between nodes that do not have direct radio contact, each node must function as a wireless router and potentially forward data traffic on behalf of the others. New routing protocols are needed as the existing protocols used in wired networks adapt too slowly to the frequent topology changes induced by mobility and are inefficient in terms of resource consumption. During the last five to ten years more than 60 different ad hoc routing protocols have been proposed, optimized and partially compared based on simulation studies. However, we believe that these simulation studies need to be complemented with real-world experiments in order to gain practical experience and reveal real-world effects that may not be visible in simulations.This thesis surveys the field of ad hoc routing and related real-world testbeds. We also describe our research results from three included papers:In our active networking approach to ad hoc routing, protocol logic is carried inside data packets, which enables new protocols to be independently deployed at run-time. As a proof of concept, we have implemented two ad hoc routing protocols and successfully used them for routing a real-time sensitive audio stream. This prototype gave us a good understanding of the practical aspects of wireless networking and prepared good ground for protocol comparisons.We have implemented a testbed called APE which enables easy management and analysis of real-world experiments. In real-world experiments, with up to 37 mobile nodes, we used the APE testbed to assess the reproducibility between repeated testruns and compared three different routing protocols. This testbed has become a crucial tool for discovering flaws in proposed protocols, as well as for benchmarking different routing styles.During our implementation of the AODV protocol we discovered a performance problem that we could relate to some invalid assumptions about the wireless link characteristics. We explored this "communication gray zone" problem and implemented three countermeasures which we compared by using the APE testbed. As a result we could eliminate the performance discrepancy between simulated AODV and its implementation in the real-world.