ADAPTIVE AND HETEROGENEOUS MOBILE WIRELESS NETWORKS
Tech report number
CERIAS TR 2004-51
This dissertation investigates two research problems: (a) designing ad hoc routing protocols that monitor network conditions, select routes to satisfy routing requirements, and adapt to network topology, traffic load, and congestion; (b) building an integrated infrastructure for heterogeneous wireless networks with movable base stations and developing techniques for network management, routing, and security. The experimental study of ad hoc routing protocols shows that the on-demand approach outperforms the proactive approach in less stressful situations, while the later one is more scalable with respect to the network size. Mobility and congestion are the primary reasons for the packet loss for the on-demand and proactive approaches respectively. Self-adjusting congestion avoidance (SAGA) routing protocol integrates the channel spatial reuse with the multi-hop routing to reduce congestion. Using the intermediate delay as the routing metric enables SAGA to bypass hot spots where contention is intense. An estimate of the transmission delay is derived based on local information available at a host. Comparison of SAGA with AODV, DSR, and DSDV shows that SAGA introduces the lowest end-to-end delay. It outperforms DSDV in the measured metrics. SAGA can sustain heavier traffic load and offers higher peak throughput than AODV and DSR. It is shown that considerations of congestion and the intermediate delay can enhance the routing performance significantly. Hierarchical mobile wireless network is proposed to support wireless networks with movable base stations. Mobile hosts are organized into hierarchical groups. An efficient group membership management protocol is designed to support mobile hosts roaming among different groups. Segmented membership-based group routing protocol takes advantage of the hierarchical structure and membership information to reduce overhead. A secure packet forwarding algorithm is designed to protect the network infrastructure. The roaming support algorithm cooperates with the proposed mutual authentication protocol to secure both the foreign group and the mobile host. The evaluation shows that the computation overhead of the secure packet forwarding is less than 2% of the CPU time, and that of the secure roaming support ranges from 0.2% to 5% of the CPU time depending on the number of hosts and their motion. This justifies the feasibility of the security mechanisms.
Department of Computer Sciences