Capacity of the wireless network is very limited. High performance routing protocols are highly desired to efficiently utilize the scarce wireless resources. But, the constraints of power and bandwidth in wireless network, in combination with the dynamic network topology by node motion, make routing in ad hoc networks extremely challenging. Scalability and mobility become the most critical challenges for routing protocol design in mobile ad hoc networks (MANETs). In this dissertation, we show that many existing routing protocols designed for wired networks experience significant performance degradation problem when applied in large ad hoc wireless networks. Moreover, node mobility is one of the parameters that most critically affect the performance of network protocols (e.g., routing or data dissemination). Several novel routing schemes are proposed in this dissertation to enhance the routing performance in large ad hoc networks under dynamic motion scenarios. Some realistic mobility models and the characteristics of motion patterns are proposed in this dissertation to help us to understand what factors critically affect the performance of network protocols.

The proposed Geo-LANMAR routing protocol is scalable in large ad hoc networks with group motion. It integrates local table-driven packet forwarding and long distance geo-forwarding and supports a hierarchical design on routing table and routing table update. The DFR and Geo-DFR routing protocols are scalable in large ad hoc networks without group motion. They integrate on demand, table driven routing (i.e., direction forwarding) with geo-routing and feature efficient recovery from dead ends with the help of direction forwarding. CIDR and GIDR are scalable inter-domain routing protocols for routing among MANETs. They handle mobility by using group motion based clustering and geo-routing schemes, which achieve the scalability and the robustness to mobility.

In this dissertation, a track-based group mobility model is proposed, which is based on a Markov Chain approach and is capable to capture heterogeneous mobility behavior with group and individual motion as well as static nodes. This diversity makes the track model a good candidate for modeling realistic military or urban scenarios in simulation experiments. A unifying parameter of Neighborhood Change Rate (NCR) for different motion patterns is also proposed in this dissertation. Simulation results show that NCR is able to characterize the intrinsic properties of complex topologies or motion patterns and feature the complex spatial and temporal dependencies observed in realistic mobility patterns. The analytic studies of NCR give a close-form expression for NCR and the analytic results of NCR are successfully applied in the scenarios of epidemic dissemination.