We present a new protocol for multi-hop routing in mobile ad hoc networks called dynamic nix-vector routing (DNVR). Our scheme is based on the nix-vector concept for efficient routing originally designed for wired networks. DNVR acquires a loop-free route and maintains it on a need basis as do other on-demand protocols. However, DNVR has several new features as compared to other existing reactive routing protocols which lead to more stable routes and better scalability. DNVR effectively validates the stored routes as well as efficiently senses the up-to-date network topology during a route discovery phase by sending a unicast probe packet. To accommodate networks with a high degree of mobility, the routing states are invalidated in a timely manner. DNVR adopts a conservative route discovery strategy by suppressing route requests in some cases, and thus only a few routes are maintained per destination. Moreover, it attains bandwidth efficiency by using a neighbor index and medium access control (MAC) addresses for routing purposes, which obviates the need for address resolution. We show via simulation that DNVR scales well to a large network with varying traffic load under diverse mobility scenarios. We compare DNVR to the well known dynamic source routing (DSR) protocol, which is believed to be one of the most efficient on-demand routing protocols. Simulation results reveal that DNVR is as efficient as DSR while at the same time providing higher packet delivery ratios (up to 46 % higher) and smaller delays (up to 23 % smaller) than DSR in most cases.

A novel load-balancing technique for ad hoc on-demand routing protocols is presented. Currently, ad hoc routing protocols lack load-balancing capabilities, and thus, they often fail to provide good performance especially in the presence of a large volume of traffic. We present a simple but very effective method to achieve load balance and congestion alleviation. The new scheme is motivated by the observation that ad hoc on-demand routing protocols flood route request (RREQ) messages to acquire routes, and only nodes that respond to those messages have a potential to serve as intermediate forwarding nodes. If a node ignores RREQ messages within a specific period, it can completely be excluded from the additional communications that might have occurred for that period otherwise. Thus, a node can decide not to serve a traffic flow by dropping the RREQ for that flow. In the new scheme, RREQ messages are forwarded selectively according to the load status of each node so that overloaded nodes can be excluded from the requested paths. Each node begins to allow additional traffic flows again whenever its overloaded status is dissolved. The new scheme utilizes interface queue occupancy and workload to control RREQ messages adaptively. The enhanced versions of protocols with this scheme are compared to the base protocols. Simulation results reveal that the new scheme greatly reduces packet latency as well as routing overhead without adversely affecting the network throughput, and it successfully balances the network load among nodes.

IP mobility support is provided by the basic Mobile IP protocol. The main drawback of the protocol is that the packets must be routed along the paths longer than the optimal one. This is known as the triangle routing problem. Recently, the route optimization protocol was proposed to solve the triangle routing problem, which allows packets to be routed along an optimal path from a correspondent node to a mobile node. However, the route optimization protocol may cause high signaling and processing costs. In this paper, a new scheme for reducing costs in route optimization is introduced to solve the above problems. Link and signaling cost functions are introduced to capture the trade off between the network resources consumed by the routing, signaling, and processing load incurred by the route optimization. In this new scheme, route optimization is performed only when it minimizes the total cost function, which provides the optimal result from the point of view of link and signaling costs. The simulation results show that the proposed scheme provides the best performance.