Direction forwarding for highly mobile, large scale ad hoc networks

In this paper, we present a novel packet forwarding scheme for wireless ad hoc networks --- "Direction" Forwarding (DFR). Popular routing protocols such as DSDV and AODV use "predecessor" based forwarding, namely, the packet is forwarded to the predecessor on the shortest path from the destination, as advertised during the last update. Predecessor forwarding may fail in large scale networks where the routing update rate must be reduced by the need to maintain link O/H below reasonable levels. However, if nodes are mobile, routing table entries may become "stale" very rapidly. DFR is designed to overcome the "state" routing table entry problem. When the routing update arrives, the node remembers not only the predecessor delivering the update, but also the update "direction" of arrival. When a packet must be forwarded to destination, it is first forwarded to the node ID found in the routing table. If the node has moved and ID forwarding fails, the packet is "direction" forwarded to the "most promising" node in the indicated direction. At first glance, DFR seems to combine the features of table based routing and geo-routing. However, direction forwarding differs from geo-routing in that the direction is learned from the routing updates, instead of being computed from the destination coordinates. Thus, DFR does not require destination coordinates, a global coordinate system, or a Geo Location Server. In the paper we show the application of DFR to a scalable routing scheme, LANMAR. Through simulation experiments we show that DFR substantially enhances LANMAR performance in large, mobile network scenarios.