Learning-based route management in wireless ad hoc networks

The nodes in a wireless ad hoc network must act as routers in a self-configuring network without infrastructure. An application running on nodes in the ad hoc network may require that intermediate nodes act as routers, receiving and forwarding data packets to other nodes to overcome limitations of noise, router congestion and limited transmission power. In existing routing protocols, the "self-configuring" aspect of network construction has generally been limited to route selection using a shortest-path routing metric as a predictor of routing efficiency. This limited, network-layer predictor fails to consider the effects of existing traffic on router loads and fails to consider the effects of noise experienced at the MAC layer. Not all network topologies are suited to efficient routing using a shortest-path metric. The location of the nodes and physical characteristics of the network environment can create topologies where shortest-path routing overloads some routers and underutilizes others. Similarly, noise sources can undermine the quality of wireless links depending on the relative distance between the noise sources and the receiving nodes. This dissertation presents a cross-layer predictor that combines the effects of noise and router congestion into a single time-based routing metric based on statistical estimation from recent experience. Also presented is a new cross-layer, adaptive routing protocol, called Warp-5, that not only uses the new routing metric to make better initial routing decisions in a noisy or congested network, but can also adjust previously existing routes as new routes or new noise sources are added to the network. Simulation results for Warp-5 are presented and compared to the existing shortest-path routing protocol AODV. The results show the cross-layer approach of Warp-5 to be superior to shortest-path routing protocols for managing router congestion and noise in wireless ad hoc networks.