Performance of Local Forwarding Methods for Geographic Routing in Large Ad Hoc Networks

We consider local geographic forwarding methods in a network of stationary wireless nodes. The task of the local forwarding is to select the next-hop for each packet so as to maximize the flow of packets in a given direction. According to our network model with a slotted ALOHA type MAC protocol, the flow of packets, as expressed by the mean density of progress, depends on a minimal number of MAC layer parameters, namely the transmission range (or power) and the probability to transmit in a time slot. Our aim is to evaluate the maximum mean density of progress with respect to these two parameters for four different forwarding methods, distinguished by the next hop selection rules, and to compare their performance. Exact modeling of the performance in the network setting is difficult, and hence simulation has been used instead. Our results show that deterministic forwarding (MFR) suffers from concentration of the traffic onto certain deterministic paths. The two randomized methods are able to spread the traffic more efficiently and achieve a better performance. However, considerably better results are obtained with an opportunistic method, where the MAC layer is enhanced with functionality to choose the best available receiver, demonstrating the benefits of using local coordination.