Publication VIII

The problem of the capacity of a massively dense wireless multihop network can be broken down into separate problems at macroscopic and microscopic levels. At the microscopic level, from the local perspective, the network appears like an infinite plane with traffic that is uniform but flowing in many directions. Previous studies have assumed that it is sufficient to find the maximum sustainable density of packet flow in a single direction and use time sharing to serve flows in different directions. We show that this time-sharing limit can be exceeded by scheduling that truly interleaves the traffic flows in different directions. Determining the forwarding capacity for multidirectional traffic defines a new problem that has not been studied earlier. For finding numerical values, we adopt a constructive approach by simulating a finite but large network using greedy maximum weight scheduling. Biand fourdirectional balanced traffic patterns are studied. For the latter, an improved greedy algorithm is developed, using insights from our earlier work. Isotropic traffic plausibly yields the highest benefits for multidirectional forwarding, and our results show that a significant gain compared with single-directional forwarding can be achieved.