Poster: Adaptive Topology Control for Wireless Ad-hoc Networks

Topology control in ad-hoc networks is the problem of adjusting the transmission power at network nodes in order to achieve the optimal topology that maximizes network performance. Recently, using effective topology control to optimize energy usage in the network has come into focus in the research community[1, 2]. The common thesis arrived at by existing works is that the transmission range used by mobile nodes should be the minimum required to keep the network connected. We refer to such a topology as the minimally connected topology in the rest of the paper. We argue in this work that the minimally connected topology does not always provide the optimal performance in typical ad-hoc networks. We show that in contrast, for typical ad-hoc networks with a few hundred nodes distributed over a few square miles area, the optimal topology is a function of the traffic load. 1. BASIS FOR THE MYTH The reasons for the presumed optimality of the minimally connected topology can be explained as follows. When the transmission range is decreased, the average hop-count for the paths traversed by flows increases linearly. However, the transmission power per hop decreases super linearly (given that the path loss exponent typically ranges from 2 to 4). Hence, the overall energy consumption in the network for the same amount of data transferred is minimized in a minimally connected topology. Furthermore, when the transmission range in a network is decreased, the average hopcount of flows in the network increases, which in turn increases the total number of one-hop flows (mini-flows) in the network increases, thus increasing the aggregate induced load in the network (We distinguish the basic load offered by the sources of the flows from the induced load that is the basic load multiplied by the average number of hops traversed by flows.). However, a decrease in the transmission range also increases the spatial re-use in the network, thus increasing the network capacity. It can be shown that, This work was funded in part by NSF grants ANI-0117840 and ECS-0225497, Yamacraw, and the Georgia Tech Broad-band Institute