Adaptive multi-metric routing in distressed mobile sensing networks

This paper proposes a data dissemination framework to support participatory sensing in a disaster response network formed by mobile, battery-driven devices in case of a large-scale outage or infrastructure failure. Such networks are often unpredictable, heterogeneous, and constrained in terms of energy, bandwidth, and storage. The delivery ratio of the messages highly depends on the routing information used to decide on optimal message forwarding and replication. Various routing metrics are employed in different network environments, ranging from a static connected network to a mobile disconnected environment. In this unpredictable and heterogeneous environment, no single metric results in optimal decisions at all times. We propose a framework that can use several widely used metrics as plug-ins thereby making more intelligent decisions through a dynamic combination of these metrics. Since every routing metric makes specific assumptions about the underlying network, its routing information becomes more accurate when these assumptions hold (e.g. static nodes or certain mobility patterns). Our protocol automatically adapts to a previously unknown network where the appropriate routing assumptions can not be determined beforehand or is different throughout the network. We evaluate the protocol rigorously to illustrate how our combination mechanism provides accurate information from a set of metrics which may disagree. We compare our protocol against widely used sensor network and DTN routing protocols and show that it provides a higher delivery ratio in a wide range of scenarios.