Probabilistic Methods for Routing in Wireless Sensor Networks

Scientific Context. Recently, there was a growing interest in wireless sensor technologies. Such technologies have large-scale applications such as earthquake monitoring or military applications and small-scale services such as home automation. Sensors are small devices that can generate data about the environment (for example, measure of temperature) and can used them for specific services (for example, emit an alarm when surrounding temperature is to high). They embed wireless communication capabilities for exchanging data between them, shaping thus an implicit network called a Wireless Sensor Network (WSN). A crucial problem for WSN is the one to transmit data from a given source sensor to a given destination. It is called routing. A routing protocol encapsulates the data into messages and then computes a path in the WSN to move them from sensor to sensor until reaching the target. Routing protocols for WSN have two specifities. First, communications within a WSN are subject to faults such as link failures due to interferences and crash failures due to the low-power batteries of sensors. The structure of a WSN is highly dynamics and routing paths cannot be computed statically and once for all. Second, sensors can only communicate with other nearby sensors: routing paths can only be computed locally, piece by piece thanks to distributed algorithms.