Henry-Joseph Audéoud - Efficient and Guaranteed Routing in Wireless Sensor Networks

The wireless sensor networks that we work with in this thesis are a set of devices connected to each other by low-rate and low-power technologies.  Their role is to produce measures on the physical environment around them (meteorological and climate condition tracking, monitoring of industrial installations, control of distribution grids, topographical surveillance…).  These measures must then be collected out of the network.  Since the sensors have short range radios, transmissions are multi-hop, the sensors close to the destination relaying the information transmitted by those which are further away from it.  Because of the movement of the nodes themselves or of objects in their environment interfering with wireless communications, the exact topology of the network is subject to change.  In addition, the battery-powered sensors are limited in energy and therefore in transmission abilities.  The power-saving techniques applied to turn off the radio most of the time impose synchronization constraints. To route information through the network, the routing protocol establishes routes, so that the sensors can relay information from and to the network border router through reliable links leading to the destination through short paths.  Due to sensor limitations, the routing must be energy efficient, i.e. the overload of the radio transmissions involved by the routing algorithm itself must be as lightweight as possible.  It must also be able to restore connectivity on a network topology change without creating routing loops that negatively impact the quality of service and the energy reserves of the nodes.
This document describes a routing protocol that meets these objectives. It is capable of creating a self-healing collection tree that extracts information out of the network, as well as from the routes to distribute command messages or acknowledgment to the nodes.  It also validates the data path of each packet to ensure that they never enter a routing loop. The protocol is run in simulations and also on real platform experiments, showing the effectiveness of the proposed mechanisms.
In order to improve its ability to choose the best available links, I also propose the use of a new estimation of their quality.  It is based on two complementary measurements: a long-term measurement of the ambient noise level on the radio channel, and a measurement of the power of the signal received from the transmitter.  These two measurements provide an estimate of the signal-to-noise ratio, and thereby the expected reception rate.  This estimate is both accurate, quick to obtain, and adapted to the constraints of sensors and networks we are talking about.