Routing in wireless sensor networks

The Internet of Things (IoT) paradigm envisions expanding the current Internet with a huge number of intelligent communicating devices. Wireless Sensor Networks (WSN) deploy the devices running on limited energy supplies and measuring environmental phenomena (like temperature, radioactivity, or CO2). Popular WSN applications include monitoring, telemetry, and natural disaster prevention. Major WSN challenges are energy efficiency, overcoming impairments of wireless medium, and self-organisation. WSN integrating IoT will rely on a set of open standards striving to offer scalability and reliability in a variety of operating scenarios and conditions. Nevertheless, the current state of the standards present interoperability issues and can benefit from further improvements. The contributions of the thesis are the following:

• We perform an extensive study of Bloom Filters and their use in encoding node characteristics in a compact form in IP addresses. Different techniques of compression and variants of filters allowed us to develop an efficient system closing the gap between feature-routing and classic approaches compatible with IPv6 networks.
• We propose Featurecast, a routing protocol/naming service for WSN. It allows to query sensor networks using a set of characteristics while fitting in an IPv6 packet header. We integrate our protocol with RPL and introduce a new metric that increases routing efficiency. We validate its performance in both extensive simulations and experimentations on real sensors on a large-scale Senslab testbed. Large-scale simulations demonstrate the advantages of our protocol in terms of memory usage, control overhead, packet delivery rate, and energy consumption.
• We introduce WEAVE, a routing protocol for networks with geolocation. Our solution does not use any control messages and learns its paths only by observing incoming traffic. Several mechanisms are introduced to keep a fixed- size header, bypass both small as well as large obstacles, and support efficient communication between nodes. We performed simulations on a large scale involving more than 19 000 nodes and real-sensor experimentations on the FIT IoT-lab testbed. Our results show that we achieve much better performance than other protocols, especially in large and dynamic networks, without in- troducing any control overhead.