Abstract EN:

Microcontrollers are small computers that can make great things. With low power consumption, they are equipped to interact with the environment and give feedback using rf. With such abilities, we could build a mesh network, to monitor a house, an industrial facility, or a whole city. However, scalability of meshed networks is a complex topic, and numerous existing solutions for managing meshed network, submitted to robustness and performance constraints, make compromises. In this thesis, we contribute to an alternative solution following our own compromises. We propose a minimalistic method to build scalable meshed networks of energy-autonomous nodes, able to self-organize without central coordination. We run distributed applications using synchronous rounds. To put such a model in practice, we discuss the synchronization topic : two nodes never run at the same speed, so there is an increasing skew between clock values. But, we need to keep neighbors in phase with each others. We give a minimalistic algorithm to set synchronous execution rounds. The shared nature of the wireless medium and its sensitivity to disturbances make it unsuitable for our synchronization algorithm without tweaks. Thus, we present a protocol giving a necessary framework to run our synchronization algorithm on wireless nodes. Once we are able to run synchronous algorithms, we aim at simplifying the deployment of networks by adding geolocation with software only. We use tiny differences in clock speeds to measure precisely propagation delays of radio waves. Finally, we consider a way to interface centralized systems with our distributed networks allowing to broadcast data from several entry points.