Optimisation de la performance d'un réseau de capteurs sans fil pour les applications de collection de données à faible énergie et courte latence

Abstract

Emerging machine learning and artificial intelligence techniques expect to receive more and more data from IoT sensors and more often. These nodes must therefore be able to handle the multitude of data coming from countless sources. TSCH (Time Slotted Channel Hopping), the MAC protocol defined in the wireless communication standards IEEE 802.15.4e, plays a crucial role in reducing latency and minimizing power. In TSCH, time is divided into timeslots and a schedule determines what action a node needs to take in each timeslot (transmit, receive, or sleep). In the case of convergecast traffic, nodes close to the root are constantly in demand and have heavy traffic because all nodes are forwarding packets to that destination. In this context, cell scheduling algorithm, which are responsible of each node’s behavior at each timeslot, has to evolve further to support convergecast traffic and to provide solutions to the growing demand for data in today's world. In this thesis, we propose OSCAR, a novel autonomous scheduling TSCH cell allocation algorithm based on Orchestra algorithm. This new design differs from orchestra by the fact that OSCAR allocates slots according to the location of the node relative to the root. The goal of this algorithm is to allocate slots to nodes according to their needs. This algorithm manages the number of timeslot allocated to each node using the value of the rank described by the RPL routing protocol. The goal being that the closer the node is to the root, the more slots it gets in order to maximize the transmission opportunities. To avoid overconsumption, OSCAR sets up a mechanism to adjust the radio duty cycle of each node by reducing the slots allocated to inactive nodes regardless of their position on the network. We implement OSCAR on Contiki-ng and evaluate its performance on a network made up of ten nodes. The performance analysis of OSCAR shows that it outperforms Orchestra on the average latency and reliability especially when the traffic is heavy and congested. Finally compared to the previous gain the energy consumption does not increase much.

Date	17 Dec 2020
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Frédéric Nabki (Supervisor)