Réseaux de capteurs sans fil avec récolte d'énergie : techniques de transmission pour les applications sensibles au délai

In recent years, the use of wireless sensors has increased. Wireless sensor networks (WSNs) have attracted the interest of the industry and have conquered many application areas. The deployment of WSNs for many applications, such as environmental monitoring and autonomous driving, is becoming an important aspect of future wireless networks. Since WSNs are composed of dense remotely-located sensor nodes, powering such devices through wired power cords or batteries and ensuring their long-term maintenance is not always viable or cost efficient. For this reason, green communications have attracted the interest of researchers, especially with the proliferation of the Internet of Things (IoT) paradigm. For example, a green sensor can harvest energy from various ambient sources. Despite the promised benefits of energy harvesting (EH) powered WSNs (EH-WSNs), green sensors face random energy supply, making it very difficult to meet stringent quality of service (QoS) requirements and real-time communications. In addition, future IoT applications will increasingly rely on the exchange of fresh and regular time-sensitive information for monitoring and control, such as in the case of autonomous vehicles. Therefore, a key challenge we address in this thesis is the design of innovative transmission techniques and resource allocation algorithms that operate under the constraints of the energy randomness and also meet the QoS requirements of ubiquitous applications, especially in terms of delay and freshness of information. In the first part of this thesis, we propose a transmission scheme in an EH-WSN to minimize the delay as well as the packet loss rate. The proposed protocol optimizes the scheduling of sensor tasks by exploiting the estimation of the wireless channel state and considering the associated time and energy costs as well as the costs associated with sensing and data transmission. Our analysis raises the following interesting question : is it more efficient to (i) immediately transmit the detected data and retransmit the packets if necessary or (ii) wait until the channel is in good condition ? To evaluate the performance of this protocol, we consider the packet loss and two complementary delay metrics : (i) the age of the delivered information which measures the time between the generation of the information at the sensor and the time of its reception by the sink, and (ii) the inter-delivery time which measures the time between two status updates successfully received by the sink.We derive the distributions and statistics of these metrics using probability theory and stochastic processes. In a second part of this thesis, we optimize the inter-delivery time of updates in an EH-WSN whose sensors are equipped with capacitors.