25 Mars – Thesis defense - Michel Bakni

10 h In videoconference

Cross-level energy-aware dimensioning tool for Wireless Sensor Networks.

In recent years, Wireless Sensor Networks (WSN) have gained attention as a major research topic in electronic and information fields. WSN consists mainly of a set of wireless nodes, the network is scalable, providing a way to contact thousands of nodes together. Nodes can have different architecture and purposes allowing a large set of implementations to be realized.
WSNs have strict requirements. For example, circuits have to fit into low-power and low-computation and limited-size conditions. Nodes need to be autonomous  and networks need to be flexible in terms of structure. Considering these necessities, developing this kind of networks has become a complex task, especially in the early stages of design, where essential decisions are to be made.
In this context, simulation is a useful solution. It provides a way to create different scenarios with hundreds or thousands of nodes with the ability to provide input configuration and obtain output results. This can save time, cost, and effort, especially when difficult conditions need to be addressed such as harsh environments.
In order for simulations to work precisely, accurate models for real-world systems need to be created. These models represent composants of the communication system such as network protocols or physical quantities like humidity or temperature. The models include parameters and algorithms that describe the behaviors of the modeled object and different approaches can be used to achieve that.
In this thesis, we started looking at the approaches used to build WSN models. The simulators are based on the models, and thus they follow the approach's modelling concepts. Based on the literature study, three approaches are used to model WSNs: single-level, multilevel, and cross-level. The latter can provide a solid framework for complex topics such as energy and routing. However, based on our survey, it is not widely implemented in the WSN simulation especially from the energy-oriented aspect.
To detect how simulators handle WSN challenges, a study was made on the most used simulators in the last decade. As a result, 4 simulators were selected and studied.
The study of simulators that were in use helped understanding how to address modelling approaches and energy problems. Regarding the modelling, none of the selected simulators follows the cross-level approach. Three are single-level, and the fourth is multilevel. In term of energy, the simulators were examined to detect how energy generation, storage, and consumption was handled. With the obtained knowledge, the main question was formed as follows: In WSN, how to model energy from a cross-level perspective?
To answer this question, we proposed an energy-aware cross-level model for Wireless sensor networks. Our model includes four levels: use case, system, node and circuit. In each level, a set of parameters resides. Those parameters can interact without limits linked to the level boundaries.
We have implemented a simulator based on the proposed model. We created a WSN oriented simulator that is both cross-level and energy-aware using MATLAB. Our simulator accepts XML configuration files and it presents results obtained from different levels of simulation, using a built-in graphical interface. The simulator has been also compared with other already in-use simulators and physical nodes.

Event localization