Subject Ph. D.

The emergence of the Internet of Things (IoT) - the networked connection of things, processes, data and people - has dramatically multiplied the number of connected devices in the world, billions of units we have today today to the tens of billions that should be deployed in the coming years in industry, the automotive sector, health, agriculture, the smart city, the general public... (IoTAnalytics, 2022). There is a broad consensus on the fact that the question will not only be to manipulate and exploit all the data that these objects generate at the Cloud level, but beyond to manage and orchestrate billions of connected objects and their data in a way coordinated as close as possible to the place of their generation by using at least Cloud network infrastructures. The expected objectives are, among other things, the reduction of the carbon footprint, and improved decision-making and reaction capacities as close as possible to the processes.

The "Edge Computing" approach is attracting growing interest across all industry sectors to address the aforementioned challenges through its ability to provide new ways to maximize operational efficiency, automate business processes and improve overall performance. through the processing and analysis of data collected by IoT objects. Adding the exponentially growing performance of microelectronic systems, as well as the increasing throughput and low latency of 5G communications to embedded applications on IoT objects opens up major areas of innovation. One of the current conceptual and scientific developments is the transformation of the management of independent connected objects towards the intelligent coordination of autonomous objects in a swarm, called "Swarm Computing" or distributed intelligence in a swarm [1].

Distributed swarm intelligence requires continuous cooperation between autonomous objects (e.g. smart objects, machines, robots, ...) and cloud service platforms, allowing applications to be executed by structures self-formed and self-organized ad hoc of distributed and heterogeneous intelligent objects.

Harnessing these advances, swarms embrace the concept of collective collaboration of heterogeneous resources and data reaching consensus for self- management for a common purpose. A Swarm is an autonomous entity that orchestrates and integrates separate autonomous objects, embracing the power of the collective and acting on a common purpose. The object in the swarm must be identifiable and authenticated while the swarm, as a whole, must be protected against malicious behavior and trustworthy to function as intended.

On the other hand, the energy consumption and network constraints of the smart nodes that contribute to the swarm must also be taken into account in order to offer sustainable swarm intelligence and maximum energy autonomy [2].

On the other hand, advances in embedded systems that involve heterogeneous devices based on multi-processor-on-chip (MPSoC), neuromorphic processing units (NPUs) like Google's Tensor Processor Unit (TPUs), are turning IoT nodes into objects intelligent, can offer a platform for experimentation and proof of concept, with considerable computing power for light artificial intelligence (Light AI and TinyML). The concept of "Swarm computing" combined with the growing capacities of IoT equipment to integrate AI models opens the way to new proposals for increasing the performance and functionalities of autonomous connected objects managed in swarms in industrial fields. of IoT.

The cyber-security issue must be taken into account in a distributed intelligence architecture. Thus dedicated security (and trust) agents distributed in the swarm can contribute to collective authentication and trust, as well as intrusion detection which will prevent and detect malicious attacks on nodes [3].

The research work of this thesis takes place within the framework of intelligent systems in the application fields of industry 4.0, the environment and energy. It will be based on the paradigms of smart objects and their interactions in cyber-physical systems (CPS) environments according to the paradigm of "Swarm computing" and embedded light artificial intelligence, under energy constraints for autonomy. items.

It aims to design, simulate and test coordinated distributed intelligence models and methods according to the "Swarm computing" approach by integrating light artificial intelligence methods into IoT nodes based on a dual hardware/software infrastructure.

The expected objectives are the improvement of performance and knowledge of systems, rapid and contextual decision-making by objects, interactions between objects and operators, data analysis by artificial intelligence methods embedded on objects. .

This work will be based on previous and ongoing work carried out in the laboratory as part of thesis, post-doctorate and an ANR project. Thus, the work previously carried out as part of the thesis [4] has made it possible to identify an innovative axis of interaction in a community of communicating objects by proposing a socio-inspired approach taking up the models of human interaction resulting from the work of sociologist and anthropologist. Objects are able to interact by requesting and exchanging embedded services within a community created in an ad-hoc and dynamic way.

The relative and global localization of autonomous communicating objects is an integral part of the coordination process of IoT entities, and the work previously carried out as part of the post-doctorate [5] has identified methods and models for approximating distances based on on wireless communication characteristics by measuring RSSI (Radio Signal Strength Indication).

The work carried out since 2021 on the ANR project [6] aims to define a methodology and tools from continuous automation and discrete event systems based on the observation and monitoring of the states and movements of assets. entities, autonomous entities and operators, involved in potentially dangerous and critical activities in storage, handling and processing operations within industrial facilities. The results of the ANR project will be used as elements of knowledge of the thesis subject.