IRL CROSSING

The particularity of the thesis is to combine approaches from Automatic Control, and in particular those of Human-Machine Cooperation, with those of Computer Science and Multi-Agent Systems. Indeed, technology allows us to imagine a real complementarity between humans and machines. So-called "human" automation has demonstrated the many combinations of shared or complementary functions between humans and machines. Often focused on a specific field of application and restricted to rigid forms of cooperation, automation levels do not take advantage of human adaptability or AI advances. Recent work has led to the definition of adaptable automation levels and a methodology. Its aim is to take into account both the skills of human and artificial agents (know-how, expertise, experience) and their capacity (workload, fatigue, energy control) to build cooperation and make it evolve according to the situations, skills and ability to carry out the tasks induced by these situations, but also according to their skills and ability to cooperate (savoir-coopérer). The models proposed by this cooperation must now be able to be instantiated with AI tools.
This thesis therefore proposes to merge the advances made by automatic control in taking the human into account in decision-making and control processes, with those of artificial intelligence on the ability of systems to learn from humans. The aim is for agents - both human and system - to learn from each other the functions they need to perform to control a situation, but also to learn from each other in order to build efficient cooperation. In particular, this topic wishes to focus on the agents' abilities to communicate so that each can build a representation of the other.

The system must therefore be able to explain its skills and also what it has understood about the human's skills. The construction of a Common Workspace will support such communication, by facilitating the sharing of information about the situation and the agents. Agents will thus develop an awareness of the team situation and be transparent with each other. A gain in human trust is also expected, while taking care to avoid excesses.
The application framework of the thesis would be crisis management and more specifically the sharing of functions between humans and robots for the control of a crisis situation, such as fire control.