Multi-agent systems

Research into distributed artificial intelligence (DAI) and, in particular, multi-agent systems (MAS) emerged in the USA in the late 1980s. The first laboratory applications were the development of software based on speech recognition. Since then, this field has seen tremendous growth, and (in our opinion) has led to a certain revival of traditional artificial intelligence. In the 1990s, a desire for formalization led to the introduction of new concepts and the extension of previous work to other types of application. Today, industrial applications are appearing, and we are becoming aware not only of the underlying theoretical problems, but also of the need for a methodology in the design of such systems. This growing momentum is undoubtedly due to the integration of various factors, such as the theoretical interest in this new issue and the integration of new computing technologies into "everyday life".

On the one hand, our scientific motivation enables us to meet new needs in fields such as artificial intelligence and human-machine cooperation. The field of multi-agents initially emerged from research into artificial intelligence (AI). AI, by definition, covers a vast set of both theoretical and practical problems concerning the modeling and development of intelligent systems, in particular knowledge-based systems. However, it has been confronted with numerous theoretical problems, in particular that of combinatorial explosion. This new approach avoids the need to manipulate a large amount of knowledge in a single entity, by distributing it across several intelligent entities. This distribution is designed to reduce the combinatorial explosion of search. In addition, the integration of new intelligent entities in the same environment has extended the work of traditional AI, since each entity must integrate the behaviors of other entities in its environment.

In addition to expanding AI research, this field has helped build a framework for cooperation between man and computer. The increased reasoning capacities of computers have further liberated individuals by assisting them in their various tasks. Many researchers are studying environments where humans and machines interact. These studies deal with aspects such as the exploitation of large quantities of information, the complexity of the tasks to be solved or the consideration of human factors (ergonomic, sociological factors). However, interactions between individuals and machines generally require cooperation, which is justified by the fact that it is impossible for an individual to achieve his or her objective alone. Multi-agent systems allow us to extend previous work, while introducing a different vision of this cooperation issue. Indeed, reality shows us that cooperation is not a simple, static phenomenon,...