8. Conclusion

Multi-agent systems have now reached maturity, with a wide range of applications and methods and tools for engineering them. We have presented here the essential concepts to enable the reader to understand the conceptual foundations of MAS and begin their practical learning methodically. In particular, we have highlighted the cognitive and social aspects that make ADMs an original paradigm for the design and development of distributed applications exhibiting intelligent collective behavior. Among the fast-growing themes, we have chosen to detail reinforcement learning in multi-agent universes, which opens up promising prospects. While many advances have been made, ambitious new challenges remain, such as ethics, responsibility and the explanation of collective decisions made by a group of agents. These challenges are made more difficult by the distributed nature of an ADM's control and its...