The explosion in the number of connected objects, assistant robots and human-machine-machine interfaces has led to a growing democratisation of cyber-physical and social-technical systems. These are systems made up of both human users, robots and artificial agents in social interactions. While the democratisation of these tools is a major element for new services on a daily basis, its dissemination faces two main barriers: on the one hand, the recognition of human activity remains unclear, both at the operational level (location, mapping, identification of objects and users) and cognitive (recognition and monitoring of intent). On the other hand, the interaction involves different vectors that must be adapted according to the context (robotic, mixed reality, virtual reality), the user and the current outrageous situation.The scientific locks surveyed by the INCA project are based on two distinct applications and a cross-cutting issue. The first concerns the design and evaluation of virtual and mixed learning environments to assess the impact of conversational agents in simulationsimmersive and interactive. The second axis examines the problems specific to robotics: the perception of the environment and the representation of space-temporal knowledge for social interaction and navigation. Finally, the problem of managing social dialogues of mixed initiative with users is common to these two applications.So, the INCA project aims to perceive and represent an environment in order to reason on it, interact naturally and socially with users and learn a user model in order to adapt the interactive behaviour of agents.