We develop approaches based on psychological and social metaphors for the learning, control, and dynamic social organization of autonomous agents within cyber-physical collectives. This includes bio-inspired mechanisms, combined with optimization tools or artificial learning, enabling the localization of autonomous agents, navigation in cluttered environments, cumulative learning of complex behaviors (constructivist approach), and the development of social behaviors influenced by mechanisms of emotional contagion.

A collaboration between the Co4Sys team and the CTSYS team focuses particularly on the testing of embedded multi-agent systems. This involves evolving and improving the algorithm for accessing information in the graphs representing the agents in the sensor network and automating the tests.

We are also developing flexible control architectures for a class of cyber-physical systems by exploiting the flat differential description of their dynamics and artificial learning methods. The targeted applications include 3D surveillance and reconstruction with a team of mobile robots, precision agriculture, and microgrid direct current (DC) energy systems. In these applications, critical tasks require the presence of multiple agents and strategies for coordination, communication, and optimization of time and energy consumption.

A complex artificial system can be composed of multiple component systems that may be emergent or self-organized systems. These emergent systems highlight collective productions obtained through the cooperation of the entities that constitute them in a completely decentralized context. The difficulty in observing and controlling the resulting emergent systems lies in detecting, registering, and manipulating these collective productions. It is necessary to work on the observation, design, and control of these particular systems, which must rely on a multi-level vision of structures and behaviors. We are particularly interested in detecting and manipulating emergent structures or behaviors, the interaction mechanisms between emergent systems, and the methodological support for their composition.

A specific project concerns the cooperation of heterogeneous systems for creating intelligent ecosystems with applications in resource management and natural disaster management. This involves developing architectures capable of seeing the emergence of different social organizations (and not just different instantiations of the same organization) and experiencing organizational changes (unplanned) to adapt to environmental variations.

We develop structured discrete approaches for modeling and controlling dynamic systems. These approaches rely either on geometric discretization preserving the properties and structure of multiphysics systems or on the discrete modeling of the physics of these systems. We are also working on unifying concepts of analysis and design of control laws based on thermodynamics, with applications notably in the management of wastewater networks (complex water systems) and the study of metabolic networks. Additionally, we are interested in the design of constraint-based control using optimization (e.g., economic MPC, NMPC) with safety guarantees and applications, particularly in industrial processes.