Modelling and control of complex systems

“Modelling and management of complex systems” theme

The McLEOD theme (Modelling and Control of Complex systems in Large Environments requiring Optimised Decision) focuses on modelling and controlling complex systems.

By nature, these systems have complex and changing dynamics that are difficult to predict over time. To control them they must first be observed, which remains an interesting challenge, especially for large-scale networked systems. The McLEOD thematic group develops various approaches to model such systems:

• data-driven modelling, which involves using data mining and learning techniques on observations and information about the system to obtain models,
• “grey box” modelling which includes expert knowledge on the system that goes beyond the available data,
• agent-based modelling, which is an emerging approach that considers that the overall dynamics of the system to be simulated is the result of individual and local behaviours.

The models designed make it easier to understand and control the system studied. To achieve this last objective, McLEOD looks for complementarities between automatic and artificial intelligence techniques. Hybrid control architectures are designed by integrating high-level decision mechanisms (Markovian models, multi-agent coordination, combinatorial optimisation model) to develop optimal orders adapted to the dynamics of the system in response to the objectives and in compliance with the inherent constraints; and low-level control methods (predictive approaches, data-based approaches) ensuring compliance with commands while guaranteeing their stability, and techniques for detecting and locating faults (unknown input observers, data-based approaches, hybrid approaches) to correct failures.