The research conducted by the SMAC team is resolutely multi-disciplinary and seeks to design ‘virtual test tubes’ that facilitate the study in silico of complex systems through the observation of autonomous entities endowed with an artificial intelligence. This approach is described as « individual-based" because its aim is to conceptualise, calibrate and simulate complex collective phenomena by specifying the individual behaviours of autonomous agents. Even if they are endowed with artificial intelligence, these agents have limited perception and rationality.

In other words, the dynamics of the multi-agent system cannot be modelled by predictive and solvable equations, but is the result of individual behaviours and their interactions. This approach is particularly well suited to the design of simulators in fields as varied as road traffic, crowd simulation, computational finance and social simulation. In these fields, the tools and platforms created by the team are used for both simulation and distributed problem solving.

In particular, the team works on the development and coding of intelligent behaviours for artificial entities, the engineering of software platforms to exploit them, and the development of methods for moving from problem analysis to implementation. The team is very sensitive to practical and applicative aspects of this approach. In particular, it offers a number of platforms for tackling these problems as efficiently as possible, with a constant concern to avoid simulation bias, and provide reproducible results.

Some links about our work:

• Github
• Demos