Research themes
Complementary research themes are developed in collaboration with other INRA scientists, in particular from the Agronomy and Environment department.
We develop methods of explicit representation of biophysical systems and decision processes in view of their analysis by simulation. Relevant disciplines include knowledge engineering, knowledge management (in particular ontological modeling) and system analysis. Conceptual modeling aims at: (i) characterizing and formalizing the generic structures of production systems so as to ease the analysis of their configuration and functioning; (ii) making it possible to reuse the elaborated concepts in different applications.
This theme concerns the development of optimisation algorithms for problems of squential decision under uncertainty. The value of the criterion to optimize is provided by a simulator of the controlled biophysical system. The methods studied belongs to various domains including reinforcement learning, stochastic algorithms and combinatorial optimization.
The sequential decision problems addressed by the team often requires to consider other options than the classical averaging maximization of a unique criterion. This theme aims at studying which other criteria could be used and addressing the algorithmic issues raised by their incorporation in decision support software. Non-linear and non-probabilistic criteria are considered in the setting of decision theory extended to deal with qualitative (ordinal) criteria and a possibilistic representation of uncertainty, which yields a notion of qualitative utility function. The methods developed have roots in possibility theory approaches of decision under uncertainty and multicriteria decision-making.
Centre INRA de Toulouse, Auzeville
BP27, 31326 Castanet Tolosan cedex,
France