MAD English

Aller à : navigation, rechercher

Modelling of Agro-ecosystems and Decision (MAD)   Mad pew.png

The MAD team members develop mathematical, statistical and computer science methods for the analysis and contro of agro-ecosystems.

The research activities of the team can be grouped into the three following domains:

Decision models (Economics, Artificial Intelligence)

  • Sequential decision under uncertainty (MDPToolbox)
  • Risk aversion of economic agents
  • Limited rationality of economic agents
  • Multi-agent / multicriteria decion models, flexible plans

Design, exploration and analysis of simulation models (Computer science, Statistics, Artificial Intelligence)

  • Heterogeneous and multi-scale simulation (VLE)
  • Multi-agent simulation (Gama)
  • Simulation-based model exploration (Mexico research network)
  • Bayesian optimisation (R packages DiceOptim GPareto GPGame)

Optimisation and learning in factored decision problems (Artificial Intelligence, Stochastic graphical models, Statistics)

  • Optimisation in decision problems (GMDPtoolbox, FA-FMDP)
  • Dynamic bayesian networks learning

These theoretical works also give rise to applied research advances in Agroecology, Ecology, Disease control, Forest management...


  • Study of the impact of the knowledge of underground water reserves (RUEdesSOLS project)
  • "Idetypes" design using numerical models (PhD thesis of Léonard Torossian, 2016-2019, collaboration with UMR AGIR, Toulouse and UMR AGAP,  Montpellier)
  • Influence of landscape composition and agregation on the reachable tradeoffs between ecosystem services (AGROBIOSE project)
  • Ecosystem services-based design and control of agro-ecosystems
  • Reconstruction of the spatio-temporal dynamics of plants with seed "dormance" (PhD thesis of Sebastian Le Coz, 2015-2018, collaboration with CEFE, Montpellier)


  • Biodiversity conservation and ecosystem services provision in species/services interaction networks (PhD thesis of Hui Xiao, 2015-2018, collaboration with  CSIRO / University of Queensland)
  • Ecological interactions network reconstruction exploiting species presence/absence data (PhD of d'Etienne Auclair, 2015-2018, collaboration with Univ. of Minnesota, Univ. of California in Santa Barbara and the Agroecology lab, INRA-Univ. of Dijon)

Disease control

  • Sequential decision under uncertainty and game theory for the control of non-regulated animal diseases (collaboration with the BioEpar lab, INRA-ONIRIS)
  • Optimisation of plant disease control exploiting landscape features: application to the sharka virus (collaboration with the BGPI lab, INRA Montpellier)

Forest management

  • Uncertain dynamics modls and Markov Decision Processes: Analysis of the impact of risk aversion on forest management under climatic threat
  • Forest owners risk preference measure (collaboration with LEF lab, Nancy)
  • Analytical models of risk cover measures adoption (collaboration with LEF lab, Nancy)

In addition, several projects of the MAD team are developped in collaboration with the RECORD team and implemented  in the RECORD platform.

Team composition, March 2018


Team leader : Régis SABBADIN (05 61 28 54 76) sabbadin[at]inra[dot]fr

Génotoul BioInfo
Outils personnels