Roger MARTIN-CLOUAIRE is director of the laboratory Unité de Biométrie et Intelligence Artificielle since October 2006.
Roger Martin-Clouaire holds a Masters in Biomedical Engineering (1982) from Saskatchewan University (Canada) and a PhD (1986) in Artificial Intelligence (AI) from Toulouse University. His research works in these graduate student years focussed primarily on developing and implementing possibility theory models of approximate reasoning in knowledge-based systems.
He joined INRA (Institut National de la Recherche Agronomique) in 1987 as a research scientist in the newly created computer science department of INRA.
Current research work
Central theme = modelling and simulation of intelligent agents involved in production management tasks
The relevance of simulation approaches to the study and design of agricultural production systems is widely claimed. The methodology and computer software appropriate to such a task have still not reached the state of a mature technology and are mainly developed in research laboratories. Suitable computer models need to represent the structure and dynamics of the underlying biophysical system together with the coordinated human activities involved in the management of the farm production process. Production process improvement involves studying interactions among biophysical processes and decision making processes at the farm level while most existing approaches tend to address one at the expense of the other: the human actor is part of the feedback process, not standing apart from it.
The objective to enhance modelling & simulation capabilities to meet the system research needs yielded to the development (in collaboration with J-P Rellier) of DIESE, a framework especially designed for building and running agricultural production system models. The classical approach to representing decision making in simulation models is to express decision behaviour through a set of decision rules. This approach becomes cumbersome as the number of rules grows beyond a threshold; the meta knowledge about the proper use of the rules (e.g. which should be applied first when several are applicable) is hard to represent and makes the rule-base hard to maintain and reuse. By contrast DIESE relies on a much richer conceptual basis under the form of an ontology of agricultural production systems. It supports the modelling of the decision process in terms of activities, resources required to realize them, and well-structured constraints bearing on the relevance and feasibility of activities, the interdependencies between them and the restrictions on the uses of resources.
Computationally the ontology comes under the form of a C++ library. In developing a farm production system model, the ontology acts as a metamodel; implementing a model amounts to particularizing the ontology concepts as required by the domain and then instantiating the corresponding classes to capture the specific aspects of the system to be simulated. A discrete event simulation mechanism realizes the step by step interpretation of the strategy and the progressive execution of the decided activities, which in turn alter the biophysical state that, otherwise, changes only in response to external factors such as weather.
DIESE is currently used in large modelling projects dealing with various kinds of production such as cash crop, vineyard, pasture-based livestock and pig systems. The development of the ontology was largely inspired by the analysis made in a modelling project on greenhouse tomato production systems. These projects attest to the wide scope of applicability of the framework.
Past research interests at INRA
- Constraint satisfaction with soft and/or uncertain constraints
- Planning under uncertainty with nondeterministic actions
- Handling uncertainties in regional soil maps
- Management problems in agricultural production systems
- greenhouse (climate control)
- grassland-based livestock systems (rotational grazing)
Some recent papers
R Martin-Clouaire and J-P- Rellier. Modelling and simulating work practices in agriculture. International Journal of Metadata, Semantics and Ontologies, 4(1-2):42-53, 2009.
N Cialdella, J P Rellier, R Martin-Clouaire, M H Jeuffroy, and J M Meynard. Silasol: A model-based assessment of pea (pisum sativum l.) cultivars accounting for crop management practices and farmers’ resources. In Proceedings of Farming Systems Design 2009, Monterey, CA, August 2009.
G Martin, M Duru, R Martin-Clouaire, J P Rellier, and J P Theau. Taking advantage of grassland and animal diversity in managing livestock systems: a simulation study. In Proceedings of Farming Systems Design 2009, August 2009.
G Martin, L Hossard, J P Theau, O Thérond, E Josien, P Cruz, J P Rellier, R Martin-Clouaire, and M Duru. Characterizing potential flexibility in grassland use. application to the french aubrac area. Agronomy for Sustainable Development, 29(2):381–389, 2009.
A Ripoche, J P Rellier, R Martin-Clouaire, A Biarnès, N Paré, and C Gary. Modeling dynamically the management of intercropped vineyards to control the grapevine water status. In Proceedings of Farming Systems Design 2009, August 2009.
C Rigolot, X Chardon, J P Rellier, R Martin-Clouaire, J Y Dourmad, A Le Gall, S Espagnol, C Baratte, and P Faverdin. A generic framawork for the modelling of livestock production systems: Melodie. In In Proceedings of Farming Systems Design, Monterey, CA, August 2009.
M Tchamitchian, R Martin-Clouaire, B Jeannequin, J Lagier, and S Mercier. Serriste: a daily set point determination software for glasshouse tomato production. Computers and Electronics in Agriculture, 50:25-47, 2006.
F Garcia, F Guerrin, R Martin-Clouaire, and J-P Rellier. The human side of agricultural production management - the missing focus in simulation approaches. In Robert M Zerger, Andre; Argent, editor, MODSIM 2005 International Congress on Modelling and Simulation, pages 203-209. Modelling and Simulation Society of Australia and New Zealand, 2005.