|
|
|
| e-Pub |
Previous | 
Home
Section: New Results
Simulator-based decision support
Participants : Philippe Besnard, Marie-Odile Cordier, Anne-Isabelle Graux, Christine Largouët, Véronique Masson, Laurence Rozé.
Ecosystem model-checking for decision-aid
Former studies of ecosystem modelling have concentrated on temporal modelling. In recent studies we have focussed on the formalization of spatial diffusion of a prey-predator trophic network composed of weeds and ground beetle. For this purpose, an approach coupling landscape representation and population models has been used. A reaction-diffusion model was developped through the synchronization ability of timed-automata. The agronomical rules of beetle migration and weeds diffusion have been translated into communications between timed automata. Landscapes have been simulated and can be evaluated thanks to landscape-metrics distance. The optimization aims to maximize the ground beetle abundance while minimizing the use of pesticides. The model obtained in this first study is quite complex but preliminary results are beeing studied.
Controler synthesis for optimal strategy search
Similarly to previous work, this approach relies on a qualitative model of a dynamical system. The problem consists in finding a strategy in order to help the user achieveing a specific goal. The model is now considered as a timed game automata expressing controllable and uncontrollable actions. The strategy represents the sequence of actions that can be performed by a user to reach a particular state (in case of a reachability problem for instance). A first approach based on a "generate and test" method has been developped for the marine ecosystem example [86] .
Recently, we generalized the work of Yulong Zhao applied in the context of a dairy production system [87] to the planning domain. The planning task consists in selecting and organizing actions in order to reach a goal state in a limited time and in an optimal manner, assuming actions have a cost. We propose to reformulate the planning problem in terms of model-checking and controller synthesis on interacting agents such that the state to reach is expressed using temporal logic. We have chosen to represent each agent using the formalism of Priced Timed Game Automata (PTGA). PTGA is an extension of Timed Automata that allows the representation of cost on actions and uncontrollable actions. Relying on this domain description, we define a planning algorithm that computes the best strategy to achieve the goal. This algorithm is based on recognized model-checking and synthesis tools from the Uppaal suite. The expressivity of this approach is evaluated on the classical Transport Domain which is extended in order to include timing constraints, cost values and uncontrollable actions. This work has been implemented and performances evaluated on benchmarks.
A datawarehouse for simulation data
In previous work we have proposed a datawarehouse architecture to store the huge data produced by deep agricultural simulation models [50] . This year, we have worked on hierarchical skyline queries to introduce skyline queries in a datawarehouse framework. Conventional skyline queries retrieve the skyline points in a context of dimensions with a single hierarchical level. However, in some applications with multidimensional and hierarchical data structure (e.g. data warehouses), skyline points may be associated with dimensions having multiple hierarchical levels. Thus, we have proposed an efficient approach reproducing the effect of the OLAP operators "drill-down" and "roll-up" on the computation of skyline queries [52] . It provides the user with navigation operators along the dimensions hierarchies (i.e. specialize / generalize) while ensuring an online calculation of the associated skyline.
Anne-Isabelle Graux, on leave from INRA (National Institute for Agronomical Research), is working on an adaptation and extension of this method for storing the simulation results of a comprehensive farm model named MELODIE [53] . The new datawarehouse will enable the analysis of simulation results within dynamic preferences, related to grassland management for instance, for identifying the data satisfying the best compromises with respect to possibly inconsistent criteria.
Post-mining classification rules
We consider sets of classification rules with quantitative and qualitative attributes inferred by supervised machine learning, as in the framework of the Sacadeau project. Our aim is to improve the human understanding of such sets of rules. First, we consider quantitative attributes in rules that often contain too many intervals which are difficult to intepret. We propose two algorithms to merge some of these intervals in order to get more understandable rules. These algorithms take into account the final rule quality. We are also working on formalizing what could be the quality of a set of rules. There are lots of studies about the quality of one rule but very few about the quality of the whole set of rules and this is still an issue.