EDEN Framework for Interactive Analysis of Ecosystems Models
Abstract
Understanding ecosystems is crucial, in particular to take conservation actions. One way to do so is formal modelling and analysis. In this paper, we present the eden (Ecological Discrete-Event Networks) framework that provides ecologists with discrete modelling languages and dedicated analysis tools. These tools are based on well-known techniques in computer science, like symbolic state-spaces or model-checking, but used in quite a different way. Indeed, most formal analysis techniques provide yes/no answers to well-defined questions, possibly with a witness execution, which is good to assess whether a system exhibits or not a given property. However, most questions in ecology are not stated as “Does the system have such behaviour?” but rather as “Why does the system sometimes has such behaviour and how can we prevent it from happening?” Moreover, these questions are often hard to express formally. With eden, we propose an exploratory way to build progressively a representation of this behaviour that is suitable to answer such questions. The question itself being formalised on the way, together with the model exploration. This approach is based on a hybrid representation of the state-space that can be incrementally split into a graph of components (symbolic sets of states) linked by transitions. The goal is thus to provide the users with a human-readable representation of the state-space that can be fine-tuned with respect to the questions of interest, resulting in an object that constitutes by itself the expected explanation. While eden is rather specific to ecology, we advocate that its analysis method and tools could be beneficial for other domains.
Origin : Publisher files allowed on an open archive