Peacock: a Benchmarks Generation Framework for High-Level Information Fusion Evaluation

Abstract

This work presents Peacock, a framework that aims at generating benchmarks for High level information fusion. Peacock makes it possible to generate several structured information sets, that are representative, coherent, diversified and controlled. The principle of Peacock lies in the generation of several information sets from one scenario. The scenario contains on the one hand, a storyboard of perfectly described events and a chronology of perfectly structured information. On the other hand, it contains communicating entities, organized in a network. These entities will alter the scenario events as well as the information exchanged. The modifications on the information consist in the introduction of imperfections (over-precision, imprecision, incompleteness, uncertainty, irrelevance, incomprehension) according to the entities communication behaviors. In this paper, we present the principles under the Peacock framework. We detail its characteristics and describe its implementation.