2019_programme: A BAYESIAN INFORMATION FUSION APPROACH TO PLANNING AND EVALUATION FOR SYSTEM OF SYSTEMS IN AUTONOMOUS NAVAL MINE-HUNTING OPERATIONS
- Session: 19. Situation-adaptive distributed networked systems
Organiser(s): Ehlers Frank, Schulz Arne
- Lecture: A BAYESIAN INFORMATION FUSION APPROACH TO PLANNING AND EVALUATION FOR SYSTEM OF SYSTEMS IN AUTONOMOUS NAVAL MINE-HUNTING OPERATIONS [invited]
Paper ID: 923
Author(s): Strenzke Ruben, Strode Christopher
Presenter: Strenzke Ruben
Presentation type: oral
Abstract: Planning and evaluating naval mine-hunting missions requires a situational picture. This picture shall include the information gathered by modern image-producing sonar sensors. Therein, a central component must fuse the information gathered by the sensors of a system of heterogeneous systems. We hence propose a novel framework for planning and evaluation for a mine-hunting system of systems. The framework calculates a geo-referenced mine presence belief by applying Bayesian statistics on detections and through-the-sensor performance self-assessment information. Thus, a mine presence probability map and an information availability map are produced; the latter showing the general confidence in either mine presence or absence. Furthermore, the framework can determine the number of future sensor looks necessary to achieve a certain target risk for vessels that are transiting the area of interest. This is done by simulation on sensor performance predictions to drive the mine presence belief to a context-derivable lower or upper threshold, which results in a sensor-specific future effort map. We describe the algorithms that generate the three before-mentioned map types from pre-processed performance assessment and target recognition data - without relying on any prior knowledge on mine density. Furthermore, we present results from simulation experiments involving systematic deviations of performance self-assessment from the actual performance. The results suggest that the geo-referenced fusion of multiple independent data collection runs improves the information availability per cell and the estimated number of mines - even if there is a constant high bias in performance self-assessment.
Download the full paper
- Corresponding author: Mr Strenzke Ruben
Affiliation: NATO STO CMRE
Country: Italy
e-mail: