Presenter:

Felipe Augusto Costa de Oliveira, DLR Bremerhaven

Inviting Professor:

Prof. Dr. Alberto Garcia-Ortiz

Abstract:

The rapid advancement of technology in recent decades has led to increasing automation in safety-critical applications, making use of complex sensor systems as an important part of decision-making processes. These advances, albeit beneficial for societal development, introduce a heightened dependence on sensor data accuracy. In this context, the need to develop robust methods for monitoring sensor integrity becomes increasingly crucial. This work introduces the concept of a multi-sensor cross layer integrity framework, as a step towards improved reliability of automated systems in safety-critical scenarios.
 
The concept of organizing and encapsulating information in different abstraction layers has been proposed in numerous fields and has been advantageous in many applications. Inspired by that idea, the proposed method relies on identifying and separating the sensor information in different abstraction layers, and then assessing and propagating the integrity information along these layers. The main hypothesis is that this layered representation can improve both the integrity assessment in multi-sensor solutions, and the overall resilience of the system being monitored by those sensors.
 
The framework, currently under development, will have the flexibility to be employed in diverse multi-sensor system configurations, integrating state-of-the-art fault detection and error estimation methods to handle various adverse conditions and fault scenarios. As a preliminary case study, a computer vision scenario, with the integrity assessment of maritime object detection, will be presented.