The safety of autonomously acting systems depends on a reliable assessment of the systems' context. We propose a framework to formalise and analyse both qualitative and quantitative measures of the context quality in terms of safety and precision. The measures are based on order-theoretic arguments by relating the ground truth (as given by the real environment) with the context information (as inferred by the sensors). We derive a hierarchy of qualitative notions that can serve as a high-level requirement description for the sensor and controller implementation of the system. The quantitative part of the framework then allows for an evaluation of a probabilistic variant of the sensor regarding its safety and precision. We in particular treat the analysis of sensor fusion based on evidence theory.
11 / 2010
SaLsA Sichere autonome Logistik- und Transportfahrzeuge im Außenbereich