Skip to Main Content
Context-dependent decisions in safety-critical applications require careful consideration of accuracy and timeliness of the underlying context information. Relevant examples include location-dependent actions in mobile distributed systems. This paper considers localization functions for personalized warning systems for railway workers, where the safety aspects require timely and precise identification whether a worker is located in a dangerous (red) or safe (green) zone within the worksite. The paper proposes and analyzes a data fusion approach based on low-cost GPS receivers integrated on mobile devices, combined with electronic fences strategically placed in the adjacent boundaries between safe and unsafe geographic zones. An approach based on the combination of a Kalman Filter for GPS-based trajectory estimation and a Hidden Markov Model for inclusion of mobility constraints and fusion with information from the electronic fences is developed and analyzed. Different accuracy metrics are proposed and the benefit obtained from the fusion with electronic fences is quantitatively analyzed in the scenarios of a single mobile entity: By having fence information, the correct zone estimation can increase by 30%, while false alarms can be reduced one order of magnitude in the tested scenario.