World-state representation is a fundamental requirement for robotic agents operating under automated planning frameworks. However, existing planning frameworks typically assume a fully symbolic discrete world state, whereas other components of real robotic systems operate on sensor-derived continuous state representations. The gap between sensorderived state representations and symbolic planning state remains largely unaddressed by existing knowledge bases, storing either only shallow symbolic facts or rich semantic representations. This paper presents a modular knowledge base approach that offers both: a structured continuous world state representation and a symbolic world model usable by automated planning.We introduce a fluent registry in which planning fluents are declaratively defined as conditions over a persistent knowledge base. The proposed knowledge base is integrated into the ROS 2-based AUSPEX multi-UAV framework and its applicability in a real-world Search and Rescue (SAR) scenario is demonstrated.     «

World-state representation is a fundamental requirement for robotic agents operating under automated planning frameworks. However, existing planning frameworks typically assume a fully symbolic discrete world state, whereas other components of real robotic systems operate on sensor-derived continuous state representations. The gap between sensorderived state representations and symbolic planning state remains largely unaddressed by existing knowledge bases, storing either only shallow sy...     »