Probabilistic Mission Landscapes for Neuro-Symbolic UAV Navigation
A neuro-symbolic architecture (ProMis) that turns geospatial maps and neural-sensor perception into Hybrid Probabilistic Logic Programs, via Probabilistic Clause Modules that map raw data into continuous and categorical distributions over spatial relations. Solving the resulting program yields a Probabilistic Mission Landscape: a scalar field giving, at each point of a drone's navigation space, the belief that legal and operator-defined mission conditions hold. Students learn how declarative legal and safety knowledge can be fused with machine-learning perception (including LLMs and vision transformers) to certify autonomous-aircraft missions in an interpretable way.
Probabilistic Mission Design in Neuro-Symbolic Systems Simon Kohaut1, Benedict Flade2, Daniel
ProMis (Probabilistic Mission Design) is a neuro-symbolic system for deciding where an autonomous aircraft may legally and safely fly. It transforms static geospatial data (roads, buildings, geofence…