When a flood or an earthquake cuts through a city, the first thing responders lose is not communication but certainty about which roads still work. Ground surveys are slow and dangerous, and satellite passes are infrequent and often obscured by cloud. This paper presents the design of a simulation-based system in which a small team of unmanned aerial vehicles surveys a disaster-affected road network, classifies each road segment by how passable it is, and feeds that classification directly into an emergency routing service that ambulances and rescue vehicles can query. The system is built as six layers: a PX4 software-in-the-loop and Gazebo simulation, a ROS2 middleware layer, a computer-vision perception stage based on OpenCV preprocessing and YOLOv8 detection, a decision layer containing an explicit Road Accessibility Score model, a FastAPI backend, and two client applications. The main methodological contribution is the separation of detection from interpretation: the neural network reports only what it observes, while a transparent, weight derived scoring function decides what those observations imply for road usability. We describe the scoring model, its weight derivation through the Analytic Hierarchy Process, the change-detection procedure that compares pre- and post-disaster surveys of the same simulated environment, and the routing formulation that treats blocked segments as removed edges rather than merely expensive ones. We set out the seventeen relations that define the pipeline, the evaluation framework and the targets against which the implemented system will be measured, and we state plainly which parts of the problem simulation cannot answer, and we extend the evaluation with a set of derived result visualisations a criterion-weight chart, a sim-to-real precision/recall comparison, a confusion-matrix heat map, a routing/change-detection summary, and a system-timing budget that make the reported targets easier to audit at a glance.

disaster response,road accessibility,path planning,object detection,Unmanned Aerial Vehicles (UAVs),PX4,ROS2,Simulation