Toward a Closed-Loop Autonomous Sensing Framework for UAS-Based Particulate Matter Mapping

Abstract

Small uncrewed aerial systems (sUAS) offer a powerful capability for high-resolution characterization of fine particulate matter (PM2.5) within the atmospheric boundary layer; however, their operational utility is limited by rotor-induced aerodynamic disturbances, platform-dependent sensor bias, and the absence of integrated architectures that couple sensing, learning, and control. This paper presents an end-to-end autonomous particulate matter sensing framework that unifies sensor calibration and validation, CFD-guided sensor placement, custom payload integration, synchronized onboard computation, physics-informed measurement correction, and autonomous flight execution within a single system architecture. Computational fluid dynamics of a DJI Matrice 100 quadcopter is used to identify an aerodynamically quiescent sampling region for an Alphasense OPC-N3 optical particle counter, while a companion-computer-based pipeline synchronizes particle measurements with vehicle telemetry using MAVLink during autonomous flight. To mitigate flight-induced measurement distortion, residual-based regression models incorporating telemetry and physics-derived aerodynamic features are employed to correct airborne PM2.5 observations relative to stationary references. The resulting corrected particulate estimates provide a reliable perceptual state for spatial mapping and higher-level autonomy. Field experiments demonstrate stable autonomous data acquisition, consistent sensor–telemetry synchronization, and statistically meaningful reduction of airborne–stationary measurement discrepancies. Collectively, this framework establishes a scalable foundation for closed-loop airborne particulate matter sensing and represents a transferable autonomy pattern applicable to tactical autonomous systems that must integrate perception, learning, and control to operate under real-world uncertainty.