Autonomous Monitoring of Wildfires with Vision-Equipped UAS and Temperature Sensors via Evidential Reasoning

Abstract

This chapter describes a novel approach for autonomously estimating the state of an evolving wildfire in real-time. Information fusion regarding the fire spread employs the geospatial probabilistic state of the current fire spread in order to forecast to a future wildfire distribution whereupon the prior belief state is updated from the observations of two distinct types of sensing agents: (i) static temperature sensors embedded on the surface, and (ii) mobile vision sensors mounted on un-manned aerial vehicles (UAVs). The guidance and information fusion methods that direct the mobile vision sensors to useful sensing locations utilize the Dempster-Shafer theory of probabilistic evidential reasoning. The recursive bi-directional feedback procedure between the mobile vision agents and the forecasting operation exemplifies the Dynamic Data Driven Applications Systems (DDDAS) framework and is evaluated via simulation on its ability to estimate an expanding wildfire in real-time. The results indicate definitive improvement with the inclusion of mobile sensors versus conventional weather-based forecasts alone, and a number of enhancements are noted that can strengthen the current UAV path-planning algorithm.