Markovian Wildfire Modeling via Evidential Reasoning

This paper presents a new evidential reasoning approach for autonomously modeling the state of an evolving wildfire. The objective is to update the current geospatial representation of a wildfire through the inclusion of the available evidence regarding a wildfire's presence at a given location. This evidential support is calculated from data provided by imperfect temperature and machine vision sensors using belief functions based in the Dempster-Shafer theory of probable reasoning. The derivation and mathematical validity of these belief functions is shown through rigorous analysis and a simplified example is considered that exhibits the benefits of this evidential reasoning approach. Based on the methods and models explored in this work, concepts for autonomous path-planning procedures are discussed for further studies.