A Graph-Search Based Navigation Algorithm for Traversing A Potentially Hazardous Area with Disambiguation

The authors consider the problem of navigating an agent to safely and swiftly traverse a two dimensional terrain populated with possible hazards. Each potential hazard is marked with a probabilistic estimate of whether it is indeed true. In proximity to any of these potential hazards, the agent is able to disambiguate, at a cost, whether it is indeed true or false. The method presented in this paper is to discretize the terrain using a two dimensional grid with 8-adjacency and approximately solve the problem by dynamically searching for shortest paths using the A* algorithm in the positively weighted grid graph with changing weight function.