Probabilistic search with agile UAVs

Abstract

Through their ability to rapidly acquire aerial imagery, Unmanned Aerial Vehicles (UAVs) have the potential to aid target search tasks. Many of the core algorithms which are used to plan search tasks use occupancy grid-based representations and are often based on two main assumptions. Firstly, the altitude of the UAV is constant. Secondly, the onboard sensors can measure the entire state of an entire grid cell. Although these assumptions are sufficient for fixed-wing, high speed UAVs, we do not believe that they are appropriate for small, lightweight, low speed and agile UAVs such as quadrotors. These platforms have the ability to change altitude and their low speed means that multiple measurements may easily overlap multiple cells for substantial periods of time. In this paper we extend a framework for probabilistic search based on decision making to incorporate multiple observations of grid cells and changes in UAV altitude. We account for observation areas that completely and partially cover multiple grid cells. We show the resultant impact on a number of simulation examples.