Probabilistic Boundary Coverage for Unknown Target Fields with Large Perception Uncertainty and Limited Sensing Range

摘要

Imagine that an unmanned aerial vehicle (UAV) is dispatched to map boundaries of excessive wind shear or low pressure regions in storm cells (see Fig. 1). The UAV has to plan its motion based on its on-board sensor readings to quickly enclose the unknown target field (UTF), which is a form of boundary coverage problem. However, UTFs often do not have a clear boundary or a priorly known dispersion function and the UAV has to get closer to the field to take multiple readings to predict field dispersion for boundary coverage. Moreover, the sensor readings often contain large uncertainties due to variations of the field itself or difficult sensing conditions. It is clear that regular boundary traversing techniques are not applicable. Such problems are not unusual. Another example is that an inspection robot is tasked to find thin hairline cracks on airport runway. These applications propose a new problem: how can we design a principled approach to ensure the robot can effectively cage UTFs under large perception uncertainty and limited sensing range.