Persistent Visual Tracking and Accurate Geo-Location of Moving Ground Targets by Small Air Vehicles

Small unmanned aerial vehicles are increasingly popular in many applications for their low cost, ease of use, and rapid deployment. One highly desirable capability is the pursuit of a target along a roadway while providing a persistent aerial view. This is made di cult by large UAV state uncertainty and limited maneuverability, however by carefully representing state uncertainty, planning in the space of available control actions, and using environmental knowledge to constrain possible target motion, adequate performance can be achieved with existing elded vehicles. This paper presents a comparative study of several target estimation and motion planning techniques and motivates the need for teams of multiple UAVs. Results from eld experiments applying these ideas to a commercial human-portable xed-wing UAV system are also presented.