Dynamic Target State Estimation for Autonomous Aerial Vehicles using a Monocular Camera System

Operations involving autonomous vehicles require knowledge of the surrounding environment including other moving vehicles. The use of vision has been regarded as an enabling technology that can provide such information. Several important applications that would benefit from this technology is autonomous aerial refueling (AAR) and target tracking. This paper considers a sensor fusion approach using traditional IMU/GPS sensors with vision to facilitate the state estimation problem of moving targets. The proposed method makes use of a moving monocular camera to estimate the relative position and orientation of targets within the image by exploiting a known reference motion. The vision state estimation problem is solved using an homography approach that employs images containing both the reference and target vehicles. A simulation involving an unmanned aerial vehicle (UAV) and two ground vehicles is documented in this paper to demonstrate the algorithm and its accuracy. This algorithm along with a system outline and practical issues for autonomous aerial refueling will be described.