We focus on the problem of having a small, fixed-wing unmanned aerial vehicle (UAV) equipped with a gimbaled camera keep sight of a moving ground target with adequate resolution for image detection and tracking. The gimbal mechanism has a limited range of pan-tilt angles, thereby inducing blind spots. Also, we make no assumptions regarding target motion. Therefore, the challenge is achieving the said objective while overcoming visibility constraints, uncertain target motion, and stochastic action effects that arise from unmodeled UAV dynamics and environmental disturbances. To solve this problem, we propose two different solutions. The first formulates the problem as a two-player zero-sum game and the second as a stochastic optimal control problem. Both problem formulations utilize dynamic programming to generate receding-horizon optimal control policies. The successful field test results of both approaches are presented and discussed.
Next, we discuss preliminary work on the optimal coordination of two UAVs performing vision-based target tracking. To focus on the nature of the optimal coordination strategy, we consider simplified, deterministic kinematics and place no limitations on camera visibility. The objective is minimizing the average geolocation (target localization) error covariance associated with the fused measurements of the target's inertial position. A surprising result, and the main contribution of this work, is that the dominant factor governing the optimal UAV routes is coordination of the distances to the target, not of the viewing directions as is traditionally assumed. Lastly, we introduce a dynamic programming method that we hypothesize will allow us to revisit the original one-UAV stochastic optimal control problem and solve it for the two-UAV scenario with a cost function incorporating the fused geolocation error covariance.
Steven A. P. Quintero developed an interest in autonomous vehicles while pursuing his B.S. degree in Electrical Engineering from Embry-Riddle Aeronautical University. During his sophomore year, he was accepted into the McNair Scholars Program, a national initiative to increase graduate degree awards for students from underrepresented groups in society. Under the supervision of Dr. Gary Gear, he conducted two summer research internships at NASA's Dryden Flight Research Center. The first summer was spent developing a global range data acquisition system for aerial science missions and the second developing an integrated vehicle health monitoring system for small UAVs. He earned his B.S. degree from Embry-Riddle in 2007 and subsequently enrolled in the Ph.D. program at UCSB, where he continues his work with small UAVs under the direction of Professor Joao Hespanha. His research interests include coordinated control, probabilistic planning, i.e., robotic motion planning under uncertainty, and applied dynamic programming.
