Simultaneous Tracking of Multiple Ground Vehicles from a Single Mulitrotor UAV
Nathan Miller · OakTrust (Texas A&M University Libraries) · 2014
An algorithm for autonomous multirotor tracking of an arbitrary number of ground targets is formulated and evaluated through simulation. The algorithm consists of a particle filter to predict target motion, trajectory generator, and model predictive controller operating over a finite time horizon. Furthermore, a target selection algorithm is included to reject targets that preclude accurate tracking of the main target set. Performance of the guidance algorithm is evaluated through simulation using real-world target data. Simulation results show that targets can be kept within the camera’s field of view when using either a gimbaled or non-gimbaled camera, but performance may be substantially degraded in the non-gimbaled case when target dynamics occur on time scales similar to tracking vehicle dynamics.