This thesis presents the coordination of an unmanned, multi-vehicle team that navigates through a congested environment. A novel approach is outlined that enables the control of multiple vehicles based on both computer vision and optimal trajectory algorithms. Various sensors are used to achieve localization in the indoor environment in lieu of global positioning data. Specifically, a Quanser Qball quadrotor is equipped with a downward-looking camera and sonar altimeter, while a Quanser Qbot ground vehicle is outfitted with sonar and infrared range finders. This equipment is complemented by an Optitrack motion-capture system. Using conventional image-processing techniques, the bird's-eye images supplied by the quadrotor provide information regarding the dynamic environment that surrounds the ground vehicle. The ground vehicle can then produce a global, optimal trajectory, assuring collision-free operations. The optimization problem is addressed by applying the Inverse Dynamics in the Virtual Domain (IDVD) method that uses both the inverse kinematics of the ground vehicle and obstacle information. Furthermore, the IDVD method enables the separation of spatial and temporal planning. As verification of the results of this research, the developed approach for path planning is executed in a fully controlled lab environment and then compared with a sonar-based, reactive obstacle avoidance technique. 14. SUBJECT TERMS: Quadrotor, Ground Vehicle.
Global Versus Reactive Navigation for Joint UAV-UGV Missions in a Cluttered Environment
2012
101 pages
Report
No indication
English
IMITATION LEARNING FOR UAS NAVIGATION IN CLUTTERED ENVIRONMENTS
TIBKAT | 2021
|Tracking multitarget in cluttered environment
IEEE | 1992
|