Скачать или смотреть RRTX: Dubins Vehicle in R^2 x S Dynamic Environment

Dubins robot using the RRTX algorithm for real-time navigation to a goal in a dynamic environment. This video shows the true speed of the robot and all planning and replanning calculations as they happen online.

The state space is a subset of R^2 x S. The robot is provided with an inaccurate map of the world and must adjust its motion plan when it discovers that obstacles have appeared or disappeared.

The robot moves at 20 m/s, it has a sensor range of 10 m (and thus a maximum reaction time of 0.5 s), and a minimum turning radius of 2 m. The robot is allowed to plan for 10 s before movement starts, but must react on-the-fly to obstacle changes.

Obstacles are black with white outlines. Color depicts distance-to-goal, the goal is a white square, the robot is the black aircraft within the white circle. Planned motion is the white path, the path already traveled by the robot is red. The search tree is light gray (although the edges of the tree are drawn as straight lines and not the corresponding Dubins trajectories along which the robot will actually move). Sensor range black/white-dashed circle. Cost-to-goal is grayed out (dark gray) at locations further from the goal than the robot.

The RRTX algorithm uses a single search tree for the duration of navigation, repairing it vs. obstacle changes. The tree is built directly in the state space of the robot, and the resulting paths approach optimality, in the limit, as planning time approaches infinity.