Guided dynamic window approach to collision avoidance in troublesome scenarios

The paper presents a new method named guided dynamic window approach(GDWA) for non-holonomic vehicle navigation in dense, cluttered, and complex scenarios. First, the local information including the local objective navigable region and relatively local orientation is obtained using nearness diagram from the central point(PND). Second, dynamics constraint, safety constraint and local navigable region constraint are applied to the velocity space. Finally, the robot chooses the velocity commands that satisfy all the constraints and maximize an objective function that trades off speed, clearance and weighted local orientation based on the local navigable region. With explicit analysis of connectivity of the free space in local environments and the ability to follow the local objective navigable region, the method can be used for mobile robots navigation in troublesome scenarios. Navigation experiments of a real vehicle with sonar ring validate this research.