| |
ABSTRACT
This paper presents a mobile robot planner that attempts to find minimum
time path between two configurations on uneven outdoor terrain. The
computed path is smooth, safe, and satisfies dynamic constraints. The
terrain is modeled as a geometric surface, the robot is Segway Robotic
Mobile Platform (RMP). It uses bidirectional rapidly exploring random
trees (RRT's) on the terrain surface with nodes connected by continuous
curvature trajectories. In standard RRT nodes are usually expanded by
applying controls for a fixed duration. In our RRT implementation, the
current path is extended with a short trajectory up to a specified maximum
length. The trajectory parametrization is such that different sets of
controls and control durations can be fitted to produce the same curve.
Thus, the robot velocity profile can be readjusted "back in time" along
previous RRT edges. An optimization algorithm is used to compute globally
optimal velocity profile resulting in near minium time path while
satisfying the dynamic constraints. We apply these methods to an
interesting problem - robot finding its quickest path on a rough terrain
while maintaining the safety of its load and itself.
|
