This paper addresses the problem of finding a nonholonomic
path subject to a curvature restriction, to be tracked by
a wheeled autonomous navigation vehicle. This robot is able to
navigate in a structured environment, with obstacles modeled as polygons,
thus constituting a model based system. The path planning methodology
begins with the conditioning of the polygonal environment by offsetting
each polygon in order to avoid the possibility of collision
with the mobile. Next, the modified polygonal environment is used
to compute a preliminary shortest path (PA) between the two
extreme positions of the trajectory in the plane (x, y). This
preliminary path (PA) does not yet consider the restrictions on
the curvature and is formed only by straight line segments.
A smoothing process follows in order to obtain a path
(PS) that satisfies curvature restrictions which consist basically of joining
the straight line segments by circular arcs of minimum radius
R (filleting). Finally, the initial and final orientation of the
vehicle are accounted for. This is done using a technique
we have called the Star Algorithm, because of the geometric
shape of the resulting maneuvers. A final complete path (PC)
is thus obtained.