Trajectory generation for car-like robots.

摘要

Autonomous robots or remotely operated vehicles have raised high hopes in the military and industrial communities because of the potential safety improvement and gain of productivity they may provide. Waste management on nuclear sites, pallet manipulation in factories, interventions on battle-fields, etc., are actively studied. A lot of these applications require powerful four-wheel vehicles, the kinematics of which is similar to that of a car. Such vehicles have three degrees of freedom: the (x,y) positions in a plane and the orientation of the vehicle. Path planning is often understood as only changing the position of the vehicle, whereas the tasks performed by this kind of robot requires a perfect orientation of the vehicle: forklifting a pallet or docking at a loading or unloading station requires accuracy in the orientation of the vehicle. It is this requirement and the kinematic constraints of the motion mode which have led to the path-planning algorithm presented in this paper. The velocity of the robot belongs to a two-dimensional vectorial space. However, we assume that there is no slipping of the wheels. Therefore, at a given position, the direction of the velocity of the rear axle, is colinear with that of the vehicle. The equation conveying this constraint is not integrable and affects the velocity but not the space of the configurations of the robot: it is a non-holonomic constraint. If the steering angle of the front wheels is constant, the vehicle moves along a circle. Since the steering angle of the car-like robots is limited, the radius of the circle is always greater than a certain value which is the minimum radius of curvature of any achievable trajectory. 3 refs., 8 figs.