Jerk bounded velocity planner for the online management of autonomous vehicles

The velocity planner for autonomous vehicles proposed in this paper is designed for the automatic generation of trajectories characterized by bounded velocities, accelerations and jerks. Differently from similar planners, velocity bounds are not constant but, more realistically, they depend on the path shape. The problem poses feasibility issues which are analyzed in the paper: conditions for the existence of a solution are proposed. A procedure is then designed for the synthesis of a feasible profile, that is subsequently used for the initialization of a minimum-time nonlinear programming algorithm. Particular attention has been given to the algorithmic efficiency, since the planner is conceived for the management of real-time planning scenarios, where reactivity to unforeseen events represents an important requirement. The algorithm has been tested by designing the velocity functions for a set of 50 randomly chosen paths. Performance comparisons are made with alternative planning schemes.