Motion Planning With Flexible Trajectory Chains

Abstract

This paper presents a novel methodology for motion planning. The basic idea is to model vehicle trajectories as chains with adjustable joints and links. The characteristic parameters of these chains are chosen in accordance with the limitations due to driving dynamics. A key feature of the approach is that trajectories can be modified in a fast and intuitive manner, e.g., in order to avoid suddenly emerging obstacles. It is possible to simultaneously generate trajectories of multiple dynamic objects, taking into account interactions and dependencies between the objects. The trajectory chain model is extended by continuous collision detection, which allows collisions along a chain link to be covered with a single collision query. The simple structure of the chain model and the ability to react to dynamic objects offers advantages in applications in which a multitude of trajectories have to be calculated in a dynamic environment. The method was successfully implemented and validated by simulations. Its functionality is shown in collision avoidance scenarios.