Clothoid spline based path planning
Date of Issue2017
School of Mechanical and Aerospace Engineering
Institute for Media Innovation
Path planning for long vehicles plays an important role in safe and efficient operation, especially for those with very large lateral and longitudinal sizes driving in narrow spaces. To address this issue, we start with generalizing a kinematic bicycle model to represent different types of vehicles, based on which a path planning framework composed of a local path planner and a global path planner is developed. During the local planning phase, feasible path primitives are generated subject to kinematic constraints of the curvature maxima and the angular acceleration upper bound. By designing a family of path primitives with clothoid arcs involved which can be efficiently configured according to the given environment and reference points in physical space referred to as waypoints, we are able to obtain a near-optimal solution without numerically solving the two point boundary value problem to generate a feasible path connecting two states under differential constraints. During the global planning phase, the local path planner can be conveniently integrated with sampling based frameworks like the rapidly exploring random tree, which complies with the kinematic constraints when sampling the new states and expanding towards unexplored space. Meanwhile, an efficient clothoid approximation method is presented based on the minimization of curvature profile difference. Quintic Bézier curves are used to ensure higher order geometric continuity with smaller approximation error, and the data is stored in a lookup table to reduce the computing time. The developed methodology is novel and effective, which is capable of addressing a class of nonholonomic planning problems successfully subject to curvature and angular acceleration constraints.