Constrained path optimization with Bézier curve primitives

Abstract

This article suggests a novel path planning algorithm for a non-holonomic wheeled vehicle operating in a semi-structured environment. The first step of the algorithm is to compute offline a finite set of feasible motions connecting discrete robot states to construct a search graph. The motion primitives based on Bézier curves are generated by solving the constrained optimization problem (COP). Applying A* on the search graph produces paths in the form of a sequence of the primitives. Although the sequence is drivable and suboptimal, we perform online path smoothing to remove the jerky or extraneous motions. The post-procedure is done by using a gradient based method to solve another COP. The numerical simulations show remarkable performance improvement in the constrained path optimization by applying our algorithm, compared to other existing works. Also, field experimental results demonstrate successful generation of fast and safe trajectories for real-time autonomous driving.