This paper presents a novel optimization-based approach to compute time-optimal trajectories for robotic systems operating in an environment with the presence of obstacles under kinodynamic constraints. The proposed approach employs a modified rapid exploring random tree algorithm (RRT) to generate a geometrical sub-optimal path inside a feasible safe region. Subsequently, a trajectory is parametrized by fourth order non-uniform B-splines and is optimized along the path with respect to kinodynamic constraints by an interior point optimizer. The optimization process is performed in the safe region without any further collision checking, which is very effective in extremely confined and complex environments. Finally, the potential and efficiency of the approach is illustrated and compared with the notable RRT* algorithm in state space by numerical simulations.
相似文献