A Hybrid Method for Near-Optimal Kinodynamic Planning

Kurzfassung

This thesis presents a novel nonlinear kinodynamic planning method including collision avoidance embedded in a global randomized search algorithm which can be applied to highly constrained orbital dynamics motion planning problems. The method is implemented in the C++ language as a software package called \texttt{p2pOpt}, for point-to-point optimization. \texttt{p2pOpt} combines four integral phases, physics simulation, virtual scene description, randomized sampling, and collision avoidance. The software has been tested on key benchmark problems including planar translation in a highly constrained, cluttered environment, 3DOF translation in orbit using the linearized Hill Dynamics and coupled, non-linear motion of a pendulum. It has been shown to be a fast, near-optimal alternative method to traditional randomized methods that is also capable of evaluating fully non-linear physical problems.