Randomized path planning for redundant manipulators without inverse kinematics

摘要

We present a sampling-based path planning algorithm capable of efficiently generating solutions for high-dimensional manipulation problems involving challenging inverse kinematics and complex obstacles. Our algorithm extends the Rapidly-exploring Random Tree (RRT) algorithm to cope with goals that are specified in a subspace of the manipulator configuration space through which the search tree is being grown. Underspecified goals occur naturally in arm planning, where the final end effector position is crucial but the configuration of the rest of the arm is not. To achieve this, the algorithm bootstraps an optimal local controller based on the transpose of the Jacobian to a global RRT search. The resulting approach, known as Jacobian Transpose-directed Rapidly Exploring Random Trees (JT-RRTs), is able to combine the configuration space exploration of RRTs with a workspace goal bias to produce direct paths through complex environments extremely efficiently, without the need for any inverse kinematics. We compare our algorithm to a recently-developed competing approach and provide results from both simulation and a 7 degree-of-freedom robotic arm.