A fast workspace-density-driven inverse kinematics method for hyper-redundant manipulators

Abstract

Hyper-redundant manipulators have a very large number of redundant degrees of freedom. They have been recognized as a means to improve manipulator performance in complex and unstructured environments. However, the high degree of redundancy also poses new challenges when performing inverse kinematics calculations. Prior works have shown that the workspace density (generated by sampling the joint space and evaluating the frequency of occurrence of the resulting end-effector reference frames) is a valuable quantity for use in ${\cal O}(P)$ inverse kinematics algorithms. Here $P$ is the number of modules in a manipulator constructed of a serial cascade of modules. This paper develops a new “divide-and-conquer” method for inverse kinematics using the workspace density. This method does not involve a high-dimensional Jacobian matrix and offers high accuracy. And its computational complexity is only ${\cal O}({\rm log}_2\,P)$, which makes it ideal for very high degree-of-freedom systems. Numerical simulations are performed to demonstrate this new method on a planar example, and a detailed comparison with a breadth-first search method is conducted.