IKFlow: Generating Diverse Inverse Kinematics Solutions

Journal Article (Journal Article)

Inverse kinematics - finding joint poses that reach a given Cartesian-space end-effector pose - is a fundamental operation in robotics, since goals and waypoints are typically defined in Cartesian space, but robots must be controlled in joint space. However, existing inverse kinematics solvers return a single solution, in contrast, systems with more than 6 degrees of freedom support infinitely many such solutions, which can be useful in the presence of constraints, pose preferences, or obstacles. We introduce a method that uses a deep neural network to learn to generate a diverse set of samples from the solution space of such kinematic chains. The resulting samples can be generated quickly (2000 solutions in under 10 ms) and accurately (to within 10 millimeters and 2 degrees of an exact solution) and can be rapidly refined by classical methods if necessary.

Full Text

Duke Authors

Cited Authors

  • Ames, B; Morgan, J; Konidaris, G

Published Date

  • July 1, 2022

Published In

Volume / Issue

  • 7 / 3

Start / End Page

  • 7177 - 7184

Electronic International Standard Serial Number (EISSN)

  • 2377-3766

Digital Object Identifier (DOI)

  • 10.1109/LRA.2022.3181374

Citation Source

  • Scopus