Skip to content

← ArchivePaper2010

Stable and Efficient Differential IK

J P Lewis, Nebojsa Dragosavac

SIGGRAPH AsiaWeta FXRigging

Differential IK formulation providing stable, efficient inverse kinematics for complex character rigs with redundant degrees of freedom.

Abstract

A differential inverse kinematics algorithm combining pseudoinverse and Jacobian transpose approaches achieves both convergence speed and stability in singular configurations. Near singular poses, the algorithm adaptively weights toward the Jacobian transpose to avoid the parameter instability of pure pseudoinverse while maintaining faster convergence than damped approaches. Singular value decomposition guides the adaptive blending based on how close the skeleton approaches singular configurations.

How to read this

Category
Method: a differential inverse-kinematics algorithm
Contributions
  • A differential IK formulation that adaptively blends pseudoinverse and Jacobian-transpose approaches for both convergence speed and stability
  • Weighting toward the Jacobian transpose near singular poses to avoid pseudoinverse parameter instability while staying faster than damped methods
  • Using singular value decomposition to guide the adaptive blend based on closeness to singular configurations
Context
Sits in the classic inverse-kinematics lineage for redundant articulated rigs, blending the well-known pseudoinverse and Jacobian-transpose families and using SVD as the switching signal.
Correctness
The central claim is stability through singular configurations via SVD-guided blending; a reader should note the trade-off is between convergence speed and the cost/robustness of the SVD-driven weighting, and the abstract states no specific rig or timing results.
Clarity
Moderately technical; a first pass conveys the adaptive-blend idea, but a second pass is needed to follow the SVD-based weighting formulation.
How to read it
Focus first on why the pseudoinverse becomes unstable near singularities and how the transpose blend fixes it; do a second pass on the SVD weighting if you intend to implement it.

Related work

Keywords

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →