Whole-Body Control and Estimation of Humanoid Robots with Link Flexibility

Overview

Real humanoid robots are not perfectly rigid: structural link flexibility introduces deflections that degrade both state estimation and control. This paper studies how to model that flexibility and account for it inside the whole-body estimation and control pipeline.

Method

• Flexibility model. Compliant links are represented so that their deflection can be estimated online rather than ignored.
• Estimation. The estimator reconstructs the robot state while explicitly considering the flexible degrees of freedom.
• Whole-body control. The controller uses the augmented model to produce commands that remain accurate in the presence of link compliance.

Results

Accounting for link flexibility improves the consistency between the model and the real robot, yielding better estimation and more accurate whole-body control on a humanoid platform.