Robust Autonomous Manipulation of Articulated Objects

This study proposes a new hybrid control framework for reliably autonomously manipulating articulated objects, in particular, the challenging case of doors, drawers and multi-jointed mechanisms in an uncertain environment.The novelty stems from an integration of articulation-aware perception, physics-grounded modelling of object internal degree of freedom, and adaptive planning that explicitly reasons about joint limits and contact dynamics.Contrary to previous studies that apply Control Barrier Functions (CBFs) on generic robotic systems, the study introduces articulationspecialized CBFs to predict and prevent joint limit violations based on object kinematics.The framework consists of modelbased planning and a real-time quadratic programming (QP) safety filter that enforces zeroing control barrier functions ZCBFs energy-based passivity constraints.Tests run on three types of articulated objects see success rates over 95% while still satisfying constraints when compared to non-attached rigid body methods for articulated objects which is a big improvement.