Contact-Rich Robotic Manipulation

Overview

This field studies how robots make and maintain useful contact with objects, where small errors in force, alignment, or friction can cause failure or damage. Approaches include tactile- and force-aware imitation learning and reinforcement learning, compliant and hybrid force-position control, contact-dynamics models, and planning methods that reason over contact modes. Recent work also brings these capabilities into vision-language-action policies and predictive world models, often using tactile feedback for fast corrections while vision guides the broader task. A central challenge is making policies safe and reliable across unseen objects, sensors, and real-world contact conditions, since more sensing or higher simulation fidelity does not consistently improve every task.

Latest papers

CardsList

454 papers44 papers in the last four weeks, down 55% on the four weeks before. 0.3% of all new papers.

  1. Periodic robust robotic rock chop via virtual model control

    Aug 4, 2025Yi Zhang, Fumiya Iida, Fulvio ForniContact-Rich Robotic ManipulationRobust Control

  2. EquiContact: A Hierarchical SE(3) Vision-to-Force Equivariant Framework for Spatially Generalizable Contact-rich Tasks

    Jul 15, 2025Joohwan Seo, Arvind Kruthiventy, Soomi Lee +5Contact-Rich Robotic ManipulationRobot Learning

  3. Passivity-Centric Safe Reinforcement Learning for Contact-Rich Robotic Tasks

    Mar 1, 2025Heng Zhang, Gokhan Solak, Sebastian Hjorth +1Robotic ControlContact-Rich Robotic Manipulation