Modular robotic grasping combines geometry and reactive control
Steerable and Reactive Grasping Through Modular Design with a Three-Point Interface
Robotics
Summary
Picking up objects with robot hands is tricky because the robot needs to choose where to grab, reach the right spot, and hold the item tightly. The authors designed a system using three key contact points that link big-picture planning with small-scale hand adjustments. Their method uses models and reinforcement learning to guide the robot hand precisely and keep a stable grip without needing complex sensors or detailed object info during the final approach. They tested the approach in simulations and real robots, showing it works well on new and known objects.
What this means in practice
- •For robotic system builders: Create robotic hands that use geometric planning combined with local finger feedback to grasp diverse objects with increased reliability.
- •For industrial automation teams: Improve robotic pick-and-place tasks by integrating reactive grasp adjustments that reduce errors from perception or reaching inaccuracies.
Authors
Andrew Nguyen, Yonghyeon Lee, Sangbae Kim
Abstract
Dexterous grasping requires deciding where to grasp, reaching the target, and maintaining stable contact. We connect these stages through a compact three-point interface that separates global geometric reasoning from local contact control. Given object geometry and optional language commands, our framework samples contact triples from a precomputed grasp-affordance heatmap. A model-based reactive controller tracks the object, avoids collisions, and guides the hand toward the selected contacts. In the final centimeters, a Reinforcement Learning (RL) policy uses proprioceptive feedback to refine and stabilize the grasp despite reaching and perception errors. It observes only finger joint states and its recent actions, with no target points, visual observations, or object geometry, so a single policy is shared across objects and grasp configurations. In simulation, we compare grasp-and-lift success against squeeze and end-to-end baselines, characterize reaching convergence, and demonstrate grasp steering; hardware demonstrations on two training objects and one unseen object illustrate the full pipeline. Our modular framework uses geometry to guide the reach and local feedback to secure the grasp.