CoorGrasp: Coordinated Contact Control for Adaptive Dexterous Grasping under Uncertainty
Mingrui Yu, Yongpeng Jiang, Yongyi Jia, Yi Ren, Xiang LI
AI summary
Problem
Open-loop execution of planned dexterous grasps often causes undesired object movement or failure due to shape and position uncertainty. Existing tactile methods typically control fingers independently and lack multi-contact coordination.
Approach
A model predictive controller (MPC) using an analytical motion-contact model that coordinates arm-hand movements during approach and adaptively allocates contact forces based on wrench balance during grasping.
Key results
- Higher grasp success rates compared to baselines
- Reduced undesired in-hand object movement
- Validated via large-scale simulation with 15k grasps across 478 objects
- Successful real-world deployment on 8 everyday objects
Why it matters
It enables the reliable execution of diverse planned grasps by adapting to real-time tactile feedback, reducing reliance on perfect perception.
Abstract
While recent research has focused heavily on dexterous grasp pose generation, less attention has been devoted to the execution of planned grasps. Under shape and position uncertainty, open-loop execution often yields uncoordinated contacts, causing undesired in-hand object motion and even grasp failures. To address this, this paper proposes a tactile- driven model predictive controller for adaptive and delicate ex- ecution of diverse dexterous grasps. Our approach emphasizes multi-contact coordination across both approaching and grasp- ing phases, with three key novelties: (i) coordination-aware phase separation, (ii) arm–hand coordination to compensate for position errors, and (iii) adaptive force coordination to increase contact forces in a balanced manner. An analytical model is employed to relate contact forces to robot joint motions for predictive control. Our formulation imposes no restrictions on grasp types or contact configurations and integrates seamlessly with state-of-the-art grasp pose generation methods. We vali- date the approach through large-scale simulations involving 15k grasps across 478 objects on three robotic hands, and real-world experiments on 8 objects. Results demonstrate that our method achieves higher grasp success rates and reduced undesired object movements. Supplementary materials are available at https://ada-grasp-ctrl.github.io/.