Multi fingered robot hands use motion patterns for better object control
Primitive-Informed Sampling-Based MPC for Multi-Fingered Dexterous Manipulation
Robotics
Summary
Controlling robot hands with many fingers is tricky because moving each joint individually is slow and inefficient. The authors introduced a way to guide hand movements using simple building blocks of finger motions called primitives, which help the robot coordinate better during tasks like rotating or grasping objects. Their method also checks action plans to throw away impossible moves early, making the process more efficient. Tests showed the approach works well even when the robot handles different sized objects or when the model isn’t perfect.
What this means in practice
- •For robotics engineers: Improve control strategies for multi-fingered robot hands to perform reliable in-hand object rotations and grasps.
- •For industrial automation teams: Enable robotic manipulation systems to adaptively handle various object sizes and orientations during assembly or packaging.
Authors
Emek Barış Küçüktabak, Karankumar Patel, Jinda Cui, Zhaodong Yang, Kazuhiro Sasabuchi, Jun Takamatsu
Abstract
We present a primitive-informed sampling-based model predictive control (MPC) framework for multi-fingered dexterous manipulation. Sampling-based MPC avoids the need for gradients through complex contact dynamics, but direct exploration of the high-dimensional joint space is inefficient and makes performance strongly dependent on the sampling distribution. Our framework biases sampling using low-dimensional manipulation primitives that encode coordinated finger motions, while simultaneously optimizing joint-level residuals to adapt these motions to the current hand-object configuration. Task-related rollout constraints reject infeasible trajectories during forward simulation, improving the effective use of the sampling budget. We evaluate the approach on a physical Allegro hand using a synchronized MuJoCo digital twin. Ablations show that both the primitive and residual are necessary for reliable continuous in-hand rotation, that increasing the sampling budget alone does not recover this coordination, and that rollout constraints substantially improve success rate. A primitive extracted from a single object remains effective across object sizes and under model mismatch, and the framework further supports grasping, object reorientation, and coordinated arm-hand reach-grasp-transport, using primitives extracted from both a simulation-trained policy and human hand-motion data.