A Reconfigurable Dual-Opposition Architecture for Single-Hand Assembly and Manipulation

William Su; Yunosuke Nakamura; Yixiao Wang; Yitong Li; Mingrui Yu; Huanan Qi; Boyuan Liang; Masayoshi Tomizuka

cs.ROSep 19, 2026

A Reconfigurable Dual-Opposition Architecture for Single-Hand Assembly and Manipulation

Authors: William Su, Yunosuke Nakamura, Yixiao Wang, Yitong Li, Mingrui Yu, Huanan Qi, Boyuan Liang, Masayoshi Tomizuka, +1 more

Organizations: Aerospace Engineering Program, University of California, Berkeley, Berkeley, CA 94720, USA. · Department of Mechanical Engineering, University of California, Berkeley, Berkeley, CA 94720, USA. · Department of Mechanical Engineering, National University of Singapore, Singapore 117575.

Abstract

In-hand assembly is constrained by the need to maintain grasps on two separate parts while controlling their relative motion within a single hand. To enable both in-hand assembly and manipulation, we present a reconfigurable dual-opposition architecture. Specifically, to support simultaneous grasping of two parts and coordinated in-hand manipulation, four independently actuated fingers are organized into two virtual finger (VF) oppositions, with their relative configuration controlled by a reconfigurable palm. To describe hand motion and simultaneous two-object grasping configurations, a kinematic model of the fingers and palm and an object-size-conditioned workspace formulation are built. To further evaluate motion performance and assembly capability, finger-joint motion and palm tracking are characterized, and in-hand assembly is demonstrated through tasks involving grasping, alignment, fastening, and pressing. Ablation experiments further demonstrate the importance of finger abduction/adduction and palm reconfiguration for successful in-hand assembly. In simulation, the proposed hand achieves a mean continuous sphere rotation success rate of 98.6% over diameters of 40-230 mm, compared with 73.8% for the LEAP Hand. After policy fine-tuning with external disturbances, the proposed hand achieves 92.8% success under disturbances from multiple directions, compared with 45.2% for the LEAP Hand. Hardware demonstrations further show in-hand rotation of objects of different sizes using policies trained in simulation. Together, these results show that the proposed architecture supports both assembly of two separately held parts and coordinated manipulation of a single object within one hand.

Figures & tables

Fig. 1: Grasp organization in bimanual and single-hand configurations. (a) Two anthropomorphic hands form the oppositions VF A $\leftrightarrow$ VF C and VF B $\leftrightarrow$ VF D, respectively. (b) The proposed hand realizes this grasp organization with two VF oppositions integrated within one hand. Palm translation adjusts their relative position. (c) Schematic mapping from inter-hand coordination to intra-hand reconfiguration.

Fig. 2: Mechanical design of the proposed 17-DOF reconfigurable hand. (a) The proposed hand mounted on a FANUC CRX-10iA robot arm. (b) Reconfigurable palm mechanism with potentiometer-based position feedback. (c) Finger kinematics showing the four degrees of freedom of each finger, including MCP flexion, PIP flexion, DIP flexion, and MCP abduction/adduction, together with their corresponding motion ranges. (d) Exploded view of the hand.

Capability	Parallel Gripper	Dexterous Hand	Proposed Hand
VF oppositions	Single	Typically single	Dual
Local dexterity	Low	High	High
Two-object adjustment	Limited	Limited	Supported
Relative object motion	Limited	Limited	Supported
In-hand assembly	Limited	Limited	Supported

TABLE I: Comparison of functional capabilities among conventional parallel grippers, dexterous hands, and the proposed hand.

Fig. 3: Sampled workspace for simultaneous two-object grasping. Each accepted hand configuration contributes one sphere center for each VF opposition, with color indicating sphere diameter. The top view shows the largest equal sphere diameter accepted in each spatial bin.

Fig. 4: Hardware characterization of the integrated hand. (a) Commanded joint ranges and encoder measurements; error bars indicate endpoint standard deviations over five cycles. (b) Commanded angle, encoder feedback, tracking error, and motor current for the MCP abduction/adduction joint of finger A. (c) Commanded palm position, potentiometer feedback, and tracking error during extension and retraction.

Fig. 5: Teleoperated assembly demonstrations using the VF A $\leftrightarrow$ VF C and VF B $\leftrightarrow$ VF D oppositions. Each row shows a task sequence progressing from left to right. (a) Through-wall fastening, with one opposition holding the nut while the other drives the screw from the opposite side. (b) Simultaneously fastening two screws at different spacings using palm reconfiguration. (c) Aligning and fitting a cap onto a glue-stick body. (d) Actuating a syringe by holding the barrel with one opposition and driving the plunger with the other. (e) Routing a cable through an opening with grasp transfer between the two virtual-finger oppositions.

Task	Full hand	MCP abduction/ adduction locked	Fixed palm
Through-wall fastening	$\checkmark$	$\checkmark$	$\times$
Simultaneous screw fastening	$\checkmark$	$\times$	$\times$
Glue-stick capping	$\checkmark$	$\times$	$\times$
Syringe actuation	$\checkmark$	$\checkmark$	$\times$
Cable routing	$\checkmark$	$\checkmark$	$\checkmark$

TABLE II: Assembly task outcomes for the full hand and two ablated configurations.

\checkmark

: success;

\times

: failure.

Fig. 6: In-hand object rotation in simulation and on hardware. Photographs show hardware experiments with a practice golf ball, a pickleball, and a volleyball. Large markers show mean simulation success rates across three training seeds; small markers show results for each seed. Error bars indicate 95% hierarchical bootstrap confidence intervals.

Fig. 7: Response to an applied force with a peak magnitude of $0.75~\mathrm{N}$ and a duration of $0.25~\mathrm{s}$ during rotation of a $100~\mathrm{mm}$ sphere. (a) Success rate for continuous sphere rotation before and after fine-tuning under disturbances from eight directions, with 32 rollouts per direction per seed across three training seeds of the original policies. (b), (c) Keyframes from a selected rollout at $225^{\circ}$ using the original and fine-tuned policies, respectively. The top and bottom rows show the proposed hand and the LEAP Hand, respectively. Frames show the motion before, during, and after the disturbance. Blue traces show the sphere COM trajectory over the preceding $2.0~\mathrm{s}$ ; red arrows indicate the applied force.

Explore similar work

Assembling Two Parts in One Hand
Sep 9, 2026Liuao Pei, Tianyue Wu, Hui Zhang +2RL for Robotics Dexterous Manipulation
The University of Hong Kong · The Hong Kong University of Science and Technology (Guangzhou) · ETH Zurich +1
A Kinematic Analysis of Palm Degrees of Freedom for Enhancing Thumb Opposability in Robotic Hands
Apr 24, 2026HyoJae Kang, Yeong Jae Park, Hyunmok Jung +2Robotic Hand Design Dexterous Manipulation
Advanced Robotics Research Center, Korea Institute of Machinery & Materials (KIMM)
A Novel Reconfigurable Dexterous Hand Based on Triple-Symmetric Bricard Parallel Mechanism
Mar 1, 2026Chunxu Tian, Zhichao Huang, Hongzeng Li +4Robotic Hand Design Robotic Grasping
Institute of AI and Robotics, Academy for Engineering & Technology, Fudan University, Shanghai 200433, China. · Department of Mechanical Engineering, The Hong Kong Polytechnic University, Hung Hom, Hong Kong.

Sep 9, 2026J/K move · Enter open · S save

Assembling Two Parts in One Hand

Liuao Pei, Tianyue Wu, Hui Zhang, Ping Luo+1

The University of Hong Kong · The Hong Kong University of Science and Technology (Guangzhou) · ETH Zurich+1

RL for Robotics Dexterous Manipulation In-Hand Manipulation Robotic Assembly

Open paper AskPDF arXiv

A hallmark of human dexterity is the cooperative use of fingers, where different fingers take on distinct yet coordinated roles to accomplish fine manipu- lation, such as capping a pen with the hand that holds it. We study this finger-level coordination through in-hand assembly: mating two rigid objects within a single dexterous hand, with no second arm and no fixture. We present a reinforcement learning formulation to solve this problem in a unified framework, which is driven by a goal relative pose between the two parts. Finger coordination is shaped by a function-based auxiliary reward and regularized toward a single human reference pose, while domain randomization and a fusion of historical proprioception and object observation confer robustness to occlusion-induced estimation noise. The same recipe solves three different assembly tasks (Bottle, Syringe, and Marker). Trained purely in simulation, the policies transfer zero-shot to hardware with a single camera, demonstrating robustness to state-estimation errors caused by oc- clusion. Our experiments also reveal that in-hand assembly places demands on hand morphology and can serve as a benchmark for modern robotic hand systems. Videos and code are available at https://ltbgbird.github.io/in-hand-assembly-page/.

Abstract

Explore similar work

Assembling Two Parts in One Hand

A Kinematic Analysis of Palm Degrees of Freedom for Enhancing Thumb Opposability in Robotic Hands

A Novel Reconfigurable Dexterous Hand Based on Triple-Symmetric Bricard Parallel Mechanism