cs.RONov 3, 2026

PACT: Pose-Robust Actuation Co-Design for Tendon-Driven Robotic Hands

Authors: Prashant Kumar, Rachael Granberry, Taylor Niehues, Andrew A. Stanley, Gabe Pirie

Organizations: Meta Reality Labs Research

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Humanoid robotic hands must provide sufficient force, speed, and backdrivability within a human-scale envelope. Tendon-driven transmissions address this challenge by relocating actuators to the palm and forearm, but routing paths, moment arms, and capstan dimensions jointly determine both force capability and joint speed, so selecting them sequentially satisfies one requirement at the expense of the other. This paper presents a pose-robust actuation co-design for tendon-driven robot hands (PACT) framework that optimizes tendon routing, moment arms, and capstan dimensions under pose-dependent force requirements and drivetrain-bandwidth constraints. A feasible-force formulation enforces actuator bounds and equilibrium at all joints, while a max-min objective maximizes task-force coverage at the weakest pose. Applied to a four-degree-of-freedom finger, the framework produces an asymmetric N + 1 tendon topology using quasi-direct-drive actuators. Across eight poses, the design achieves task-force coverage of 0.88-1.00 and a predicted 5 Hz minimum isolated-joint flexion/extension bandwidth within a 70th-percentile human-size envelope. Prototype measurements deviate from predictions by up to 24.5%, primarily due to unmodeled transmission losses. These results indicate that co-designing routing with the drivetrain can meet pose-dependent force requirements without relaxing the bandwidth constraint, paving the way for human-sized robotic hands with human-level performance.

Figures & tables

Fig. 1: Pose-robust actuation co-design for tendon-driven robot hands (PACT) framework. The framework combines a force-pose library, task-specific bandwidth constraints, and actuator properties to jointly optimize tendon-driven robot hands. It shapes feasible force sets across representative poses, maximizes the workspace coverage under worst-joint bandwidth limits, and selects design parameters that balance force capability, motion speed, and pose robustness. The resulting parameters define the tendon-routing and actuation configuration of the optimized hand.
Fig. 2: Actuation layout of the four-DoF finger. Flexion/extension uses an N+1N+1 tendon arrangement, while MCP abduction/adduction is actuated separately.
PoseMCP A/AMCP F/EPIPDIP
Pose 1: Small tip pinch0606030
Pose 2: Medium pad pinch0454520
Pose 3: Large pad pinch0253015
Pose 4: Medium tripod0405025
Pose 5: Small tripod0556030
Pose 6: Extended reach0152010
TABLE I: Index-finger task poses in degrees.
ParameterLow-torque QDDHigh-torque QDD
MotorGB2204Robstride QDD-05 [ 20 ]
Gear ratio5:15{:}17.75:17.75{:}1
Motor Rotor inertia ( kg m2\mathrm{kg\,m^{2}} )6.8×10−76.8\times 10^{-7}3.0×10−63.0\times 10^{-6}
Continuous torque ( N m\mathrm{N\,m} )0.141.14
Peak torque ( N m\mathrm{N\,m} )0.405.4
Output speed ( rpm\mathrm{rpm} )512480
TABLE II: Actuator parameters used in the optimization. All torque and speed values are specified at the gearbox output.
Fig. 3: Flexion/extension-plane cross-sections of the feasible force polytopes for the optimized shared design across all eight poses.
QuantityBaselineIndependentShared
Minimum reported score0.241.000.88
Maximum reported score0.691.001.00
Minimum F/E bandwidth (Hz)5.05.05.0
TABLE III: Results across the eight-pose task library.
Fig. 4: Cross-pose performance of the Pose 1-specific design compared with the shared design.
Fig. 5: Force sets at Pose 1: (a) baseline, (b) pose-specific design, and (c) shared design.
Joint Moment Arm(mm)m1m_{1}m2m_{2}m3m_{3}MM
MCP A/A (fixed during Optimization )+3.0+3.0−3.0-3.00.00.00.00.0
MCP F/E+6.9+6.9−3.0-3.0−8.0-8.0+5.6+5.6
PIP F/E−3.4-3.4−3.0-3.00.00.0+4.6+4.6
DIP F/E−6.9-6.90.00.00.00.0+3.1+3.1
Capstan diameter12.712.711.011.013.013.036.036.0
TABLE IV: Optimized shared-design parameters.
Fig. 6: Prototype and experimental setup with fixed finger base and six-axis fingertip load cell
Fig. 7: Predicted and measured fingertip forces at qtest\mathbf{q}_{\mathrm{test}} . The blue polytope represents the predicted force set at the experimental actuator limits. Circles indicate predicted forces, and diamonds indicate measurements.
More in this topic

Robotic Hand Design

Robotic hand design develops mechanical structures, actuation, and sensing for grasping and dexterous manipulation across real-world tasks.

See all 71 papers

Explore similar work

CardsList

  1. H-PAC Hand: Control-Oriented Modeling and Tendon-Elasticity Compensation for an Underactuated Robotic Hand

    Aug 17, 2026Teng Yan, Jiongxu Chen, Teng Wang +5Robotic Hand DesignTendon-Driven Actuation

    The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China · Southeast University, Nanjing, China

  2. Tendon-Actuated Robots with a Tapered, Flexible Polymer Backbone: Design, Fabrication, and Modeling

    Mar 19, 2026Harald Minde Hansen, Nandita Gallacher, Nicholas B. Andrews +3Continuum Robot ControlTendon-Driven Actuation

    Mechatronics, Robotics and Operation Section, European Organization for Nuclear Research (CERN), Meyrin, Switzerland · Department of Aeronautics and Astronautics, University of Washington, Seattle, WA, USA · Department of Engineering Cybernetics, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

  3. Design and Validation of an Antagonistic Tendon-Driven Dexterous Robotic Hand with Bidirectional Operation

    Sep 28, 2026Chunghyeon Lee, Hyukjun Kwon, Sungeon Kim +2Robotic Hand DesignRobotic Grasping

    Department of Mechanical Engineering, Texas A&M University, College Station, TX 77843, USA · Department of Mechanical Engineering, Sogang University, Seoul 04107, South Korea