Design Paradigms Based on Spring Agonists for Underactuated Robot Hands: Concepts and Application

Tianjian Chen, Tianyi Zhang, Matei T. Ciocarlie

Introduction

Underactuated robotic hands offer a compelling solution for complex grasping tasks using a minimal number of actuators. Traditionally, these systems employ tendons as "agonists" to pull fingers closed and simple springs as "antagonists" to return them to an open state. However, this standard approach suffers from significant energy inefficiency—as motors must constantly overcome return spring tension—and a lack of fail-safety, where power loss results in the immediate release of a grasped object. Researchers at Columbia University have proposed a shift in this design philosophy by utilizing "Springs as Agonists" (SA), where springs serve as the primary closing mechanism.

The "Springs as Agonists" Paradigm

The core innovation lies in a design framework that categorizes underactuated hands based on agonist choice and coupling methods. While traditional Tendons as Agonists (TA) are effective for high-force flexion, the SA paradigm is particularly suited for maintaining static grips without continuous power consumption. By integrating these concepts, the team developed a hybrid hand optimized for NASA’s Astrobee robot, designed for operation on the International Space Station (ISS).

The proposed underactuated hand design, showing the internal tendon routing (left) and the physical 3D-printed prototype (right).

Mechanical Implementation and Force Cancellation

To overcome the "spring tax"—the torque wasted fighting return springs—the researchers implemented a technique called Spring Force Cancellation. By winding flexion and abduction tendon systems in opposite directions on a single motor shaft, the opposing spring forces effectively neutralize each other at the actuator level. This configuration utilizes a combination of traditional tendon-driven flexion (TA+MJT) and spring-driven adduction balanced against a motor-driven antagonist (SA+MTS).

Diagram illustrating the coupling of multiple tendons and springs on a single motor shaft to balance forces and reduce motor load.

This resulting system allows the internal friction (stiction) of the motor's gearbox to hold the hand in any position. This enables "power-off perching," where the robot can maintain a firm grip on a handrail indefinitely without drawing electrical power. Furthermore, in the SA+MTS configuration, if one finger's movement is blocked, the tendon simply goes slack, allowing the motor to continue moving the other fingers—a mechanical form of adaptive grasping.

Performance and Results

A physical prototype, driven by a single Dynamixel servo, demonstrated the practical utility of this approach. Kinematic optimization was used to determine ideal pulley radii and spring constants, ensuring synergistic finger motion. Key findings include:

  • Torque Efficiency: The resultant torque on the motor shaft remained below the gearbox's stiction limit (84 Nmm) throughout the entire range of motion.
  • Fail-Safe Grasping: The hand successfully maintained its grip on an ISS-style handrail even when completely disconnected from power.
  • Human-Robot Interaction: The balanced force profile allows humans to manually adjust the finger positions while the system is powered off without risking damage to the internal mechanics.

Conclusion and Future Directions

The "Springs as Agonists" framework provides a robust path toward energy-efficient and fail-safe robotic manipulation. While the current design excels at static holding, the researchers noted that grip strength is largely limited by spring stiffness, making active force regulation difficult. Future work will address these challenges by introducing series compliance to allow for better force control and evaluating the hand's performance in simulated microgravity environments.

Summary updated

BibTeX
@inproceedings{chen2021design,
  author = {Tianjian Chen and Tianyi Zhang and Matei T. Ciocarlie},
  title = {{Design Paradigms Based on Spring Agonists for Underactuated Robot Hands: Concepts and Application}},
  booktitle = {IEEE International Conference on Robotics and Automation},
  year = {2021},
  doi = {10.1109/icra48506.2021.9561832},
}
Figure