Biologically informed geometry and force distribution improve task performance in agonist/antagonist tendon-driven prosthetic hands

The study demonstrates that the HAMR framework, which generates customized agonist/antagonist tendon-driven prosthetic hands based on individual morphology and functional needs, significantly improves task performance and reduces user burden compared to designs with generalized anthropomorphic geometry, particularly for gross motor activities.

The Gist

Imagine your hand as a master chef's tool. It's not just a rigid clamp; it's a flexible, intelligent instrument that can pick up a delicate egg, crush a walnut, or pour a glass of water without spilling a drop. Now, imagine trying to do all that with a pair of stiff, plastic tongs. That is essentially the challenge faced by many people using current prosthetic hands. They look like human hands, but they often move like rigid robots, making daily tasks frustrating and difficult.

This paper introduces a new approach to building prosthetic hands that tries to fix this by asking a simple question: "What if we built the hand to work exactly like a real human hand, not just look like one?"

Here is the story of their discovery, broken down into simple concepts.

The Problem: The "One-Size-Fits-All" Robot Hand

Most prosthetic hands today are like mass-produced tools. They have a standard shape that looks somewhat human, but they don't account for how your specific hand moves.

• The Rigid Grip: Imagine trying to pick up a bumpy potato with a pair of tongs that are locked in a straight line. If the potato is slightly off-center, the tongs slip. Traditional prosthetics often work this way: if one finger hits an object, the whole hand stops moving.
• The Cognitive Load: Because the hand doesn't adapt naturally, the user has to think constantly about how to position it. It's like driving a car where you have to manually calculate the friction of every tire on every turn. It's exhausting.

The Solution: The "Human Blueprint" (HAMR)

The researchers developed a new process called HAMR (Human-inspired Actuator Modeling and Reconstruction). Think of this as a "3D printer for your specific hand."

• Instead of using a generic mold, they scanned a real human hand to get the exact proportions, joint angles, and thumb placement.
• They used this "blueprint" to build a new prosthetic hand called TAPH (Tendon Actuated Prosthetic Hand).

The Secret Sauce: The "Whippletree" and Flexible Tendons

The real magic isn't just in the shape; it's in how the muscles (motors) pull the fingers.

• The Old Way (BSH): Imagine a puppet where all the strings are tied to a single knot. If you pull the knot, all fingers move at the exact same speed. If one finger hits a wall, the whole puppet stops. This is the "Bionic Skeletal Hand" (BSH) used for comparison.
• The New Way (TAPH): The researchers added a special mechanism called a Whippletree (named after the wooden bars used on old horse-drawn wagons to distribute weight).
  • The Analogy: Imagine a team of rowers. In the old hand, if one rower hits a rock, the whole boat stops. In the new hand, the "Whippletree" acts like a flexible steering system. If one finger hits an object, the others can keep moving and adjusting their grip. It distributes the force so the hand can wrap around irregular shapes naturally, just like a real hand does.

The Experiment: The "Gym Class" for Hands

To test if this new design actually worked, they didn't use people with limb differences (to avoid other variables). Instead, they asked 12 people without limb differences to wear a special suit that hid their real hands and gave them the prosthetic to use.

They had to perform three types of "daily life" tasks:

1. The Block Move (Gross Motor): Moving wooden blocks from one box to another. This tests basic grabbing and lifting.
2. The Checker Stack (Fine Motor): Picking up thin checkers and stacking them. This tests precision and delicate control.
3. The Carton Pour (Dynamic Task): Picking up a carton of marbles, moving it, and pouring the marbles into another box. This tests handling changing weight and balance.

The Results: The "Smart Hand" Wins

The results were clear, especially for the first two tasks:

• Speed and Efficiency: People using the TAPH (the human-shaped, flexible hand) moved blocks and stacked checkers significantly faster and with fewer mistakes than the rigid BSH.
• Less Frustration: When asked how hard the tasks felt, users reported that the TAPH felt much easier. They felt less "mentally drained" and less frustrated. It felt more intuitive, like their brain didn't have to work as hard to tell the hand what to do.
• The "Pour" Surprise: In the pouring task, both hands performed similarly in terms of accuracy. Why? Because pouring requires constant, dynamic adjustments to a changing weight (the marbles moving inside the carton). Even the "smart" hand struggled a bit because it lacked a fully active wrist to compensate for the shifting weight. This taught the researchers that while a human-like hand is great for grasping, it still needs help with balancing in dynamic situations.

The Big Takeaway

This study proves that biology matters. You can't just copy the look of a human hand; you have to copy the mechanics of how it works.

By building a hand that respects human anatomy (the right finger lengths and thumb angles) and uses a "flexible rope system" (tendons with a Whippletree) to share the load, the prosthetic becomes less of a tool you have to fight and more of an extension of your body.

In short: A prosthetic hand that thinks and moves like a human hand is faster, easier to use, and less frustrating than a rigid robot hand. It turns the "heavy lifting" of daily life from a chore into something that feels natural again.

Technical Summary

1. Problem Statement

Current upper-limb prosthetic hands often rely on rigid end effectors with generalized anthropomorphic geometries that do not align with specific human biomechanics. This mismatch creates a disconnect between the device's mechanical behavior and the user's natural grasping strategies, leading to:

• High Cognitive and Physical Burden: Users must exert significant mental effort and physical compensation to manipulate objects.
• Limited Adaptability: Traditional designs often couple hand aperture with impedance (stiffness), making it difficult to adapt to object shape variability.
• Low Adoption Rates: Due to poor usability and comfort, many users rely on their residual limb for dexterous tasks, using the prosthesis only for stabilization.

While tendon-driven designs offer potential for decoupling impedance from position, it remains unclear whether biologically informed geometry (derived from actual human anatomy) and adaptive force distribution improve functional performance compared to generalized designs, provided control strategies remain constant.

2. Methodology

A. Experimental Framework: HAMR Process

The authors developed the Human-inspired Actuator Modeling and Reconstruction (HAMR) process. This user-centered workflow integrates physical molding, 3D scanning, and CAD to derive prosthetic geometry directly from human anatomy. It defines:

• Anatomically aligned joint axes.
• Human-derived finger proportions and thumb orientation.
• Tendon routing paths consistent with the scanned morphology.

B. Experimental Setup

• Participants: 12 individuals without limb loss (8 female, 4 male), all with prior experience using myoelectric prosthetics.
• Platform: The Tendon Actuated Modular Prosthesis (TAMP), a modular platform with a forearm socket, two Maxon RE30 motors, and sEMG control (flexor/extensor mapping).
• Control: A First-In-First-Out Finite State Machine (FIFO FSM) maps sEMG signals to motor current. Both devices used identical motors, control logic, and user interfaces to isolate design variables.

C. Comparative End Effectors

Two tendon-driven hands were tested under identical actuation conditions:

1. Bionic Skeletal Hand (BSH): A generalized anthropomorphic design.
   • 4 fingers, 3 joints each (12 actuated joints).
   • Force Distribution: Rigid coupling. All finger tendons converge at a single ring; if one finger is blocked, the entire hand stops.
   • Geometry: Elongated segments, minimized palm, passive thumb rotation.
2. Tendon Actuated Prosthetic Hand (TAPH): A HAMR-derived design.
   • 4 fingers, 2 joints each (8 actuated joints).
   • Force Distribution: Adaptive underactuation via a Whippletree mechanism. This equalizes tendon forces, allowing unconstrained fingers to continue closing even if others are blocked by an object.
   • Geometry: Explicitly mapped from human anatomy (finger proportions, joint axes, thumb opposition). Features a semi-passive thumb with indexed positions (lateral, pinch, power).

D. Experimental Tasks

Participants performed three ADL-inspired tasks (Box and Blocks, Checker Transfer/Stack, Carton Transfer/Pour) across three trials each. Metrics included:

• Objective: Number of items transferred/stacked/poured, and pour accuracy.
• Subjective: NASA Task Load Index (TLX) for mental demand, physical demand, effort, frustration, and performance.

3. Key Contributions

1. HAMR Framework: A validated workflow for translating human anatomical data directly into prosthetic kinematic and tendon routing specifications.
2. Isolation of Design Variables: A controlled study demonstrating that geometry and force distribution alone (with constant control and actuation) significantly impact performance.
3. Adaptive Underactuation: The implementation of a Whippletree mechanism in a prosthetic hand to mimic biological force redistribution, allowing for robust grasping of irregular objects.
4. Empirical Evidence: Quantitative data showing that biologically derived designs outperform generalized anthropomorphic designs in specific motor tasks.

4. Results

Task Performance

• Gross Motor (Block Transfer): TAPH significantly outperformed BSH ($p < 0.001$) in the number of blocks transferred. No learning effect was observed, suggesting the advantage is intrinsic to the design.
• Fine Motor (Checker Transfer/Stack): TAPH resulted in significantly more checkers transferred and stacked ($p < 0.001$). A learning effect was observed only for stacking, indicating that while TAPH aids precision, repeated practice still benefits performance.
• Dynamic Interaction (Carton Transfer/Pour): TAPH enabled more transfers and pours ($p < 0.01$). However, pour accuracy did not differ significantly between devices.
  • Interpretation: While TAPH improved efficiency, the dynamic nature of the task (changing mass distribution) and the lack of an active wrist degree of freedom limited the advantage in precision pouring.

Subjective Workload (NASA TLX)

• TAPH consistently reduced mental demand, physical demand, effort, and frustration compared to BSH across all tasks ($p < 0.001$ for most metrics).
• In the dynamic carton task, workload differences were less pronounced, likely because the task's inherent complexity required high attention regardless of the hand design.

5. Significance and Conclusion

The study demonstrates that biologically informed design is not merely aesthetic but functionally critical.

• Mechanism of Improvement: The TAPH's success is attributed to two factors:
  1. Anatomical Geometry: Proper thumb opposition and finger proportions allow for more natural grasp strategies (e.g., stabilizing objects during pouring).
  2. Adaptive Force Distribution: The Whippletree mechanism allows the hand to conform to objects by redistributing force, preventing "lock-up" when a single digit encounters resistance.
• Limitations: The benefits are most pronounced in stable or repetitive tasks. Dynamic tasks requiring continuous reorientation (like pouring) highlight that hand geometry alone cannot compensate for the lack of active wrist control or dynamic sensory feedback.
• Future Impact: These findings support the shift toward task-specific, biomimetic prosthetic design. By integrating human-derived geometry and adaptive mechanics, future prosthetics can reduce user cognitive load and improve adoption rates without necessarily increasing control complexity.