💻 computer science

Energy-Based Injury Protection Database: Including Shearing Contact Thresholds for Hand and Finger Using Porcine Surrogates

This study establishes the first energy-based Injury Protection Database by expanding existing porcine surrogate data to include shearing contact scenarios, revealing that collision angle significantly influences injury outcomes and enabling the development of more robust energy-limiting controllers for safe human-robot interaction.

Original authors: Robin Jeanne Kirschner, Anna Huber, Carina M. Micheler, Dirk Müller, Nader Rajaei, Rainer Burgkart, Sami Haddadin

Published 2026-02-25

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Robin Jeanne Kirschner, Anna Huber, Carina M. Micheler, Dirk Müller, Nader Rajaei, Rainer Burgkart, Sami Haddadin

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are building a robot that is supposed to work side-by-side with humans in a factory or a home. You want it to be strong and helpful, but you also need to make sure it doesn't accidentally hurt anyone if they bump into each other.

For a long time, safety engineers have treated robot collisions like a hammer hitting a nail. They assumed that if a robot arm hit a human hand, it would be a straight-on, "bonk" impact (perpendicular). They set safety limits based on how much energy that "bonk" could have before it broke a bone or bruised skin.

But here's the problem: In the real world, collisions are rarely perfect "bonks." They are more like sliding, scraping, or slicing. Think of a robot arm brushing past your hand, or a sharp edge gliding along your finger. This is called "shearing."

This paper is like a safety detective story that asks: "Do our current safety rules work for these sliding, scraping accidents, or do we need new ones?"

The Experiment: The "Piggy" Test Dummies

Since we can't ethically test dangerous robot crashes on real people, the researchers used pig paws as stand-ins for human hands.

Why pigs? Pig skin and human skin are very similar in texture and thickness. It's like using a very realistic, high-tech mannequin.
The Setup: They built a giant swinging pendulum (like a wrecking ball) that could hit the pig paws.
The Variables: They changed three things:
1. The Shape: Did the robot hit with a flat sheet, a sharp wedge, or a sharp edge?
2. The Angle: Did it hit straight on (90°) or slide across at an angle (45° or 70°)?
3. The Force: How heavy was the robot and how fast was it moving?

They ran 1,080 experiments. That's a lot of "oops" moments to gather data!

The Big Discovery: Sliding is Safer (Surprisingly!)

The researchers expected that sliding (shearing) might be worse because it could slice the skin. But they found the opposite.

The "Hammer" (Perpendicular Hit): If a sharp object hits your hand straight on, it punches right through. This is the most dangerous.
The "Knife Slide" (Shearing Hit): If that same sharp object slides across your hand, the skin and tissue can move with the object a little bit, absorbing some of the energy. It's like the difference between a car crash where you hit a wall head-on versus a car that just scrapes along the side of a building. The scrape is still bad, but it's less likely to cause a catastrophic injury than the head-on crash.

The Result: The "sliding" accidents allowed for more than twice the energy before causing an injury compared to the "straight-on" accidents.

The New "Injury Protection Database"

The authors created a new Safety Rulebook (an Energy-Based Injury Protection Database).

Think of this like a speed limit sign for robots, but instead of miles per hour, it measures Joules of Energy.

Old Rule: "If the robot has more than X energy, it's dangerous." (Based only on straight hits).
New Rule: "If the robot is sliding, it can have a bit more energy before it's dangerous. But if it's hitting straight on, it must be much slower."

They found specific limits for different shapes:

Flat Sheet: Can handle a bit more energy.
Sharp Edge: Very sensitive; even a tiny bit of energy can cause a cut.
Wedge: Somewhere in the middle.

Why Does This Matter?

Right now, robot safety systems are often too cautious. Because they only know about "straight hits," they might make robots move incredibly slowly to be safe, even when the robot is just gently brushing past a human. This makes robots slow and inefficient.

With this new data, engineers can program robots to be smarter:

If the robot knows it's about to slide past a human, it can move a bit faster safely.
If the robot knows it's about to hit head-on, it will slam on the brakes immediately.

The Bottom Line

This paper gives us the first clear map of how different types of robot collisions hurt humans. It tells us that angle matters. By understanding that "sliding" is different from "hitting," we can build robots that are not just safe, but also fast and useful, allowing them to work safely alongside us in our daily lives.

In short: We used pig hands to learn that robots don't always have to be slow and clumsy to be safe. If they know how they are going to bump into you, they can adjust their speed to keep you safe without slowing down the whole factory.

1. Problem Statement

Current safety standards for Human-Robot Interaction (HRI), such as EN ISO 10218-2, rely heavily on energy thresholds derived from blunt, perpendicular impacts (e.g., 90° collisions). However, these standards lack data for:

Edged or pointed collisions: Scenarios involving sharp tools or robot edges.
Shearing contacts: Real-world collisions where the robot moves tangentially or at an angle across the skin (unconstrained collisions), rather than striking it perpendicularly.
Scalable datasets: There is a lack of clinically grounded, energy-based datasets covering diverse contact geometries and angles necessary for validating safety controllers in dynamic environments.

Existing force-based thresholds are often impractical for real-time control due to the complexity of modeling contact forces. Therefore, there is a critical need for an energy-based Injury Protection Database that accounts for shearing dynamics to ensure safe HRI.

2. Methodology

The study utilized porcine surrogates (pig front paws and dew claws) to simulate human hands and fingers, leveraging the anatomical and tissue similarity between porcine and human skin.

Experimental Setup:
- A dual-pendulum system was used. Pendulum I represented the human limb (with effective masses of ~2.6 kg for hand/lower arm and ~6.4 kg for the entire arm) and was free to recoil. Pendulum II represented the robot, equipped with various impactors and variable effective masses (up to 9.46 kg).
- Impactors: Three geometries were tested: a 90° wedge, a 90° edge, and a 1.6 mm thick aluminum sheet.
- Variables:
  - Collision Angles ( $\alpha$ ): 45° (shearing) and 70° (shearing), compared against previous 90° (perpendicular) data.
  - Velocities: 1.0, 1.5, and 2.0 m/s.
  - Masses: Various effective masses to simulate different robot weights and human limb inertia.
- Total Experiments: 1,080 trials.
Injury Analysis:
- Injuries were categorized into four classes: No visible damage (S0), Skin imprints (S1), Skin cuts/abrasions (S1/S2), and Tendon/Bone injuries (S2).
- Severity was mapped to ISO 12100 standards (S0–S4), focusing on preventing even minor reversible injuries (S1).
- Statistical Analysis: A Chi-Square test was performed to determine the statistical significance of the relationship between impact angle and injury occurrence.
Energy Calculation:
- Kinetic energy ( $K_{max}$ ) was calculated based on the effective mass of the robot pendulum and the measured impact velocity: $K_{max} = \frac{1}{2} m_{eff} v_c^2$ .
- Thresholds were defined using the 5th percentile (lowest whisker of box plots) of the energy distribution for each injury type to ensure a conservative safety margin.

3. Key Contributions

First Energy-Based Injury Protection Database for Shearing: The paper establishes the first dataset linking kinetic energy thresholds to injury outcomes for unconstrained shearing contacts, filling a gap left by existing perpendicular-impact-only standards.
Quantification of Angle Influence: It provides empirical evidence that collision angle significantly alters injury probability and energy tolerance.
Re-evaluation of Safety Thresholds: The study re-analyzed previous perpendicular impact data alongside new shearing data to create a unified framework for safety limits across different geometries (wedge, edge, sheet).
Decision Matrix for Controllers: It proposes a scenario-based selection matrix (Fig. 11) for engineers to choose appropriate energy limits based on the specific contact geometry (sharp vs. blunt) and collision type (perpendicular vs. shearing).

4. Key Results

Impact Angle Significance:
- A Chi-Square test confirmed a statistically significant association between impact angle and injury probability ( $p < 0.0001$ ).
- Shearing (45°/70°) vs. Perpendicular (90°): Unconstrained shearing contacts resulted in fewer injuries than perpendicular impacts. Specifically, shearing scenarios allowed for more than twice the energy threshold before injury onset compared to perpendicular impacts.
- Injury Probability: The probability of injury at 45° was 0.11, whereas at 70° it was 0.19, both significantly lower than typical perpendicular impact rates.
Energy Thresholds (Conservative 5th Percentile):
- Sheet Impactor: No skin injury up to 1.10 J.
- Wedge Impactor: No skin injury up to 2.68 J.
- Edge Impactor: No skin injury up to 0.78 J (the most critical threshold).
- Severity Thresholds:
  - For S1 injuries (minor reversible), the threshold for 45° impacts was 1.61 J.
  - For 70° impacts, the S1 threshold was 0.78 J (driven by edge impacts), and the S2 threshold was 3.78 J.
Observations on Injury Types:
- Skin abrasions and cuts showed closely aligned energy values, making distinct boundaries difficult to define.
- Tendon injuries were rare, occurring mostly at higher velocities (2.0 m/s) with specific mass configurations.

5. Significance and Implications

Controller Design: The findings enable the development of energy-limiting controllers that are less conservative for shearing scenarios (allowing higher speeds/energies safely) while maintaining strict limits for perpendicular or sharp-edge contacts.
Safety Standards: The paper argues that relying solely on perpendicular impact data (like EN ISO 10218-2) is overly conservative for shearing scenarios but potentially insufficient for sharp-edged contacts. It advocates for a geometry-aware safety approach.
Real-World Application: By validating that shearing contacts are generally less injurious than perpendicular ones, the study supports the deployment of humanoid robots in constrained, dynamic environments where tangential contact is inevitable.
Future Work: The authors recommend future validation using human tissue and expanded statistical evaluations to refine these thresholds further, noting that porcine tissue is robust and human thresholds may be slightly lower.

In conclusion, this study provides a foundational dataset that moves HRI safety beyond simple blunt-force metrics, offering a nuanced, energy-based framework that accounts for the complex dynamics of real-world robot-human collisions.