# Scientists Develop Robotic Hand That Walks Like Addams Family's Thing

Researchers at ETH Zurich enable an off-the-shelf robotic hand to walk independently.

- **Source:** [Science & tech Articles | Mail Online](https://www.dailymail.com?ns_mchannel=rss&ns_campaign=1490&ito=1490)
- **Published:** 2026-11-09
- **Topic:** [technology](https://brifing.si/c/technology)
- **Original article:** <https://www.dailymail.com/sciencetech/article-16278347/robotic-hand-addams-family-thing.html?ns_mchannel=rss&ns_campaign=1490&ito=1490>
- **On Brifing:** https://brifing.si/s/4d69e581-2d97-40b4-830e-e36651eff7ae

## Autonomous Robotic Locomotion

Researchers at ETH Zurich have successfully engineered a robotic hand capable of walking independently on its fingertips. Unlike traditional designs that rely on bulky arm attachments, this 1.8lb device functions autonomously. It mimics the movement of the fictional character Thing, demonstrating the ability to navigate diverse terrains and perform complex physical tasks.

## Hardware Modification Strategy

The team utilized an off-the-shelf WUJI robotic hand rather than building a custom machine from scratch. By integrating a compact battery, a small onboard computer, and a motion sensor, they transformed existing hardware into a mobile unit. This approach highlights the potential for repurposing standard industrial components for advanced, independent robotic mobility.

## Reinforcement Learning Implementation

To achieve stable movement, engineers employed reinforcement learning, a method where the system practices through trial and error in simulations. This process allowed the hand to calculate optimal finger movements without manual programming. By simulating thousands of iterations, the robot learned to maintain balance while navigating across 14 different surface types.

## Performance and Versatility

Testing confirms the hand can right itself after falling, succeeding 21 out of 25 times. It demonstrates high precision, pushing blocks to within two-thirds of an inch of a target. Furthermore, the device successfully operated a keyboard during trials, hitting the correct keys 29 out of 32 times while maintaining its balance.

## Future Operational Utility

The primary goal is to enable humanoid robots to reach confined spaces by detaching limbs for specific tasks. Lead researcher Amirhossein Kazemipour envisions a future where robot parts possess independent sensing and mobility. This modular design would allow robots to separate for specialized inspections and reconnect when a full-body task is required.

## Technical Control Mechanisms

Currently, the hand operates via remote control, with an onboard computer translating operator commands into specific finger movements. An overhead camera provides essential spatial information, allowing the system to interact with its environment effectively. This integration of remote guidance and autonomous balance control represents a significant step toward more complex, multi-part robotic systems.

## Overcoming Mechanical Challenges

Teaching a hand to crawl is inherently difficult due to the varying lengths and movement ranges of human-like fingers. Every step changes the hand's support structure, requiring constant adjustments to prevent tipping. The researchers addressed this by combining virtual simulation data with real-world measurements of the hand's 20 moving joints and friction.

## Implications for Robotics

This development shifts the paradigm of robotic design from monolithic structures to modular, collaborative components. By proving that existing hardware can be modified for independent movement, the study paves the way for more agile, versatile machines. Future iterations may focus on increasing autonomy and refining the hand's ability to manipulate objects in tight spaces.

## Sources covering this story

- https://www.dailymail.com/sciencetech/article-16278347/robotic-hand-addams-family-thing.html?ns_mchannel=rss&ns_campaign=1490&ito=1490 — 

---

_Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._
