
![]() |
Charlotte Stonestreet
Managing Editor |
Home> | AUTOMATION | >Robots | >Physical AI framework enables rapid learning of complex skills in robotics |
Editor's Pick
Physical AI framework enables rapid learning of complex skills in robotics
07 August 2025
DESPITE SIGNIFICANT advancements, robotic automation can struggle with reliably performing delicate, tactile tasks like inserting connectors or handling flexible materials. A new breakthrough – Tactile Skills – claims to be changing this.

Sami Haddadin, VP for Research at Mohamed bin Zayed University for Artificial Intelligence (MBZUAI) has worked with his former PhD student Lars Johannsmeier and colleagues Yanan Li from the University of Sussex and Etienne Burdet from Imperial College, and led the work behind the paper published in Nature Machine Intelligence in June.
The paper introduces the groundbreaking embodied AI framework coined Tactile Skills as a theoretically sound, practical and scalable framework, inspired by the human neural system and human vocational training. Their method uses a clearly structured taxonomy based on expert-defined process specifications. In simple terms, it’s like giving robots their own specialised curriculum, helping them rapidly learn and master new physical tasks.“Our framework bridges the gap between human expertise and robotic capability. It’s rewarding to see robots now reliably mastering intricate tasks with precision and adaptability – this isn’t just incremental progress; it’s a meaningful leap forward for practical automation in everyday life,” said Sami Haddadin, VP for Research and Professor of Robotics at MBZUAI.
This approach has been tested extensively on 28 distinct industrial tasks, including complex operations like plug insertion and precision cutting. Remarkably, the robots achieved nearly 100% success and industrial grade performance, even when encountering unexpected changes in object positioning or environmental conditions. These robots not only performed reliably but did so quickly and with minimal error.
A crucial advantage of this method is that it doesn’t rely solely on extensive trial-and-error or massive datasets, as traditional machine-learning methods often do. Instead, it blends expert process knowledge with reusable tactile control and adaptation components, significantly simplifying and speeding up the robot’s learning process. This reduces energy consumption and increases performance dramatically compared to current deep-learning-based approaches.
One of the framework’s standout achievements was successfully assembling a complex industrial device used in bottle-filling plants, demonstrating its practicality for real-world manufacturing scenarios. By bridging expert process definitions and robotic learning in a structured way, this method empowers operators who may not have deep robotics expertise to effectively deploy robots on diverse tasks, greatly reducing setup times and costs.
“This research represents an important leap toward widespread automation, transforming robots from specialised tools into adaptable, skilled assistants and ultimately physical AI agents. Industries can now envision greater automation of complex, tactile tasks, unlocking new potential for efficiency, safety, and scalability. Ultimately, this breakthrough doesn’t merely improve how robots handle specific tasks—it reshapes what’s possible for automation, making versatile and reliable robotic skills broadly accessible for industries and everywhere, and ultimately our homes,” said Haddadin.
- Keeping heavy loads moving without cage creep
- Sophisticated end of arm tooling enhances automated manufacturing
- UK a European hub for innovation in 3D printing
- Honeywell and Hai Robotics to collaborate
- Three layers of predictive maintenance
- Modern approaches to industrial maintenance
- Digital transformation clinic for SME makers
- Laser Interferometry & automated manufacturing
- Elisa Smart Factory & TCS partner to deliver advanced analytics solutions to manufacturing sector
- Accelerating robotics from cloud to edge
- Balls to Gravity
- ROBOTS GIVE FASTER CELL TOOL CHANGE TIMES
- ROBOT FOR DELICATE ASSEMBLY
- NUMBER OF ROBOTS IN UK ON THE RISE
- The future is bright - The future is automation!
- High Demand For Grant
- More Capacity, Less Space
- FREEBIRD - THE THIRD DIMENSION
- PRECISION ROBOT CELL FOR THE PHARMACEUTICAL INDUSTRY
- ROBOT SALES REACH ALL TIME HIGH