TUM Introduces Industry-Standard Evaluation Protocol to Certify Robot Tactile and Motion Capabilities

TUM Introduces Industry-Standard Evaluation Protocol to Certify Robot Tactile and Motion Capabilities

(IN BRIEF) TUM MIRMI has unveiled a comprehensive testing framework that measures a robot’s tactile sensitivity and overall fitness for safe human collaboration. Developed by Alessandro Melone, Robin Kirschner, and Kübra Karacan at the AI Robot Safety & Performance Centre, the protocol combines 25 tactile metrics with standard motion assessments. Results are visualized via a “Tree of Robots,” classifying systems into industrial robots, cobots, soft robots, and tactile robots. This initiative aims to establish an industry-wide seal of approval, empowering businesses with transparent performance data to guide robot selection and deployment.

(PRESS RELEASE) MUNICH, 15-Jul-2025 — /EuropaWire/ — TUM researchers at the Munich Institute of Robotics and Machine Intelligence (MIRMI) have developed a groundbreaking evaluation protocol to benchmark autonomous robots’ sensitivity and suitability for safe human interaction. When machines collaborate with people and perform tasks independently, their tactile awareness and adaptability are essential for both safety and versatility. Until now, however, no universal framework existed to quantify a robot’s touch sensitivity or overall fitness for close physical work. The new scheme, devised at TUM MIRMI’s AI Robot Safety & Performance Centre, initially targets industrial robotic arms and can later be expanded to mobile platforms, humanoid robots, and dexterous robotic hands.

The research team at the AI Robot Safety & Performance Centre at TUM MIRMI, from left to right: Kübra Karacan, Alessandro Melone and Robin Kirschner.

At the Centre, experts Alessandro Melone, Robin Kirschner, and Kübra Karacan have meticulously analyzed manipulator sensitivity over several years. “Our testing methodology at MIRMI has the potential to set a new industrial standard,” asserts Achim Lilienthal, Deputy Director of MIRMI and Professor of Intelligent Systems Perception. He envisions a recognized seal of approval that delivers transparent performance data to manufacturers and end users alike. “Clear insights into robot capabilities enable companies to deploy systems more effectively,” says MIRMI Executive Director Lorenzo Masia. “I’m confident our Centre will become Germany’s leading robotics testing authority.”

Categorizing Single-Arm Manipulators

The first phase evaluated an array of single-armed robots from prominent industrial and academic vendors. Though similar in appearance, these manipulators diverge significantly in sensor configurations, actuator technologies, and control architectures—yielding distinct performance profiles. Some excel in brute strength and pinpoint accuracy; others distinguish themselves through compliant motion and refined sensitivity to environmental contact.

To illustrate these variations, the team crafted the “Tree of Robots,” modeled after Darwin’s Tree of Life. This visual taxonomy maps robots according to specialized capabilities—path-following precision, positioning accuracy, contact gentleness, and collision safety—revealing each model’s niche strengths in a clear, hierarchical layout.

Measuring Tactility with 25 Metrics

The protocol employs 25 discrete measurements to capture a robot’s tactile performance: force regulation, surface interaction fidelity, and collision response, among others. These data points are plotted on a spider diagram, offering an immediate portrait of sensitivity levels that stakeholders can interpret without specialist expertise.

Mapping Strengths, Weaknesses, and Use Cases

By integrating these tactile metrics with conventional motion-performance indicators, MIRMI’s framework delivers the first holistic comparison of a robot’s core interactive skills. Based on the results, robots are classified into four categories—industrial robots, collaborative robots (cobots), soft robots, and tactile robots—each tailored to specific applications. Surgical systems demand uncompromising precision, while warehouse and factory units prioritize robustness and endurance for prolonged repetitive tasks.

Publications

Robin Jeanne Kirschner, Kübra Karacan, Alessandro Melone and Sami Haddadin; Categorizing robots by performance fitness into the tree of robots; Nature Machine Intelligence, 24.3.2025; https://www.nature.com/articles/s42256-025-00995-y

Further information and links

The Munich Institute of Robotics and Machine Intelligence (TUM MIRMI) is an integrative research institute at the Technical University of Munich, where researchers develop intelligent machines that interact with humans and are capable of learning. More than 70 professors from TUM are involved in TUM MIRMI, focusing on robotics applications, perception and artificial intelligence. The focus is on applications in the fields of health, work, environment and mobility. A five-member Board of Directors defines the research and innovation strategy as well as the core teaching content of TUM MIRMI, establishes and coordinates new focus groups, appoints new research and innovation leaders, and proposes new members for the Science Advisory Board and the Industry Advisory Board. The Board of Directors consists of Prof. Lorenzo Masia (Executive Director), Prof. Achim Lilienthal (Deputy Director and Director of Strategy and Partnerships), Prof. Angela Schoellig (Director of Industry and International Affairs) and Prof. Eckehard Steinbach (Director of Start-ups and Infrastructure). More information: https://www.mirmi.tum.de/mirmi/startseite/

Media Contacts:

Corporate Communications Center
Andreas Schmitz
presse@tum.de

Contacts to this article:

Robin Kirschner
Laboratory Manager, AI Robot Safety & Performance Centre
Munich Institute of Robotics and Machine Intelligence (MIRMI)
Technical University of Munich (TUM)
robin.kirschner@tum.de

Prof. Achim Lilienthal
Chair for Perception for Intelligent Systems
Technical University of Munich (TUM)
achim.j.lilienthal@tum.de

SOURCE: Technical University of Munich