Over the past decades, roboticists have introduced a wide range of systems with distinct body structures and varying capabilities. As the number of developed robots continuously grows, being able to easily learn about these many systems, their unique characteristics, differences and performance on specific tasks could prove highly valuable.
Researchers at Technical University of Munich (TUM) recently created the “Tree of Robots,” a new encyclopedia that could make learning about existing robotic systems and comparing them significantly easier. Their robot encyclopedia, introduced in a paper published in Nature Machine Intelligence, categorizes robots based on their performance fitness on various tasks.
“The aspiration for intelligent robots that can understand their environment as we humans do, and execute tasks independently, has existed for ages,” Robin Jeanne Kirschner, first author of the paper, told Tech Xplore.
“The active development of tactile robots—robots capable of understanding their surroundings through the sensation of touch—began approximately 20 years ago. This journey started with the creation of lightweight systems equipped with torque sensors in every joint. Since then, we have witnessed improved technology, better controllers, and new reaction schemes, which have enabled the development of systems proficient in executing tasks and perceiving the environment through touch.”
Most standards and approaches for classifying robots introduced to date do not account for the ability of systems to adapt to their surroundings and successfully interact with nearby objects by touching them. This crucial capability influences both the safety of robots and their performance on specific tasks, spanning various real-world applications.
“The focus of system classification remains separated based on, e.g., individual mechanical properties, new controller features, and certifications remains based solely on the mechanical structure of sensing systems instead of their actual performance,” said Kirschner. “This narrow approach often overlooks the interplay of components and the core purpose of a robotic device: to assist in executing tasks, which demands specific capabilities.”
To overcome the limitations of existing robot classification methods, Kirschner and her colleagues started testing various existing systems, focusing on features that influence their safety, such as their ability to detect contact with other objects. Concurrently, they also conducted an in-depth analysis of robotics tasks, deriving multiple metrics that indicate the capabilities of robots beyond safety, for instance, impacting their ability to successfully execute tactile tasks and comfortably interact with humans.
“By testing multiple robot manipulators, we were then able to derive all these metrics and show that the tactility fitness of these systems significantly varies, calling for a proper classification and encyclopedia—the Tree of Robots,” said Kirschner.

“As a result, we established the AI Robot Performance and Safety Center—a dedicated laboratory equipped with advanced measurement devices to evaluate robot performance. With these resources, we aim to further grow the ‘Tree of Robot,’ an essential encyclopedia for the field of robotics.”
The Tree of Robots encyclopedia is meant to be continuously updated over time, ultimately serving as a Wikipedia-like platform that contains information about robots and their capabilities. It includes a wide pool of information ranging from the robots’ fundamental body structures to the motors and/or sensors they rely on and their resulting capabilities, specifically the sensitivity and reliability of their physical interactions (i.e., tactility fitness) and precision of their movements (i.e., motion fitness).
“While we began with analyzing and classifying existing stationary manipulators using fitness metrics we defined specifically based on for industrial applications, the encyclopedia must grow to encompass other robotic systems for service tasks, such as humanoids or mobile robots,” explained Kirschner. “Its purpose is to efficiently guide both hardware and software development in robotics.”
In contrast with many previously devised robot categorization approaches, the Tree of Robots encyclopedia clearly outlines the specialized capabilities of different robots. In addition, it groups robots into three main groups based on their tactility fitness, which indicates the extent to which they are suitable for completing specific tasks.

“This fundamental insight should be integrated into application design, standardization efforts, and future robotics development,” said Kirschner. “By aligning hardware and software components to achieve optimal performance for a given process—rather than designing processes to fit the system’s constraints—we can advance robotics to new levels of efficiency and effectiveness.”
The new encyclopedia developed by Kirschner and her colleagues could inform future research, for instance, by helping other computer scientists and roboticists to identify the best systems to test their algorithms. Meanwhile, the researchers plan to continue adding information to the Tree of Robots, including other robotic systems and other relevant metrics.
“We are now expanding our work in several directions,” added Kirschner. “My focus is on linking these critical findings to ensure human safety in collaborations, emphasizing a robot’s tactile capabilities. The goal is to achieve certifiably safe applications with tactile robotic systems. Alongside other teams, we are also exploring how to extend the tree of robots in other areas, such as systems designed for service and care tasks and including, e.g., humanoid systems.”
More information:
Robin Jeanne Kirschner et al, Categorizing robots by performance fitness into the tree of robots, Nature Machine Intelligence (2025). DOI: 10.1038/s42256-025-00995-y.
© 2025 Science X Network
Citation:
An evolving robotics encyclopedia characterizes robots based on their performance (2025, March 16)
retrieved 16 March 2025
from
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no
part may be reproduced without the written permission. The content is provided for information purposes only.