Neuromorphic sensors revolutionize industrial welding

Cobots (collaborative robots), which can work closely with humans, are experiencing impressive growth rates and offer enormous potential for a wide range of industries. Until now, however, cobots have only been able to carry out pre-programmed actions and cannot react flexibly to unexpected events. With the CORINNE project, fortiss aims to overcome this limitation through the use of neuromorphic technology and enable robots to react to human gestures in real time and continuously learn new ones. In collaboration with Neura Robotics GmbH from Metzingen and the Technical University Chemnitz, the research results are to be applied in the field of welding work.

In the rapidly advancing field of robotics, fortiss is setting new standards in the development of innovative neuromorphic visual sensors with the CORINNE (Cobots' Relational Interface with Neuromorphic Networks and Events) project. These sensors are designed to optimize gesture recognition and enable online learning in welding applications. The CORINNE project will be realized with the Maira cobot, an intelligent robot arm developed by Neura Robotics GmbH and which will be equipped with the new neuromorphic sensors.

Gesture recognition and online learning for intelligent welding robots

In contrast to conventional robots, which can only perform pre-programmed actions, robots with neuromorphic sensors learn continuously and adapt flexibly to new situations. This ability enables them to recognize and react to the welder's gestures in real time. The main advantage of neuromorphic technology lies in its ability to adapt to the individual user without the need for extensive models or large amounts of training data. The approaches of this innovative research can be seen in the video “Smart Melding with MAiRA”. The neuromorphic AI algorithms developed by fortiss enable robots to learn new gestures and gesture combinations in real time and with minimal training, significantly reducing reliance on large data sets and increasing cost-effective applicability.

fortiss scientist Jules Lecomte works in the fortiss Neuromorphic Computing competence field and will lead the project for the next two years. He is convinced of the promising potential of the technology and explains: “Our neuromorphic visual sensors enable robots to learn dynamically and react to human gestures, which increases their flexibility and usability in complex industrial tasks such as welding. This technology represents a paradigm shift in the way robots interact with human operators.”


By activating this video, you consent to transmitting data to YouTube.

Copyright © NEURA Robotics GmbH

Federated learning enables autonomous, adaptable robots

Enabling even more flexibility and seamless industrial application, another technical goal of the project is federated learning, developed by Prof. Röhrbein at Chemnitz University of Technology, in which several robots can share and learn new knowledge. This approach not only accelerates the learning process, but also ensures data sovereignty by minimizing dependencies on external data sources.

The impact of this innovation goes beyond industrial welding. Neuromorphic sensors have the potential to revolutionize various industries by enabling robots to adapt and learn autonomously, paving the way for more efficient and adaptable automation solutions.

The integration of neuromorphic sensors into industrial robotics represents an important milestone in the development of automation technology. As research in this area progresses, the potential applications and benefits of neuromorphic robotics are likely to increase, driving further innovation and reshaping the future of manufacturing. The expected research results from fortiss could therefore significantly accelerate the introduction of cobots and other autonomous systems into production environments. This will lead to higher productivity and flexibility.

  Marketing & press

Your contact

Marketing & press