Make industrial robots more flexible with embodied AI
Join EIDOS_SBO and explore how embodied AI can help you train robots faster, automate complex tasks and adapt to changing production environments with less programming effort.
Challenge
Industrial robots are becoming increasingly important in manufacturing and logistics. However, traditional robotic automation often struggles with tasks that require flexibility, adaptability and complex decision-making. Handling different product variants, manipulating objects in unpredictable positions or performing multiple interconnected actions still requires extensive programming and engineering expertise.
Most robotic systems are designed for predefined tasks and controlled environments. When products, materials or operating conditions change, robots often need to be reprogrammed or retrained. This makes automation costly and difficult to justify, particularly in high-mix, low-volume production environments.
Recent advances in embodied AI offer new opportunities for more flexible robotic systems. However, current AI models still face challenges in learning new skills from limited data, combining individual actions into reliable workflows and responding safely to unexpected situations. These limitations prevent manufacturers from fully exploiting the potential of intelligent robotics.
Project Goals
In this project (EIDOS_SBO), we aim to develop and evaluate embodied AI technologies that enable industrial robots to learn new tasks efficiently, perform complex operations autonomously and adapt safely to changing environments.
The project focuses on three complementary developments: data-efficient robotic skill learning, intelligent task orchestration and real-time monitoring. By combining AI-based learning with vision, force and tactile feedback, robots will learn manipulation skills from demonstrations and apply them to previously unseen objects and configurations.
An intelligent orchestration system will coordinate individual skills into complex, multi-step operations, while independent monitoring and recovery mechanisms will help robots detect errors, adapt their actions and operate more safely.
In this project, we target at least 80% success on previously unseen object configurations using no more than 50 demonstrations, 90% autonomous completion of six-step industrial tasks and detection of at least 95% of injected faults within 200 milliseconds.
The project also aims to reduce the engineering effort needed to introduce a new production variant to less than one day, without requiring a robot programmer.
By making robotic automation more adaptable and easier to deploy, EIDOS aims to help manufacturers, technology providers and system integrators unlock new automation opportunities, particularly in high-mix, low-volume production.
Join our research
We are looking for companies that want to participate in our research and work with us on the valorisation of the project.
Interested?
Then complete the form below.
