Would you like to work toward ensuring that humans and robots can collaborate safely and reliably as a team in the future? In this research project, you will combine formal safety mechanisms with multimodal large-language models on embedded systems. The goal is to develop autonomous robotic systems that can act intelligently even when performing complex tasks—such as large-scale assembly—understand their environment, and simultaneously meet the highest safety requirements.
They enable the safe operation of robots that autonomously plan and execute tasks in collaboration with humans based on Vision-Language-Action (VLA) models. To this end, you will specifically investigate how to make the output of VLA models accessible to formal safety mechanisms based on contracts (neuro-symbolic reasoning) and how these mechanisms can be used for feedback and training of the VLA model. When executing actions, you will investigate a fallback mechanism that ensures minimally acceptable robot behavior even in the event of a disruption to AI functions. The research project is being conducted in close international collaboration with partners in Taiwan. As part of this collaboration, you will also work on-site with the project partners. To this end, multiple research stays in Taiwan, each lasting approximately one to two months, are planned during the project period.