Advanced learning for generalizable and reliable robotic skills

Robotics is crucial in Europe for driving innovation and maintaining global competitiveness in advanced manufacturing, service and automation, etc. At present, one of the primary limitations hindering the further development of robotics today is the lack of sufficient intelligence, which prevents robots from being deployed in complex tasks such as those in home-services, and healthcare. Robotic skill learning aims to provide robots with the necessary intelligence to acquire the skills needed for performing these complex tasks. The explosion of AI technologies in recent years has provided a large platform and space for robot skill learning. The primary objective of this proposal is to develop a generalizable robotic skills learning system leveraging advanced AI techniques. This system will be designed to emulate the human skill-learning process, thereby enhancing both the reliability and explainability of the learning behaviors. The researcher will investigate foundation model techniques to provide human-level generalization ability for robots. Moreover, The researcher will present novel biologically inspired sensorimotor policy learning algorithms to significantly enhance the reliability and explainability of the system. The system will be validated via various applications such as assembly across different objectives. During the project, the researcher will acquire a solid multidisciplinary scientific and transferable skills training, covering from AI theorems, robotic applications, project management skills and career planning, etc. The final aim of the project is to push the Europe in a good position to ride the coming wave of ""AI + Robots"".""