Robotics is undergoing a rapid transformation driven by advances in both imitation learning and reinforcement learning (RL). Emerging Vision–Language–Action (VLA) models and policy-learning frameworks promise to augment—and in some cases replace—traditional pipelines for perception, planning, and control.
Nikolaus Correll,
Notre Dame
In this talk, I will describe recent efforts of my group toward autonomous dismantling of EV batteries using a combination of collaborative industrial robots and humanoids working in concert with humans. Here, we are using both large (vision) language models (L(V)LM) and optimal control to strive for a balance between generalizability and robustness. In particular, I will describe how we use LLMs and VLMs to leverage common sense knowledge to improve classical planning and control to improve manipulation, as well as how optimization techniques in conjunction with physics-based simulation can be used to create robust controllers for full-body humanoid motion.
Nikolaus Correll is the Viola D. Hank Professor in Aerospace and Mechanical Engineering at the University of Notre Dame, which he joined in Fall 2026. Before that, he was a Professor of Computer Science at CU Boulder. Nikolaus obtained a degree in electrical engineering from ETH Zurich in 2003, a PhD from EPFL in Computer Science in 2007 and did a post-doc at MIT CSAIL from 2007-2009. Nikolaus is the recipient of a 2012 NSF CAREER award, a NASA Early Career Faculty Fellowship, and the 2016 Provost Achievement award.