ResWORK Seminar by Dr. Olivier Stasse and Dr. Vincent Bonnet
Talk 1: Vision Language Action models, Whole Body Model Predictive Control and Reinforcement Learning to Generate Motion on Legged Robots
 | Speaker: Dr. Olivier Stasse Tenured Senior Research Scientist, Laboratory of Analysis and Architecture of Systems, CNRS, Toulouse, France |
The talk reviews the Gepetto group’s two decades of work using humanoid robots to explore motion generation, highlighting optimization-based approaches linking motion planning and control. It presents recent progress in deploying end-to-end sensor-driven policies for dynamic quadruped motions, discusses challenges in extending these methods to humanoids, and underscores hardware importance through experiences with robots like TALOS and the ODRI initiative.
Talk 2: Capturing, Modelling, and Understanding Human Dynamics for Assistive Robotics
 | Speaker: Dr. Vincent Bonnet Associate Professor, University of Toulouse, Member of the GEPETTO team at LAAS-CNRS, Toulouse Adjunct Visiting Associate Professor, School of Computing, National University of Singapore |
The talk addresses the challenge of designing human‑centred controllers in assistive and collaborative robotics. After introducing LAAS‑CNRS, it highlights three pillars for seamless physical interaction:
- Real‑time estimation of human kinodynamics for safe contact.
- Predicting human motion and intent using explainable models that integrate multi‑modal signals and task constraints.
- Efficient calibration of human and robot models to reduce the sim‑to‑real gap through tailored identification methods