Diffusion-Based Humanoid Motion
Master's thesis work on physics-based humanoid motion, diffusion priors, and stability under perturbations.
My Master’s thesis at the Max Planck Institute for Informatics focuses on diffusion-driven physics-based humanoid motion. The project augments a motion prior with biomechanics-inspired loss guidance to improve stability under external perturbations.
As part of this work, I designed and trained a neural network to estimate ground reference points that influence stability, supervised using insole sensor data.
Topics: embodied AI, generative models, diffusion models, physics-based simulation, motion synthesis.