Andreas Sochopoulos

PhD in Robot Learning | University of Edinburgh | Honda Research Institute Europe

profile_pic.jpeg

I am currently a PhD student at the University of Edinburgh (UoE), supported by Honda Research Institute Europe (HRI-EU), and advised by Sethu Vijayakumar (UoE), Michael Gienger (HRI-EU).

My research focuses on imitation learning using continuous-time generative models, such as diffusion and flow models. So far, I have worked on accelerating the inference of flow models for action generation without additional training overhead and efficient RL post-training . I am particularly interested in ways to make diffusion/flow-based policies force-aware, fast, and training/adapting them with as few demonstrations as possible.

I am currently exploring demonstration-efficient VLA steering. If you are interested in my research or would like to discuss related topics, feel free to reach out!

News

Aug 01, 2025 Our paper, Fast Flow-based Visuomotor Policies via Conditional Optimal Transport Couplings, has been accepted at the Conference on Robot Learning (CoRL), 2025!
Jul 01, 2025 Thrilled to announce that I will be spending this July as a visiting PhD in the University of Tokyo. Feel free to reach out if you’re in Tokyo.
Jun 01, 2024 Our paper, Learning Deep Dynamical Systems using Stable Neural ODEs, has been accepted at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024!

Selected Publications

  1. otql_thumbnail.gif
    Optimal Transport Q-Learning for Flow Policy Steering and Acceleration
    Andreas Sochopoulos, Esmeralda S. Whitammer, Nikolaos Tsagkas, João Moura, Michael Gienger, and Sethu Vijayakumar
    In ArXiv Preprint, 2026
    arXiv HTML
  2. w2vla_thumbnail.gif
    Decoupling the Declarative from the Procedural in Vision-Language-Action Models
    Nikolaos Tsagkas, Andreas Sochopoulos, Chris Xiaoxuan Lu, Oisin Mac Aodha, and Alexandros Kouris
    In ArXiv Preprint, 2026
    arXiv HTML
  3. afa_thumbnail_smaller.gif
    Attentive Feature Aggregation, Or: How Policies Learn to Stop Worrying about Robustness and Attend to Task-Relevant Visual Cues
    Nikolaos Tsagkas, Andreas Sochopoulos, Duolikun Danier, Sethu Vijayakumar, Alexandros Kouris, and Oisin Mac Aodha
    In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Pittsburgh, USA, 2026
    arXiv HTML Code
  4. temporal_thumbnail.gif
    The Temporal Trap: Entanglement in Pre-Trained Visual Representations for Visuomotor Policy Learning
    Nikolaos Tsagkas, Andreas Sochopoulos, Duolikun Danier, Chris Xiaoxuan Lu, and Oisin Mac Aodha
    In International Conference on Robotics and Automation (ICRA), Vienna, Austria, 2026
    arXiv HTML Code
  5. zipper_16fps_240.gif
    Fast Flow-based Visuomotor Policies via Conditional Optimal Transport Couplings
    Andreas Sochopoulos, Nikolay Malkin, Nikolaos Tsagkas, João Moura, Michael Gienger, and Sethu Vijayakumar
    In Conference on Robot Learning (CoRL), Seoul, South Korea, 2025
    arXiv HTML Code
  6. pvrobo_thumbnail_distractors_short.gif
    When Pre-trained Visual Representations Fall Short: Limitations in Visuo-Motor Robot Learning [Oral EAI@CVPR25]
    Nikolaos Tsagkas, Andreas Sochopoulos, Duolikun Danier, Chris Xiaoxuan Lu, and Oisin Mac Aodha
    In ArXiv Preprint, 2025
    arXiv HTML Code
  7. stablenode.gif
    Learning deep dynamical systems using stable neural ODEs
    Andreas Sochopoulos, Michael Gienger, and Sethu Vijayakumar
    In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024
    arXiv