Nikolaos Tsagkas
Postdoctoral Researcher at the University of Amsterdam
I work on Physical AI: teaching robots how the world works and how to act in it. During my PhD at the University of Edinburgh, I studied how robots can reuse what foundation models already know about objects, so they only need to learn how to physically interact with them. Put simply: a foundation model already knows what a mug is. A robot still has to learn how to handle one.
What interests me most is what remains missing. Pixels can show a robot what a mug looks like, but not how heavy it is, how it slides, or when it will tip. These properties reveal themselves only when the robot acts, and every outcome, failures included, is evidence about how the world works. Learning from that loop of acting, observing and revising is what I'm now pursuing with Efstratios Gavves at the University of Amsterdam.
- 2025
- 2024
- 2023
- 2019
2026
Oct 2026
π Thrilled to share that I have accepted a PostDoc position at the University of Amsterdam with Prof. Stratis Gavves. Stay tuned for exciting research updates on Physical AI. See
you in Amsterdam! π³π±π
Sept 2026
π Our paper, Optimal Transport Q-Learning for Flow Policy Steering and Acceleration, has been accepted at CoRL, 2026. See you in
Austin, Texas, USA! πΊπΈπ
June 2026
π New paper published on ArXiv! You can access it here:
Decoupling the Declarative from the Procedural
in Vision-Language-Action Models.
June 2026
π Our paper, Attentive Feature Aggregation
or: How Policies Learn to Stop Worrying about Robustness and Attend to
Task-Relevant Visual Cues, has been accepted at IROS, 2026. See you in
Pittsburgh, USA! πΊπΈπ
Jan 2026
π Our paper, The Temporal Trap: Entanglement
in Pre-Trained Visual Representations for Visuomotor Policy Learning, has
been accepted at ICRA, 2026. See you in Vienna, Austria! π¦πΉπ
2025
Sep 2025
πΌ I am thrilled to announce that I have joined the Samsung Research Artificial Intelligence
Center as a research intern in Cambridge, UK. π¬π§
Aug 2025
π Our paper, Fast Flow-based Visuomotor
Policies via Conditional Optimal Transport Couplings, has been accepted at
the Conference on Robot Learning (CoRL), 2025. See you in Seoul, South Korea!
π°π·π
June 2025
π Our paper, Learning Precise Affordances from Egocentric Videos for Robotic
Manipulation, has been accepted at the IEEE/CVF International Conference on
Computer Vision (ICCV), 2025. See you in Hawaii, USA! πΊπΈπ
June 2025
π A short version of When Pre-trained Visual Representations Fall Short: Limitations
in Visuo-Motor Robot Learning has been accepted for oral and poster
presentation at the 6th Embodied AI Workshop at
CVPR 2025. See you in Nashville, Tennessee! πΊπΈπ
May 2025
π New paper published on ArXiv! You can access it here:
Fast Flow-based Visuomotor Policies via
Conditional Optimal Transport Couplings.
Feb 2025
π New paper published on ArXiv! You can access it here:
When Pre-trained Visual Representations Fall
Short: Limitations in Visuo-Motor Robot Learning.
2024
Oct 2024
π£οΈ I will be presenting our paper, Click to Grasp, during the Robot Vision IV
session from 09:00 to 10:00 on Fri 18 Oct at Room 4, IROS'24 in Abu Dhabi,
UAE.
Sep 2024
π£οΈ Our paper, Click to Grasp, will be presented at the BMVA Symposium: Robotics
Foundation & World Models in London.
Aug 2024
π New paper published on ArXiv! You can access it here:
Learning Precise Affordances from Egocentric
Videos for Robotic Manipulation.
Jun 2024
π Our paper, Click to Grasp: Zero-Shot Precise Manipulation via Visual
Diffusion Descriptors, has been accepted at the IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS), 2024. See you in Abu Dhabi,
UAE! π¦πͺπ
2023
May 2023
π Our paper, VL-Fields: Towards Language-Grounded Neural Implicit Spatial
Representations, has been accepted (spotlight) at the Workshop on
Effective Representations, Abstractions, and Priors for Robot Learning, ICRA
2023. See you in London, UK! π¬π§π
April 2023
π Our paper, Inference and Learning for
Generative Capsule Models, has been accepted at the Neural
Computation Journal, Volume 35, Issue 4.
2022
Sep 2022
π Thrilled to announce that I have been accepted at the
Centre for Doctoral Training in Robotics and
Autonomous Systems PhD programme π€π¦Ύ, at the University of Edinburgh, UK!
π΄σ §σ ’σ ³σ £σ ΄σ Ώ
2021
Sep 2021
π Thrilled to announce that I have graduated from the
University of Edinburgh with a MSc in Artificial Intelligence, with
Distinction!
2020
Sep 2020
π Excited to announce that I have been accepted at the
Artificial Intelligence MSc programme, at the University of Edinburgh, UK!
π΄σ §σ ’σ ³σ £σ ΄σ Ώ
2019
Nov 2019
πΌ Excited to announce that I have joined Ernst
& Young as a Data Scientist!
Aug 2019
π Excited to announce that I have graduated with a
Diploma of Electrical & Computer Engineering (BSc, MSc), 4th in my class, from
the University of Patras, Greece!
July 2019
π Our paper, On the Use of Deeper CNNs in Hand Gesture
Recognition Based on sEMG Signals, has been accepted (oral presentation) at
the 10th International Conference on Information, Intelligence, Systems and
Applications (IISA),. See you in Patras, Greece! π¬π·π
July 2019
π Honored to receive the "Best Student Innovation
Challenge Award" at IEEE Worldhaptics 2019 in Tokyo,
Japan, π―π΅ for our prototype "Gamified haptic UI for
increasing trust in autonomous vehicles"! Thank you Hyundai Motors
for sponsoring us! ππΉοΈ
Postdoctoral Researcher · Amsterdam, Netherlands
- Working on Physical AI with Prof. Efstratios Gavves.
- Funded by the Toyota Research Institute (TRI) and Toyota Motor Europe (TME).
Oct 2026 – Present
AI & Robotics Research Intern · Cambridge, UK
- Designed a novel Vision-Language-Action architecture, termed w2VLA, that enables compositional generalization, with Alexandros Kouris.
Sep 2025 – Dec 2025
PhD in Robotics & Artificial Intelligence ·
School of Informatics
- Thesis title: "What Robots Should Not Have to Learn: Declarative and Procedural Knowledge in Manipulation", advised by Prof. Oisin Mac Aodha (UoE) and Prof. Chris Xiaoxuan Lu (UCL).
- Funded by the Edinburgh Centre for Robotics through UKRI grant EP/S023208/1.
- Coordinator of the Edinburgh Vision Group (2026) and Teaching Assistant for Applied Machine Learning (2023–2025).
Sep 2022 – Dec 2026
MSc in Artificial Intelligence
- Graduated with Distinction.
- Worked on inference and learning for Generative Capsule Models with Prof. Chris Williams, later published in Neural Computation (2023).
Sep 2020 – Nov 2021
Diploma in Electrical & Computer Engineering
(integrated BSc & MSc) · Patras, Greece
- Graduated 4th out of 202 in the ECE class of 2019.
- Worked on real-time hand-gesture recognition from sEMG signals with Prof. Athanassios Skodras, presented at IISA 2019.
- Received the Best Student Innovation Challenge Award at IEEE World Haptics 2019 in Tokyo, for a gamified haptic interface for trust in autonomous vehicles.
- Teaching Assistant for Digital Signal & Image Processing Laboratory (2018–2019).
Sep 2014 – Aug 2019