My research sits where machine learning, robotics, and control meet human-machine interaction, from brain-computer interfaces to learning-based manipulation and legged robots. I care most about work that eventually runs on real hardware, or real signals, in unstructured environments.
Six interconnected themes that have guided my research since 2011. Every project touches at least two of them.
Decoding intent from brain signals in real time is where my research began. I work on classifying motor imagery from EEG, including the NF-EEG and IS-EEG models developed during my PhD, and on building them into a real-time system that can control computers and external devices.
Closing the physics gap between simulation and the real world is a central challenge in robot learning. My work investigates domain randomization schedules, residual physics learning, and adaptive system identification to make policies robust to unmodeled dynamics, tested on manipulators and legged platforms.
A good controller starts with an honest model of how a system moves. I study model-based and learning-enhanced control, including model predictive and adaptive methods, built on accurate dynamic models of mechanical systems. The goal is controllers that stay stable and precise when conditions change.
Reliable manipulation depends on combining what a robot sees with what it feels. I work on vision- and tactile-guided manipulation, sensor fusion for contact-rich tasks, and AI-enabled soft grippers that adapt to objects they have never encountered.
Machine learning is most valuable when it answers a real engineering question. I apply it to condition monitoring, such as detecting tool wear from vibration signals, and to LLM-based analytics that turn academic and administrative data into detailed insights and recommendations.
Simulation can shorten the path from concept to physical part. I combine finite element analysis, topology optimization, and additive manufacturing to design components that are tested virtually before they are built, so prototypes arrive closer to the final design.
Research projects in various stages, from data collection to paper submission. Most are collaborations with researchers in Europe and North America.
We are equipping a UR5e arm with a custom tactile sensor array and investigating how force-torque signals can close the loop on sub-millimetre peg-in-hole tasks that are currently beyond the precision of vision-only policies. Data collection is ongoing at the ITU Robotics Lab.
A systematic empirical study investigating what fraction of sim-to-real transfer failures can be attributed to (a) inaccurate physics parameters, (b) sensor simulation errors, and (c) distribution shift in visual inputs. We are running paired experiments across Isaac Gym and MuJoCo against physical UR5e trials to produce a variance decomposition across failure modes.
Following our ICRA 2023 work on efficient EfficientDet variants, this project investigates whether self-supervised contrastive pre-training on unlabelled factory floor imagery can reduce the labelled data requirement for defect detection by an order of magnitude. Paper currently in writing; targeting IEEE RA-L submission by Q3 2026.
We are investigating whether a quadruped locomotion policy conditioned on a learned latent terrain prior — estimated online from proprioceptive history alone — can generalise zero-shot to terrain categories unseen during training. Building on prior work with ANYmal-C; current experiments running in Isaac Gym with ETH ASL.
Graduate
Supervision
Undergraduate
My work has appeared at the leading robotics and machine learning conferences and journals.
The goal of research is not to produce papers. It is to understand something well enough that you could build a system that works, whether it's a controller, a learning algorithm, or a brain-computer interface. Papers are the record of that understanding. If the system can't be built, the understanding isn't complete.
If your work overlaps with mine, I'm always interested in collaborating, especially on projects that involve real hardware or real signals.