Legged navigation & traversability
Where can a legged robot step, and how should it get there? Semantic and geometric traversability, footstep and path planning, world models.
Robotics · Cognition · Learning
Dimitrios Kanoulas
Professor in Robotics and AI, UCL Computer Science · UKRI Future Leaders Fellow · Head of the Intelligent Robotics Group and the RPL Lab
Robots are no longer short of bodies. What they lack is judgement. My group builds the cognitive layer — perception, mapping, and task/path planning — that lets legged robots work out where they are, what is around them and what to do next, in terrain and clutter too unpredictable for scripted behaviour. Our methods run on more than a dozen humanoid and quadruped platforms.
160+peer-reviewed publications
£2.4M+as Principal Investigator
26PhD students & postdocs supervised
15+robot platforms deployed on
Research themes
All projects →
Mapping, localisation & SLAM
Real-time metric-semantic mapping, structure-free topometric maps, visual localisation with few images, dense RGB-D SLAM on embedded devices.
Learning for loco-manipulation
Reinforcement, imitation and diffusion-based policies for agile gaits, parkour, grasping and whole-body loco-manipulation, with sim-to-real transfer.
Teleoperation & human–robot teaming
Wearable motion capture, VR and shared autonomy for legged manipulators in hazardous environments; usability evaluation.
Selected recent papers
All publications →Robots I have worked with
Also: Unitree A1, Go1, Go2, B1 + Z1 arm, Jueying, Summit-XLS, AgileX, MOCA, Pyrène and more.
Consulting, advisory and industry collaboration
I advise companies on legged locomotion, robot perception, manipulation and reinforcement learning, and I have served as scientific advisor to several robotics start-ups. If you are building robots that have to work outside the lab, let's talk.
How I can help →