Aerial robots can reach places nothing else can, but they are still poor at interacting with the physical world. Batteries limit how long they fly, and precise, autonomous physical interaction needs advances in control, perception, collaboration, and learning.
We are working on floating-base cooperative robotic manipulation, with aerial manipulation as one example. In this setting, we train aerial robots to perform contact-rich cooperative behaviours using only onboard sensing, such as RGB-D and inertial cues. Central themes include energy-aware autonomy, cooperation, and dexterous manipulation.
Multimodal robots have clear advantages, but their design space is largely unexplored. Instead of a single “optimal” configuration, we look for morphologies tailored to specific tasks, found by inverse design that links task requirements to morphology, actuation and aerodynamics. In parallel we study how fluid–structure interaction, such as passive thrust vectoring and proximity to walls and ceilings, can improve efficiency and control authority.
The work the lab builds on, from Caltech, ETH Zurich and EPFL. Each topic opens a page with its papers, figures and videos; the full list is on the Publications page.
Robots that fly and drive with the same body. We design morphobots that reuse one set of actuators across locomotion modes and transform in mid-air to land on their wheels, and we study the aerodynamics that make such transitions possible: ground effect close to touchdown, and rotor wake that can be redirected rather than lost.
