Research Areas

01

Robotics & control

swarms · mechanisms · embodied intelligence · estimation

We build simple robots and mechanisms to see how local rules give global function: light-seeking swarms that build and dig, a scissor linkage that morphs and writes, and an elastic strip that senses and actuates itself. We also work on optimal estimation and control.

›Background reading

02

Scientific ML

generative models · reinforcement learning · statistical physics

We use statistical physics and control theory to understand why machine learning works: the dynamics of generative models, reinforcement learning as control, and how models generalise.

Recent work

  • Statistical performance of large vision models

    Work in progress

  • Generalization of neural networks close to bifurcation points

    Work in progress

›Background reading

03

Behaviour

decision making · posture & locomotion · stigmergy · collectives

We model how animals, alone and in groups, act with limited information and noisy senses: a snake balancing upright, ant colonies choosing food through pheromone trails, bees recruiting with airflow, and people trading speed against accuracy when they bet.

›Background reading

04

Soft mechanics & geometry

instability · filaments & sheets · elastocapillarity · design

We study how thin sheets, filaments and droplets change shape when elasticity, surface tension and geometry compete, and how geometry can be used to design a response, as in popup structures and reconfigurable assemblies.

›Background reading