Diffusion policies in milliseconds, directly from the demonstration dataset and fast inference on a mobile CPU. Plus inference-time neural diffusion policy editing to steer actions or add new demonstrations to the policy

Inequality constrainted Model-Based Diffusion using interior-point inspired time-varying barrier functions to improve performance in highly constrained environments without using computationally expensive operations like projections.

An end-to-end differentiable simulation pipeline with orbital propagation and physically based rendering for on-orbit inspection operations to optimise inspection trajectories for image quality.

An appearance-aware impulsive trajectory optimisation using differentiable Lambert problem solver, using Blender-in-the-loop as an appearance model and using gradient-based optimisation.

We augment a conventional universal gripper with a computer-controllled audio exciter to show it can improve holding strength. We build an automated test rig to allow large-scale data collection to explore the effects of active vibration.