Bucky
A cyber-physical cooking prototype — a KUKA arm that watches, learns and reproduces human gesture in real time.
- Brigid White, Aswin Lasingtan Pious
A two-week proof of concept at the University of Sydney DMAF Lab, developed in collaboration with Bosch.
Bucky asks whether an industrial robot can acquire cooking skill from its immediate environment — not from pre-scripted paths, but from watching a person cook. Live webcam input is understood and filtered, then amalgamated with predefined movement functions so the KUKA replicates cooking nuances. The enquiry sits at the intersection of cyber-physical systems, computer vision and fabrication robotics, with clear extensions into AEC manufacturing and assembly.
Workflow
Gestural transition → live webcam feed → TensorFlow + MediaPipe pose estimation → data synthesis and filtering in Processing → keypoint transfer as JSON → deserialization and plotting in Grasshopper → G-code generation via KUKA PRC.
Gesture becomes motion data; motion data becomes toolpath.
Mentors: Dagmar Reinhardt, Eduardo De Oliveira Barata, William McKee, and the DMAF support team.