Open Cabinets

Diffusion-driven robot policy learning for reliable cabinet door opening.

A sleek, white robotic arm with articulated metal joints and a soft silicone gripper poised in front of a closed matte-gray kitchen cabinet door in a high-fidelity simulated environment. The cabinetry has clean, modern lines and subtle reflections on its surface, with a virtual marble countertop beneath. Cool, even studio-style lighting illuminates the scene, revealing fine textures on the arm’s brushed aluminum segments and the cabinet’s paint. The atmosphere feels precise and professional, emphasizing technical sophistication. Captured at eye level with a slightly wide frame, the composition centers the interaction point between gripper and handle, with a shallow depth of field softly blurring the rest of the kitchen. Photographic realism and a clean, modern aesthetic reinforce the robotics research context.
A photorealistic close-up of a robotic gripper gently wrapping around a slim, stainless-steel cabinet handle in a simulated kitchen. The gripper’s soft rubber pads show faint wear patterns, contrasting with the polished metal of the handle and the smooth, off-white cabinet surface. Neutral, diffused lighting from above casts soft shadows under the handle and emphasizes the subtle curvature of the gripper fingers. The mood is focused and analytical, highlighting precision manipulation. Shot with a macro, shallow depth of field, the point of contact between gripper and handle is tack-sharp while the surrounding environment fades into a smooth bokeh of kitchen textures. The image feels like a scientific documentation photograph, clean and minimal, perfectly suited for a professional robotics portfolio.

Teaching robots kitchen skills

Our CS188 final project trains a simulated robot arm to open kitchen cabinet doors using diffusion policy and action chunking.

Team Member

  • Jiawei Miao 4th year Math of Computation @UCLA
    Nannan Wang 3rd year CS @UCLA