PinchBot: Long-Horizon Deformable Manipulation with Guided Diffusion Policy

Alison Bartsch¹ Arvind Car¹ Amir Barati Farimani¹

¹Carnegie Mellon University Mechanical Engineering

[Paper] [Dataset] [Hardware CAD] [Github]

Abstract

Pottery creation is a complicated art form that requires dexterous, precise and delicate actions to slowly morph a block of clay to a meaningful, and often useful 3D goal shape. In this work, we aim to create a robotic system that can create simple pottery goals with only pinch-based actions. This pinch pottery task allows us to explore the challenges of a highly multi-modal and long-horizon deformable manipulation task. To this end, we present PinchBot, a goal-conditioned diffusion policy model that when combined with pre-trained 3D point cloud embeddings, task progress prediction and collision-constrained action projection, is able to successfully create a variety of simple pottery goals.

12cm Goal

Extending to Air-Dry Pottery: New Material & Variable Volumes

Our dataset and experiments were both conducted with reusable Playdoh clay with a constant volume. However, we deployed our trained PinchBot policy for the 8cm and 10cm diameter pottery goals using air-dry clay with variable volume to create a small set of permanent robot-created bowls. Each pot was varnished to better preserve them. Some images of the final pottery are shown below. Interestingly, this air dry clay required more force for each squeeze but because PinchBot's action space is controlling for the final finger squeeze distance, the policy was able to handle this change. Additionally, the firmer clay resulted in even softer indentations and surface smoothness as compared to the Playdoh pottery.

Bibtex

@article{bartsch2025pinchbot,

title={PinchBot: Long-Horizon Deformable Manipulation with Guided Diffusion Policy},

author={Bartsch, Alison and Car, Arvind and Farimani, Amir Barati},

journal={arXiv preprint arXiv:2507.17846},

year={2025}}

Page updated

Report abuse