Overview

Robotic manipulation of rigid objects has embarked on a remarkable trajectory, as demonstrated by increasingly advanced manipulation skills and real-world deployments. This progress has been enabled by improved perception pipelines, standardized 6DoF rigid object representations, mature planning, and control frameworks. However, real-world environments contain a diverse set of non-rigid objects: deformable ones, such as clothes, cables, food, and articulated ones, such as dishwashers, laptops, drawers, whose shape, topology, and behavior dynamically changes during interaction. Manipulating these objects remains a fundamental open challenge that requires advances across the entire robotic pipeline, from perception and representation to modelling, planning, and control.

This workshop focuses on robotic manipulation of non-rigid objects, with particular emphasis on deformable and articulated ones. We aim to bring together researchers working on complementary approaches that span end-to-end learning systems and modular pipelines, addressing challenges across the entire manipulation process.

The relevance of this workshop stems from the growing need for robots that can robustly operate in unstructured, human-centered, and natural environments, where interaction with non-rigid objects is unavoidable. Progress in this area is critical for applications such as household assistance, logistics, agriculture, healthcare, and industrial automation, where handling deformable and articulated objects remains one of the key bottlenecks to autonomy and real-world deployment. Currently, two dominant paradigms have emerged: end-to-end approaches that aim to learn manipulation policies from large-scale (often teleoperated) datasets, and modular pipelines inspired by the successes of rigid object manipulation. Each comes with distinct advantages and limitations in terms of robustness, generalization, data efficiency, and interpretability. By fostering discussion between researchers from both paradigms, this workshop seeks to identify common principles, open challenges, and promising directions toward scalable and reliable non-rigid object manipulation.