About

Robotic assembly has traditionally focused on well-specified industrial settings with fixed parts, structured environment and predefined goals. However, emerging applications in creative designs, household robotics, and human-robot collaboration highlight the need and interest in customized solutions for assembly tasks.

In these scenarios, the problem is often underspecified or ambiguous: the target assembly may be provided only as a 2D sketch, silhouettes, semantic descriptions or high-level concepts. Additionally, the available components may consist of generic primitive shapes or irregular non-exact corresponding parts. Solving such problems requires reasoning beyond classical task planning: the system must be able to abstract solutions, by inferring missing constraints, consistent geometry and physically stable assemblies under uncertainty. Furthermore, in assembly, the contact-rich nature of the task poses yet another challenge in motion planning and execution.

This workshop aims to address those challenges, by bringing together researchers working on assembly under uncertainty, with a focus on:

• Assembly design in under specified targets;

• Geometric and physical reasoning from incomplete information;

• Integration of Large Language Models (LLMs) as general reasoning agents for assembly planning under uncertainty;

Addressing these challenges could unlock new strategies that extend robotic assembly beyond well-defined manufacturing domains, enabling robots to reason about and perform open-ended assembly tasks in uncertain and creative settings.

Invited Speakers

Organizers