Finding Common Language for Human-machine Collaboration

28.6.2025 - 6.7.2025 | ICD Workshop DigitalFUTURES 2025, Tongji University, Shanghai, China

Xiliu Yang, Samuel Slezak, Lasath Siriwardena, Felix Amtsberg, Achim Menges

DigitalFUTURES 2025 - Finding Common Language for Human-machine Collaboration

This workshop integrates embodied interactions in augmented reality (AR) with the reasoning capabilities of large language models (LLMs) for interactive design and robotic fabrication. Participants worked with a HoloLens 2 AR headset and a UR10-e collaborative robot, adapting an agentic framework (MAVE) to craft custom workflows that connect AI agents, robots, and human interactions. 

Over seven days, participants learned how AR grounds LLMs in design and fabrication contexts, built custom agents on the MAVE framework, and applied these tools to human-robot collaborative scenarios. The workshop began with an introduction to robotic fabrication and the use of AR interfaces with LLM agents. Days two and three cover the MAVE software architecture and hands-on tutorials with the codebase. Students then developed concepts for interactive design and fabrication workflows, considering how agents with different domains of expertise can contribute towards novel outcomes. Days four through six were dedicated to system development, prototyping, and documentation. The workshop concluded on day seven with project presentations and an exhibition of the results throughout CDRF 2025.

© Video and photo credits: ICD/IntCDC University of Stuttgart  

Workshop Projects

Talk-Create-Build proposed a conversational assistant for furniture design. The Supervisor agent conducts an interview to gather user requirements on functional aspects (weight, load capacity, budget) and aesthetic preferences (complexity). These responses map to parameters controlling density, height, and twist of layered assemblies generated in Grasshopper. (with Yaqi Chen, Jiayi Liu, Ahmad Mansuri, Xiaoye Ren, Jingwen Song)

Digital_futures_2025_Group1 ©
Hero Image of the Talk-Create-Build project. The image shows a person interacting with AR interface which she uses to collaborate with the robot and a agent to build furniture.

Sync-Point developed an assembly balancing game where human and robot collaborate to add elements to a structure without causing collapse. The human places a beam in AR, the Design agent computes a balancing solution, and both parties synchronise their placements. A new tool analyses beam dimensions and positions to determine valid configurations. (with Yichu Liu, Xuanyu Nick Lu, Han Nan, Haoyi Xu, Hanyu Xue, Yunfan Zhao)

Digital_futures_2025_Group2 ©
Hero Image of the Sync-Point project. The image shows a person playing an interactive balancing game with a robot using an AR environment.

Orchestrating Chaos implemented a workflow for designing with waste materials. A new Material agent catalogues as-found cardboard offcuts, capturing geometry through AR control points and qualitative notes on material condition. The Design agent uses this catalogue to run Kangaroo physics simulations that generate folded configurations. (with Ruoxi Li, Julian Tesche, Cheng Zeng, Chi Zhang, Zhiyuan Zhang)

Digital_futures_2025_Group3 ©
Hero Image of the Orchestrating Chaos project. The image shows multiple pople interacting with robot and AR holograms in order to design a structure from waste material.

Instructors

Xiliu Yang, Samuel Slezak, Lasath Siriwardena, Felix Amtsberg, Prof. Achim Menges

Technical and Administrative Support:

Xingjie Xie (CAUP Tongji)

 

Participants

Yaqi Chen, Jiayi Liu, Ahmad Mansuri, Xiaoye Ren, Jingwen Song, Yichu Liu, Xuanyu Nick Lu, Han Nan, Haoyi Xu, Hanyu Xue, Yunfan Zhao, Ruoxi Li, Julian Tesche, Cheng Zeng, Chi Zhang, Zhiyuan Zhang

Financal Support and Sponsoring

DigitalFUTURES

To the top of the page