Code Agents: AI-powered end-to-end solutions for flexible manufacturing
| Reference number | |
| Coordinator | Chalmers Tekniska Högskola AB - Chalmers Tekniska Högskola Inst f Industri- & materialvetensk |
| Funding from Vinnova | SEK 2 000 000 |
| Project duration | November 2024 - December 2025 |
| Status | Ongoing |
| Venture | Advanced digitalization - Enabling technologies |
| Call | Advanced and innovative digitalization 2024 - one-year projects |
Important results from the project
The primary objective of generating robot code was accomplished by employing two RAG models specialized in Python-based UR script and RAPID code generation. The resulting code was validated through a digital twin of a kitting operation. Additionally, a framework for AI-powered agents, referred to as Verification and Validation Agents, was developed to highlight their individual capabilities and define their expected roles.
Expected long term effects
The long-term effect of this project is divided into two phases. First, to expand Retrieval-Augmented Generation (RAG), an effective way to use large language models like ChatGPT with intellectual property-protected data, ensuring data integrity. Next, artificial intelligence will be used for robotic code generation as RAG models become more capable. This will enable individuals with limited robot programming experience to develop solutions for robotic applications.
Approach and implementation
The project was highly demanding and exceeded its budget for the partners. Still, all partners’ strong interest in AI code generation and Digital Twin validation kept work ongoing. Minor software implementation delays were resolved efficiently. Company visits and workshops involved various stakeholders from participating organizations. Communication channels stayed open, with bi-weekly meetings initially and monthly meetings later.