A research team from Tsinghua University has published a groundbreaking paper in the Journal of Manufacturing System, proposing the Embodied Intelligent Industrial Robotics (EIIR) framework — the first complete concept and technical architecture for bringing embodied AI to industrial manufacturing.
Led by Professor Feng Pingfa and Associate Professor Zeng Long from Tsinghua Shenzhen International Graduate School (Tsinghua SIGS), the paper defines the three evolutionary stages of industrial robotics: the Automation Era (1960s-1980s), the Perception Era (1980s-2020s), and the Embodied Intelligence Era (2020s-present). The EIIR framework represents the third wave, where robots transition from executing fixed programs to understanding factory environments, planning tasks autonomously, and executing precision operations.
The team identified a critical gap: existing large language models trained on internet data cannot handle industrial scenarios that require knowledge of CNC tolerances, production line cycle times, and cleanroom standards. Their solution is a knowledge-driven approach requiring three types of knowledge: general knowledge (language understanding), workspace knowledge (factory layout and constraints), and operational knowledge (product structures, process flows, and quality requirements).
The paper introduces a five-module technical architecture combining a world model, high-level planner, low-level controller, simulator, and physical system — working in concert to enable factory-ready embodied intelligence.
Crucially, the research has already moved beyond theory. Through industrial partner Guangzhou Fuwei Intelligence, the EIIR technologies have been deployed on real production lines serving Tesla (North America), Foxconn, and Biel Crystal. The product lineup includes ICR series composite robots for CNC machine tending, Fuzhi series humanoid robots for automotive assembly, and delivery robot-AMR collaborative systems for logistics centers.
The paper is published in Journal of Manufacturing System (CAS Zone 1, Impact Factor 14.2), with first author Zhang Chaoran (Tsinghua SIGS master student) and corresponding author Associate Professor Zeng Long. Funding was provided by the National Natural Science Foundation of China, Guangdong Natural Science Foundation, and Shenzhen Major Science and Technology Projects.
