南京工业大学学报（自然科学版）2026，Vol.48Issue(1)：1-11,11.DOI:10.3969/j.issn.1671-7627.2026.01.001
物理人工智能:重塑材料化学研发的新范式
Physical artificial intelligence:a new paradigm for research and development in materials chemistry
摘要
Abstract
Traditional materials research and development often suffer from inefficiency due to the trial-and-error approach.This paper aims to explore the transformative role of Physical Artificial Intelligence(PAI)in reshaping the research paradigm of materials chemistry.This paper reviews the architectural features of PAI,focusing on how it integrates machine learning algorithms with embodied intelligent systems(robots)and real-world interactions.The paper also evaluates the application mechanisms of PAI in autonomous laboratories,intelligent synthesis,and characterization.The study identifies three major breakthroughs in PAI:the development of an autonomous laboratory for discovering high-throughput materials,which significantly improves experimental efficiency;the creation of an AI-guided robotic platform,enabling precise customization of nanomaterials and single-atom catalysts;and the advancement of neuromorphic devices and flexible sensors,providing a physical hardware foundation for embodied intelligence.Additionally,by combining computational models with physical feedback loops,PAI supports reverse material design based on target performance.PAI can shorten the time needed for discovering new materials and plays a key role in transitioning from an"automated"to an"intelligent"approach.Future studies include developing improved strategies for data generation,integrating hardware and software,and enhancing model interpretability.Addressing these will unlock the full potential of PAI in sustainable chemistry and advanced material innovation.关键词
物理人工智能/机器化学家/自动驾驶实验室/高通量实验/材料基因组Key words
Physical Artificial Intelligence/robot chemist/autonomous laboratory/high-throughput experiment/materials genome分类
化学化工引用本文复制引用
陈逸飞,杨文韬,朱熹..物理人工智能:重塑材料化学研发的新范式[J].南京工业大学学报（自然科学版）,2026,48(1):1-11,11.基金项目
国家重点研发计划(2022YFC2403500) （2022YFC2403500）
