我院与美的集团AI研究院共探智能家居产学研融合新路径

Fudan University IIINN and Midea Group AI Research Institute Explore New Pathways for Industry-Academia-Research Integration in Smart Home

2025年6月19日上午,复旦大学智慧纳米机器人与纳米系统国际研究院(以下简称“我院”)与美的集团AI研究院成功举行合作交流研讨会。本次会议特别邀请到微软研究院首席科学家陆子恒参会。双方以“推动智能家庭机器人技术革新”为核心目标,围绕校企联合实验室共建、产学研协同路径创新等关键议题展开深度探讨,旨在打通从前沿科研到智能家居产业应用的创新链。

研讨会上,我院副院长黄高山教授向美的集团AI研究院代表介绍了我院在AI前沿领域的系列突破性成果:开发了微纳机器人强化学习控制框架,解决复杂陆地环境中多自由度协同控制难题,实现无人为干预的自主运动编程,获《PNAS》报道;在光谱成像领域,创新性提出AI驱动的高精度光谱重构算法,有效解决欠定问题,实现高精度光谱重建,研究成果发表于《PNAS》;突破微纳米传感系统的空间多维信息串扰瓶颈,基于深度神经网络显著提升分辨率,实现入射光角度探测,核心技术获《Nature Communications》刊发;聚焦多模态生物传感信号有效识别,基于Transformer神经网络训练实现对多种微表情传感信号的精准鉴别,相关成果发表在《Device》。

美的AI研究院HR刘思雨、人机交互算法负责人高羽分别介绍了AI研究院的战略定位与核心成果。作为美的集团AI技术攻坚的引擎,美的AI研究院汇聚了全球数百名顶尖人才,依托上海、顺德双研发中心,深耕大模型、语音交互、边端智能等前沿领域,致力以技术重塑家电产业格局。其自主研发的中国首个家电垂域大模型“美言”,已服务于超1.2亿家庭,覆盖设备量1.16亿台;其在超低信噪比环境下的语音交互技术处于行业领先,自研AI边缘语音芯片FL119,支持空调、冰箱、洗衣机等全品类智能家电量产,可实现云-边-端的AI服务协同。

微软研究院首席科学家陆子恒在会上结合双方的交流内容,从产业实践与研究前沿融合的角度提供了宝贵的专业见解。

我院执行院长梅永丰教授出席会议并强调,期望以本次研讨会为契机,实质性推进校企联合实验室的建设。他指出,双方应重点围绕智能机器人、纳米系统与AI融合等方向寻求创新突破,发掘高价值合作接口,将实验室打造成产研协同的关键枢纽,加速AI技术在研发、制造、服务等核心场景的深度渗透与应用落地。

AI研究院人力总监邓健丽表示,未来希望能与复旦大学共建覆盖联合技术研发、高端人才共育、技术成果转化的多维长效合作机制。基于本次洽谈,美的集团将评估并推进AI Lab与复旦的深度协同计划,期待引入复旦顶尖科研力量,共同突破AI技术应用难点。

目前,我院正将前述核心技术积极整合并应用于新一代智能实验室的建设中,依托大模型与具身智能技术,拟构建具备自主决策能力的微纳加工全链条控制系统,推动纳米机器人从基础研究快速转化至产业化应用。




Fudan University IIINN and Midea Group AI Research Institute Explore New Pathways for Industry-Academia-Research Integration in Smart Home



On the morning of June 19, the International Institute for Intelligent Nanorobots and Nanosystems at Fudan University (hereinafter referred to as IIINN) successfully held a cooperation exchange seminar with Midea Group’s AI Research Institute. Dr. Ziheng Lu, Principal Scientist from Microsoft Research, attended as a special guest. Focusing on the core goal of advancing technological innovation in smart home robotics, the two parties engaged in in-depth discussions on key topics, including the joint establishment of university-enterprise laboratories and innovation in industry-academia-research collaboration pathways, aiming to bridge the innovation chain from cutting-edge research to smart home industrial applications.

At the seminar, Professor Gaoshan Huang, Deputy Director of IIINN, introduced a series of breakthrough achievements by the institute in AI frontier fields:

Development of a reinforcement learning control framework for micro-nano robots, which resolves multi-degree-of-freedom coordinated control challenges in complex terrestrial environments and enabling autonomous motion programming without human intervention (PNAS);

An AI-driven high-precision spectral reconstruction algorithm in spectral imaging, which effectively addresses underdetermined problems to achieve high-accuracy spectral reconstruction (PNAS);

A breakthrough in overcoming spatial crosstalk bottlenecks in multi-dimensional information for micro-nano sensing systems, which significantly enhances resolution via deep neural networks to enable incident light angle detection (Nature Communications);

Precise identification of multimodal biosensing signals through Transformer neural network training, which enables accurate discrimination of micro-expression sensing signals (Device).

Ms. Siyu Liu (HR) and Mr. Yu Gao (Head of Human Computer Interaction Algorithm) from Midea AI Research Institute elaborated on the institute’s strategic positioning and core achievements. As the engine for Midea Group’s AI technological breakthroughs, the institute has gathered hundreds of global top talents. Leveraging dual R&D centers in Shanghai and Shunde, it focuses on frontier fields like large models, voice interaction, and edge intelligence, striving to reshape the home appliance industry through technology. Its self-developed Meiyan – China’s first large model specialized for home appliances – serves over 120 million households and covers 116 million devices. Its voice interaction technology in ultra-low signal-to-noise ratio environments leads the industry, and its self-developed FL119 AI edge voice chip supports mass production of smart appliances (e.g., air conditioners, refrigerators, washing machines), enabling cloud-edge-end AI collaboration.

Dr. Ziheng Lu, Principal Scientist at Microsoft Research, provided valuable insights on integrating industry practices with research frontiers based on the discussions.

Professor Yongfeng Mei, Executive Director of IIINN, emphasized leveraging this seminar to substantively advance the establishment of a university-enterprise joint laboratory. He stressed that both parties should pursue breakthroughs in smart robotics, nano-systems-AI integration, and other key areas, identifying high-value collaboration interfaces to transform the lab into a pivotal hub for industry-research synergy. This will accelerate the deep integration and application of AI technologies in R&D, manufacturing, and service scenarios.

Ms. Jianli Deng, HR Director of the AI Research Institute, expressed hopes to establish a multidimensional, long-term cooperation mechanism with Fudan University, covering joint technology R&D, high-end talent cultivation, and technology transfer. Following the talks, Midea will evaluate and advance deep collaboration plans between its AI Lab and Fudan, aiming to integrate the university’s top-tier research capabilities to overcome AI application bottlenecks.

IIINN is now actively integrating the aforementioned core technologies into its next-generation smart laboratory. Powered by large models and embodied intelligence, it aims to build a full-chain micro-nano processing control system with autonomous decision-making capabilities, accelerating the industrialization of nano-robotics from fundamental research.










 

撰稿:杨帆

校稿:程绍平 陈茜

审核:梅永丰 黄高山