Nature发文介绍依托同济建设的上海自主智能无人系统科学中心：力争站在人工智能创新的最前沿！
日前，Nature发表文章《Engineering intelligent systems with AI——Shanghai Research Institute for Intelligent Autonomous Systems leverages Tongji University’s strengths in autonomous systems and aims to stand at the forefront of artificial intelligence innovations.》，全面介绍了依托同济大学建设的上海自主智能无人系统科学中心，同济将通过整合来自上海、中国和全球的专业力量，努力将中心打造成为人工智能创新的全球枢纽。
原文详见阅读原文！
The information technology revolution is rapidly changing our lives. Breakthroughs in computing and telecommunications, based on information theory, the communication and control theory, Turing’s computational model, and powerful processors, are dramatically boosting productivity. Emerging artificial intelligence (AI) technologies are bringing a new information revolution.
As big research powerhouses look to AI to boost their technological innovation and economic development, China has also launched a plan to build up its AI industry. In Shanghai, China’s commercial, and science and technology hub, the establishment of Shanghai Research Institute for Intelligent Autonomous Systems by Tongji University, announced by the mayor in December 2018, is the natural result of ever-growing investment.
Aiming to accelerate AI research, the new institute is built upon Tongji’s traditional strengths in intelligent autonomous systems. “We expect to leverage the university’s interdisciplinary research resources to tackle cutting-edge research problems and cultivate high-end talent in AI,” said Jie Chen, president of Tongji University, and chief scientist of the newly established institute. “By integrating multilateral forces from Shanghai, China and worldwide, we will be a global hub of AI innovations.”
A hub for AI innovation
The institute focuses on three key issues in AI research — autonomous and sensing systems, emergent intelligence, and collaborative and swarm intelligence. Specifically, the institute is dedicated to harnessing the novel properties of super materials to develop devices that extend our senses, and with real-time collection of massive data, supporting autonomous decision-making based on multi-source big data. In intelligent cognition, the team strives to develop collaborative autonomous agents with cognitive capabilities, like those of humans, by simulating the neural mechanisms of cognitive behaviours, and to enhance the capabilities of processing big data and developing brain-like intelligence. Another focus of the institute is the collaboration between biological, mechanical, and electronic systems, as well as the mechanisms of self-organisation and emergence for multi-agent collaboration and decision-making. Researchers hope to optimise integration and collaboration between humans and machines.
By gathering a multidisciplinary group of talented researchers, and building research infrastructure, the institute will undertake major national science and technology innovation projects. With research centres on novel algorithms and devices, and on autonomous agents for aerial, underwater and land uses, as well as R&D platforms on advanced technologies and applications of unmanned systems, the institute is bound to advance breakthroughs in AI with its leading research.
A leader in smart technologies
As China’s pioneer in basic and applied research on intelligent autonomous systems, Tongji University researchers have made breakthroughs in the application of AI technologies, ranging from intelligent computing and distributed control, to smart transportation and cities, reported in leading international journals.
Tongji’s team has investigated novel sensors for a wide range of applications by exploring new sensing mechanisms, theories, sensor structures, materials, and effective strategies for enhancing sensing performance. Their works on the control of molecular structures, semiconductor microstructures, and interface charges effects have led to the development of artificial skin, flexible photo sensing membrane and high-performance electronic nose, which can mimic various sensing functions of human bodies. By revealing a mechanism for multi-scale sensing, they have proposed new composite devices based on nanotube films and piezoelectric wafers. The team’s idea for a small hyperspectral sensing system is a world first.
Working on integration between biological, mechanical, and electronic systems, another team has created a worm-like bionic robot. Linking motor neurons and biomorphic structure with the environment, the robot is optimised for improved movement. The team has also made breakthroughs in designing bionic joints by integrating bionic sensing and robotic control.
Based on the natural interaction and autonomous learning for developing robot intelligence, the team studying autonomy and interaction has built a model for robot development. Based on big sensing data, the model shows effective skill learning by robots without any human intervention.
In heterogeneous systems and decision-making, a Tongji team has invented methods for automatic, real-time traffic flow analysis to assess and control risks, ensuring safe driving. They have independently developed China’s first intelligent traffic system, and first testing site for intelligent connected vehicles. Simulations of realistic urban and rural settings for intelligent connected vehicles, featuring highway and country roads, have enabled real-world testing for different types of driverless vehicles. The team has also participated in the construction of a photovoltaic roadway fitted with solar panels and the intelligent traffic planning for 60% of China’s subway lines.
In smart city design, Tongji’s urban planning research team has used machine learning to collect and mine sample data from 13,810 cities across the world. By training the machine to categorize these cities into different urban development phases, they have, in a world first, defined seven major categories of urban growth. Precise identification of a city’s development stage will enable capturing of the driving forces or problems underlining its development, so land use layout and construction can be calibrated with the city’s needs.
The group studying brain-like intelligence and bionic applications is dedicated to advancing AI technology through understanding of human neural networks. They have pioneered the use of single-cell transcriptomics to reveal the signal pathway to activate dormant neural cells, and built a new model for neural network structures. Using brain-wide association analysis, integrating imaging technologies and big data analysis, they have identified brain regions and connectivity significantly associated with autism.
Working on multi-agent collaboration, another Tongji team has established the theoretical framework that balances exploration and exploitation to improve machine learning. Led by Chen, who is well known for his research on control science and engineering, the research team has also disclosed how the uncertainty of information transmission, and the unmodeled dynamic environment influence the stability of collaborative control. They have proposed methods for multi-criteria collaboration of complex systems and multi-objective decision-making to improve system operations. Chen’s research on multi-criteria optimization and control of complex systems in dynamic environments, and collaborative control of multiple intelligent agents has wide industrial applications.
Dedicated to creating world-class research environment, and attracting and fostering top talent to lead AI research, the institute aims to forge university-industry collaborations to improve autonomous systems. “We would like to contribute to AI technologies with our expertise,” said Chen, “We hope more talented researchers will join our endeavour.”
人工智能已经是全球学术领域的热点，科学家们迫切希望寻求通用的人工智能技术在复杂环境、大数据背景下为人类幸福生活提供科学根源和创新思想。自上世纪香农的信息论、维纳的控制论和图灵计算模型为人类带来信息和科学技术的革命，带来了诸如工业处理器、逻辑控制器、路由器、计算机等科技革命，推动信息在农业、制造业等生产率大幅提升，全球GDP在短短半个世纪增长了40多倍。全球的人工智能面临新一轮的科技革命，也将为信息技术革命和社会经济发展的提供新的核心推动力。
依托同济大学建设的上海自主智能无人系统科学中心（Shanghai Research Institute for Intelligent Autonomous Systems）充分发挥多学科优势，广聚上海及全球精英，致力于解决人工智能重大前沿科学难题，培养紧缺的人工智能高端人才，努力建设成为人工智能领域的重要创新中心。
中心首席科学家、中国工程院院士、同济大学校长陈杰教授，为中国自动化学会副理事长、中国人工智能学会常务理事，长期从事控制科学与工程等相关学科领域的教学与科研工作，主要研究方向是动态环境下复杂系统的多指标优化与控制、多智能体协同控制等。
中心立足人工智能重要科学前沿，围绕“自主与感知”“智能与涌现” “协同与群智”三大关键科学问题，聚焦“超材料感知”“多尺度融合”“类自然计算”“自主智能体”“生机电共融”“自主与交互”“异构与决策”“多体与协同”“类脑与仿生”等九大研究方向，着力汇聚世界一流人才团队，建设重大科技基础设施，搭建下一代器件和算法研究中心、陆海空天自主智能体及多体集成研究中心、无人系统先进技术研发功能平台、无人系统应用研发功能与转化平台，承接国家科技创新2030—重大项目等，努力持续产出引领性原创大成果，加快推动在人工智能领域的重大突破和跨越。
同济大学是国内自主智能无人系统研究的先行者与实践者，在基础研究和应用转化领域已有多年积累和丰硕的成果，在智能感知、智能计算、自主控制、分布式控制、智慧城市、智能交通、智能农业、智能制造、智能医疗等方向取得了一系列重要突破，并在多个领域取得了重要创新的成果，相关成果均有多篇高水平研究论文发表于Nature、IEEE Trans. SMC、IEEE Trans. Auto. Control、IEEE Trans. on CSVT等国际著名学术杂志。
在超材料感知方向，相关团队创建了力、热、声、光、电等多物理场和化学成分融合感知的新理论，开创“物理量+化学量”联合探测的新原理、新材料、新器件和新系统，国际上首次提出基于准三维亚波长结构材料的小型化超光谱感知系统，创制基于分子结构调控、微结构调控和界面电荷调控的新型电子鼻和人工皮肤等。揭示了C-PZT多尺度融合传感机理，提出了新型纳米管薄膜-压电晶片复合器件设计方法。
在生机电共融方向，同济大学教授团队首次研制共融移动环境-驱动-结构一体化的优化蠕虫型仿生机器人，建立了微结构尺寸和表形的湿吸附力综合修正模型，突破了仿生机构关节、感知及控制单元机电一体化设计理论方法。
在自主与交互方向，研究团队提出了基于自然交互与自主学习的机器人心智发育方法，建立了基于海量类人感知数据的机器人心智发育模型，突破了无人干预条件下机器人高效技能学习。
在异构与决策方向，同济大学教授创建了主动安全实时车流运行风险研判、控制理论与方法、自主研发全国首个智能交通系统（ITS），建立全球首条光伏路面试验段、“智能网联汽车测评基地”，涵盖城市区域、快速路、乡村和越野区域等智能网联汽车典型应用区域，可用于各等级、各类技术方案的自动驾驶汽车的场地测试。完成了全国60%的地铁线路运行智能规划。
在多体与协同方向，建立了探索-开发权衡的理论框架，发现了状态有限时间一致性与拓扑结构间的本质联系，揭示了信息传输不确定性以及未建模动态环境对协同控制稳定性影响的内在规律，提出复杂系统的多目标协同与多目标决策方法
在类脑与仿生方向，揭示了早期胚胎发育组蛋白修饰建立与调控机理，率先使用单细胞转录组技术揭示室管膜静息态神经细胞激活的信号途径，建立了全新神经网络结构模型。提出全新的脑功能影像大数据分析方法和脑影像学大数据分析技术，成功发现与自闭症显著关联的脑区之间的连接。
在智慧城市领域，研究团队聚焦世界城市的深度学习和人工智能城市推演，首次发现并定义七大城市发展类型，全息感知城市数据，揭示全球13810个城市时空发展规律，精准诊断现有城市运行动力需求和发展瓶颈，自动完成用地布局和建设时序推演过程。中心建设以汇聚人才、创新体制为核心，致力于形成国际一流的科研环境和模式，培养并引进一批拔尖和国际大师级领军人才，努力建设成为集人工智能理论基础、自主无人系统核心技术为一体的世界一流产学研基地和示范中心，为世界人工智能科学与技术贡献“中国智慧”“中国方案”。