Against the backdrop of rapid advances in digital and intelligent technologies and the deepening implementation of the Chinese path to modernization, understanding modern infrastructure and the construction of its system from a systemic perspective has become a key issue for high-quality socio-economic development and the modernization of national governance. This study conceptualizes modern infrastructure as a comprehensive technological-institutional system formed through the integration of physical entities, cognitive intelligence, and institutional rules. Through technological supply, scenario responsiveness, and rule embedding, modern infrastructure continuously expands its capacity to support innovation and governance. It exhibits four intrinsic characteristics: system-wide perception enabled by digital-physical isomorphism, model-driven cognitive iteration, resilient self-adaptation enabled by endogenous regulation, and scenario-driven ecosystem co-creation. Building on a collaborative evolution model of "forward driving‒reverse guidance‒lateral regulation", this study elucidates three mechanisms governing the interaction between modern infrastructure and socio-economic systems: the value-creation mechanism from factor support to the incubation of new quality productive forces, the intelligent iteration mechanism from passive use to demand-driven redefinition, and the institutional governance mechanism from external constraint to resilient adaptation. At the system level, it proposes a construction pathway in which strategic orientation clarifies development priorities, structural optimization strengthens network coordination, mechanism innovation ensures stable operation, and feedback calibration enhances system adaptability. The study can provide a reference for deepening the theoretical understanding of modern infrastructure and advancing the coordinated construction of a modern infrastructure system.
The new-type power system exhibits the characteristics of high proportions of renewable energy and power electronic equipment. From the perspective of energy security, it is necessary to identify the key issues faced by the new-type power system in terms of technological evolution, structural transformation, and market mechanisms, and propose actionable response strategies. Considering the development status of current power systems, this study analyzes the safety risks associated with new-type power systems from five aspects: power supply, power operation, power infrastructure, industrial chain, and extreme scenarios. A power security support system focusing on power supply guarantee and restoration, flexible mutual assistance, and comprehensive defense is proposed. Specifically, the study summarizes the issues related to power supply guarantee and recovery under extreme events as well as the reliable operation of power systems with a high proportion of renewable energy. In terms of flexible mutual assistance, related issues include the configuration of a large power grid amidst the development of marine‒land renewable energy, flexible resources for new-type power systems, entry of renewable energy into the market, and establishment of a national unified power market. Regarding comprehensive defense, related issues involve digital security protection in the context of distributed resource aggregation, adaptive security defense against strong uncertainties, and cross-energy, cross-domain security defense under energy system coupling. Furthermore, development ideas for future new-type power systems are proposed from the perspectives of enhancing power supply and restoration capabilities, strengthening flexible mutual assistance in power supply, and establishing a comprehensive defense system. These efforts aim to facilitate the safe, stable, and high-quality transformation of the power system, and provide effective support for energy security and the transition to a green, low-carbon energy system.
Fostering an open, shared, and secure unified data market is a pivotal task for advancing the construction of Digital China and for serving the new forms of the smart economy. The new data infrastructure serves as the core carrier for supporting data element circulation and directly determines the effectiveness of data value release. Currently, the development of China's data infrastructure still faces deep-seated challenges. The root cause lies in the longstanding constraints of the "impossible trinity" of security, cost-effectiveness, and compatibility, alongside difficulties in synergizing existing resources with new development. This study summarizes the fundamental engineering challenges in constructing new data infrastructure, and elaborates on the vision, basic principles, and theoretical support elements for its development. Moreover, it innovatively proposes a "cloud‒network integrated hyper-convergence" architecture for data infrastructure, based on the theory of polymorphic intelligent network environment. Additionally, it employs a technical logic of "elementalizing network resources" and "dynamic resource aggregation," empowered by AI-enabled generative networks, to generate dedicated, scenario-adaptive network modalities on demand. Drawing on live network demonstrations, this study elaborates on the design philosophy and key technologies of the new data infrastructure architecture, and summarizes a synergistic construction model of "stock efficiency + incremental optimization." Practice shows that this architecture can provide an engineering solution to the "impossible trinity" and the stock‒increment synergy dilemma. Finally, the study proposes recommendations covering engineering advancement, technological innovation, platform support, disciplinary development, and institutional safeguards, aiming to provide theoretical and practical insights for building a nationally unified data market.
The intelligent economy is shifting demand for computing power from point-based, static, and general-purpose forms to scenario-oriented, real-time, heterogeneous, and low-carbon forms. As a result, the construction of new computing power networks is gradually transitioning from resource expansion to collaborative operation across the industrial chain. From the perspective of industrial chain collaboration, this study analyzes the structure, mechanisms, and pathways of new computing power networks in the context of the intelligent economy. It addresses problems including inadequate unified management and control of computing power resources; inadequate coordination of networking and computing; inefficient integration of computing power, electricity, and carbon; regional mismatches in supply and demand; insufficient compatibility between upstream and downstream sectors; and imperfect rules for ecosystem governance. The study argues that the industrial chain of new computing power networks comprises an upstream basic supply layer, a midstream platform operation and scheduling layer, a downstream application layer, and four horizontal support systems related to network, energy, security, as well as standards and policies. Its collaborative operation takes the form of a three-dimensional coupling process among factors, space, and actors. Core tasks include the integration of heterogeneous computing power, coordination of networking and computing, computing‒electricity‒carbon synergy, regional computing power collaboration, matching supply and demand while cultivating upstream and downstream linkages, and ecosystem governance collaboration. Computing power operation agents can perform technical operation functions such as resource sensing, task scheduling, contract performance monitoring, risk early warning, and decision support within established rules and authorized boundaries, and they constitute important support for enhancing collaboration efficiency across levels, regions, and actors. In the near term, priority can be given to resource collaboration as well as coordination of networking and computing. In the medium term, efforts can be intensified to deepen supply‒demand, computing‒electricity‒carbon, and regional collaboration. In the long term, it is advisable to improve standard interfaces, transaction rules, security responsibilities, reputation evaluation, and international cooperation mechanisms, thereby forming a systematic pathway for the industrial chain collaboration of new computing power networks.
Against the backdrop of climate change, water network systems face systemic risks involving multiple overlapping hazards and cross-scale transmission. Traditional analysis paradigms centered on engineering structural safety are insufficient to meet the security requirements of complex mega-systems under changing environmental conditions. Starting from the current status of China's water network development, this study reviews major achievements, identifies key challenges in building a safe and resilient modern water network, and analyzes existing gaps in the basic theories and methods of water network resilience. Considering the multi-objective, multi-level, and multi-element coupling characteristics of water network systems, and focusing on the three core functions of water supply, flood control, and ecological protection, this study defines the scientific connotation of a safe and resilient modern water network and proposes a corresponding theoretical framework. On this basis, a strategic framework for developing a safe and resilient modern water network is established. Strategic objectives are proposed around cascading risk prevention and control, functional maintenance, recovery and reconstruction, and digital-intelligent iteration. An implementation pathway is clarified through the coordination of engineering systems, network structures, dispatching mechanisms, and digital-intelligent systems. A closed-loop operation mechanism that integrates perception, analysis, decision-making, execution, feedback, and optimization is also recommended. Furthermore, from the three scales of macro-level planning, meso-level dispatching, and micro-level control, this study proposes a multi-scale collaborative promotion strategy for constructing a safe and resilient modern water network. Systematic development recommendations are also provided from five aspects: engineering foundation, dispatching capacity, scientific research, institutional mechanisms, and industrial layout. The findings can provide important support for building the theoretical system and promoting the engineering application of safe and resilient modern water networks.
As urban underground utility networks shift from incremental expansion to the renewal of existing assets, their governance faces a core question: Whether independently established discipline-specific conditions can become jointly feasible and remain effective once they enter shared underground space and a continuous engineering chain. The corresponding engineering-science logic is to successively transform dispersed discipline-specific conditions into jointly feasible conditions, executable project conditions, and conditions for subsequent engineering decisions. This study defines a cross-utility engineering linkage as a situation in which a discipline-specific condition materially alters the feasibility of another facility. Along the three stages of joint judgment, project execution, and subsequent updating, four condition states are identified: independently feasible discipline-specific conditions, jointly feasible conditions, executable project conditions, and conditions for subsequent engineering decisions. Three condition transitions connect these states across the engineering chain, while common-boundary identification, implementation-path differentiation, governance translation, and cross-period feedback explain the corresponding transition mechanisms. The results indicate that discipline-specific compliance cannot substitute for joint checking at common boundaries; implementation paths are determined, under non-negotiable safety baselines and relevant discipline-specific constraints, by the degree of spatial lock-in, data reliability, and construction windows; conclusions from joint checking become executable at the project level only after they are incorporated into project documents, assigned to responsible entities, and used to modify project conditions; and cross-period feedback is established when as-built and operational information is retrieved and used in subsequent projects to revise engineering decisions. For engineering applications, the framework can therefore be used to identify common boundaries, select implementation paths, incorporate joint-check conclusions into project documents and assign responsibilities, and use as-built and operational information to update subsequent decisions. Accordingly, disciplinary conflicts, discontinuities in project implementation, and failures in information reuse can be interpreted uniformly as interruptions in the transformation of engineering conditions. Systemic governance can thus be evaluated at a directly testable engineering scale: Whether engineering conditions remain effective across successive stages. This engineering-science logic can provide a basis for the transition of urban underground utility networks from facility-specific management toward lifecycle governance at the infrastructure-network level and, after sustained validation, may further enrich basic understanding of the development and long-term operation and maintenance of modern infrastructure networks.
Extending urban development underground has become a natural choice in response to land resource constraints and limitations on sustainable development. The construction of underground eco-cities is directly driven by national strategies such as those for deep-Earth and deep-space exploration, and aligns with major needs such as coping with extreme environments. Conventional underground development models are characterized by single functionality, rigid morphology, and insufficient ecological circulation capacity, making them inadequate to support emerging demands. There is a pressing need to research new paradigms and formulate their development pathways. Based on a systematic understanding of the development context of underground eco-cities and an analysis of key research breakthroughs, this study proposes a new paradigm for constructing underground eco-cities based on the concept of cross-scale bionics. It precisely analyzes the micro-geometric morphology and mechanical mechanisms of ant nests, revealing their excellent mechanisms in stress management, energy consumption optimization, and system synergy. Core design principles at the meso-scale, such as resilience, sustainable circulation, and distributed intelligence, are identified and then mapped onto the macro-scale functional topological structure and deep stress-adaptive morphology of underground cities. Guided by this methodology, a three-dimensional strategic blueprint for a depth-scalable, morphologically adaptive underground eco-city is constructed. Based on cutting-edge engineering practices both in China and abroad, it is preliminarily judged that the pathway from biological wisdom to engineering practices is viable. For future applications, key enabling technologies across the entire chain of "perception–modeling–regulation–decision-making" and developmental elements like synergistic "legal–social–economic" management are elaborated. This study will directly promote a fundamental shift in underground development from passive adaptation to active shaping. The new paradigm proposed possesses both universality and good transferability, providing theoretical references for serving deep space strategies and constructing subsurface habitats on extraterrestrial celestial bodies.
In the context of advancing urban resilience and modernizing the national governance system, dual-use public infrastructure has emerged as a pivotal spatial carrier for coordinating urban development and security. However, the implementation of the dual-use concept at the planning level faces systemic challenges, including the lack of top-level design, disconnections in transmission mechanisms, and barriers to cross-departmental coordination. This study reviews the conceptual evolution and policy trajectory of dual-use public infrastructure, and explores the jurisprudential logic and integration challenges concerning its alignment with the five-level, three-category territorial space planning system. The study reveals that the development of dual-use infrastructure is by no means a mere stacking of individual projects; rather, it represents a profound transformation of the logic underpinning territorial space governance. Through a comparative case study of three typical cities (i.e., Beijing, Chengdu, and Xi'an), the study identifies four core bottlenecks: jurisprudential conflicts in planning, ineffective indicator transmission, spatial misallocation, and incomplete standards and specifications. Drawing upon the system theory, planning integration theory, collaborative governance theory, and whole-life cycle management theory, this study constructs a closed-loop embedding framework consisting of strategic leadership, spatial constraints, facility coordination, and practical implementation. Accordingly, a systematic embedding pathway is proposed, encompassing four dimensions: strengthening top-level design guidance, optimizing the planning compilation system, reconstructing a collaborative governance framework, and refining laws and standards. This research aims to translate dual-use requirements from policy discourse into statutory planning control requirements, providing theoretical support and strategic references for achieving dual-use coordination and enhancing urban resilience.
Current communication, navigation, and remote sensing satellite systems remain largely isolated from one another, leading to a prominent contradiction with the high-quality application demands for all-domain real-time response, intelligent autonomous operation, and inclusive public services. Promoting innovation in positioning, navigation, timing, remote sensing, and communication (PNTRC) services and establishing a space-based spatiotemporal intelligent system that integrates communication, navigation and remote sensing represent a key direction for resolving this contradiction. Commercial aerospace serves as the core carrier and industrial foundation for building such a system. This study reviews the research background and technical progress of space-based spatiotemporal intelligence, analyzes development challenges regarding technical capabilities and system architecture, and discusses the supporting role of commercial aerospace in space-based spatiotemporal intelligence as well as the enabling role of space-based spatiotemporal intelligence for commercial aerospace. Furthermore, a five-dimensional enhancement architecture for space-based spatiotemporal intelligence oriented toward commercial aerospace is proposed, comprising a perception enhancement layer evolving from full downlink transmission to on-orbit intelligent screening; a cognition enhancement layer evolving from multi-source heterogeneous data to knowledge fusion; a service enhancement layer evolving from data sales to decision output; a system integration layer evolving from single-satellite single-function to integrated communication, navigation, and remote sensing; and a value co-creation layer evolving from chain-based supply to ecosystem collaboration. Drawing on engineering practices such as the Luojia-3 and Oriental Smart Eye projects, four progressive typical scenarios (i.e., global, regional, urban, and individual) are selected to analyze the applicability of the five-dimensional enhancement architecture, confirming that it can support business models such as Data-as-a-Service, Platform-as-a-Service, Decision-as-a-Service, and capability co-construction. Finally, recommendations for large-scale application of the five-dimensional enhancement architecture are proposed from the perspectives of system integration, service processes, service models, standards systems, business models, and talent cultivation, providing theoretical support and practical references for optimizing the industrial layout of commercial aerospace.
Geospatial multi-agent collaboration is a new intelligent collaborative paradigm centered around geospatial large models, where multiple agents collaborate, self-organize, self-learn, and self-adapt through continuous optimization and iteration. Based on an analysis of the current development status of geospatial multi-agent collaboration, this study explores its essence and characteristics. It also presents a five-layer system architecture that comprises physical perception, data governance, intelligent engine, multi-agent, and business application layers, as well as three collaboration modes: horizontal, vertical, and hybrid collaboration. Moreover, the study summarizes key technologies such as multimodal data perception and representation, deep integration and understanding of geographical knowledge, intelligent utilization of specialized tools and quality control, as well as collaborative scheduling and self-optimization of multiple agents. Demonstration applications for disaster emergency support service scenarios are also carried out. This study aims to lay a foundation for constructing intelligent systems with capabilities such as task self-planning, tool self-invocation, result self-evaluation, and memory self-evolution. This will advance geospatial applications from static analysis to dynamic intelligent response, providing next-generation decision support for fields such as natural resource monitoring, emergency management, and smart cities.
The rapid advancement of artificial intelligence, particularly large language models (LLMs), has become a central catalyst for the digital transformation across multiple sectors. In the water conservancy field, the efficient circulation of data elements is critical for advancing water resource management, optimizing infrastructure operations, and enhancing disaster resilience. However, current LLM-enabled data circulation faces multiple challenges, including security vulnerabilities, weak cross-domain coordination, latency in data processing, and limited interpretability. Leveraging LLM technologies, this study starts by identifying critical challenges in the water conservancy domain, which include inadequate data sharing, subpar monitoring capabilities, poor data quality, constrained analytical abilities, underutilized data values, and elevated operational and maintenance expenses. Guided by the principles of open access, elastic scalability, full-chain security assurance, and standards-based interoperability, this study proposes a six-layer, data-centric architecture, encompassing multi-source data acquisition, hybrid storage, intelligent data governance, agile service orchestration, scenario-driven applications, and endogenous security. Furthermore, it analyzes the coupling mechanisms and collaborative requirements across five systemic dimensions: regulatory frameworks, data and model standards, platform technologies, talent ecosystems, and pilot-scale experimentation. Building on these insights, a development pathway is proposed for constructing an LLM-enabled ecosystem for the market-oriented, secure, and intelligent circulation of water conservancy data, providing theoretical support and practical guidance for facilitating the market-oriented allocation and efficient circulation of water conservancy data. The proposed architecture can significantly enhance data usability, interpretability, and circulation efficiency, thereby propelling the water conservancy sector toward high-quality digital transformation, and providing a solid theoretical and technical foundation for the establishment of a smart water conservancy system.
China has a vast bridge infrastructure inventory that, on the whole, has entered a phase marked by intensive performance degradation and centralized maintenance. Traditional operation and maintenance paradigms, which heavily rely on manual labor, find it challenging to handle systemic safety assurance pressures under complex service environments. This study delves into the evolution trajectory of four types of intelligent technology paradigms in bridge engineering: logical models, discriminative models, generative models, and agent-based systems. It also reviews relevant research advances of the paradigms. Logic-based bridge engineering intelligence covers bridge sensing technologies and bridge health monitoring systems; discriminative bridge-engineering intelligence encompasses bridge-deck moving-load monitoring, component anomaly identification and defect detection, bridge monitoring data auditing, and structural numerical simulation with intelligent computing; and generative bridge-engineering intelligence involves applications of large language models, geometric model construction, and numerical-analysis mesh generation. Future directions are further discussed, including lightweight edge intelligence, domain-specific multimodal large models, full-lifecycle knowledge graphs, engineering swarm intelligence and embodied intelligence, and agent-based foundations for predictive maintenance. The study concludes that existing intelligent methods and technologies remain insufficient to support highly reliable engineering decision-making in complex service environments and cannot yet meet the growing demands for bridge safety, resilience enhancement, and service-life extension. Accordingly, research and development of agent-based systems for bridge engineering should be prioritized and their large-scale deployment promoted, thus to upgrade the capabilities of intelligent technologies in bridge engineering and advance the high-quality development of transportation infrastructure.
Extreme disasters in China are characterized by high frequency, concurrent hazards, and chain transmission, making it urgent to develop a modern emergency rescue system. Reshaping and upgrading the modernization level of the emergency rescue system from the perspective of digital and intelligent governance has become a core issue in the construction of the emergency management system. Focusing on digital and intelligent governance for emergency rescue, this study analyzes its theoretical implications, clarifies the demand characteristics from multi-dimensional perspectives, reviews the development status in terms of full-domain perception, survivable communication, unified command, intelligent decision-making, collaborative operation, and precise allocation, and points out that it will evolve in the directions of panoramic, intelligent, precise, edge-oriented, resilient, and integrated development. The study holds that in the “intelligence‒command‒execution” chain of emergency rescue under the “three disruptions” (i.e., disruptions of roads, electricity, and Internet access) scenario, the capabilities of full-domain perception and communication guarantee need to be strengthened in the “intelligence” system; the capabilities of unified collaboration, precision, and efficiency need to be improved in the “command” system; and the efficiency of task synergy and orderly support needs to be enhanced in the “execution” system. Furthermore, targeted construction ideas are proposed: developing a modern emergency rescue system with data-driven, intelligent empowerment, and collaborative efficiency as the core features, to enhance the system’s resilience and precise disposal capacity in response to major disasters; and constructing a parallel and integrated framework of “strategic guidance‒organizational management‒science and technology” for digital and intelligent governance of emergency rescue, thereby forming a new governance paradigm that serves emergency rescue tasks. Therefore, it is suggested to strengthen digital supervision to improve the full-domain perception and communication guarantee capabilities of the emergency management system; establish an emergency command center to enhance the system’s capacities in collaborative research and scientific decision-making; promote the digital and intelligent integration of equipment and material support to strengthen the efficiency of rescue operations; and improve the industrial support system to boost the high-quality development of the emergency digital and intelligent industry.
Artificial intelligence (AI) is increasingly being integrated into knowledge acquisition, skills training, and competency assessment in medical education, prompting changes in training models and governance mechanisms. To examine the progress and implementation challenges of AI-driven paradigm reshaping in medical education, this study combines literature analysis with engineering-oriented reasoning to review applications, challenges, and practices in China and abroad from the perspective of students' learning experiences and to propose response pathways for China. The analysis suggests that AI has expanded from assisting individual tasks to supporting teaching and assessment processes, with teachers becoming learning facilitators and designers of human‒AI collaborative teaching, and students becoming active learners. The implementation of AI in medical education faces three mismatches: technological systems versus learning capabilities, algorithmic objectives versus educational goals, and efficiency versus humanistic values. These mismatches are manifested as cognitive overload, goal substitution and automation bias, as well as erosion of student data sovereignty and insufficient humanistic education, respectively. Based on an analysis of coordination across technological, organizational, and governance levels, a three-dimensional pathway encompassing capability building, system optimization, and governance improvement is proposed. It includes integrating digital literacy and human‒AI collaboration training at successive stages, improving the controllability, explainability, and reviewability of AI teaching platforms, and strengthening data protection, ethical standards, and responsibility allocation. The integration of AI into medical education should be guided by medical talent training goals and supported by the coordination among technology, educational organization, and governance, thus to cultivate medical professionals with digital literacy and humanistic qualities.
Driven by technologies such as whole-slide imaging (WSI), spatial omics, and self-supervised pretraining, digital-intelligent pathology (DIP) is shifting pathology from traditional experience-based diagnosis toward a new data-intelligence-driven engineering paradigm. Building on these advances, DIP is establishing an integrated engineering system tailored to clinical and pharmaceutical needs, thereby forming a closed loop of diagnosis‒treatment‒Response. This study outlines the technological evolution of DIP, focusing on large-scale self-supervised foundation models, cross-modal virtual staining, pathology‒omics alignment, and tumor microenvironment (TME) modeling. It also outlines DIP's core functions: automated grading, multi-task learning, TME characterization, and genotype-phenotype mapping. Through examples in target discovery, mechanism studies, toxicity assessment, and precision trials, the study demonstrates how DIP enables functional phenotype mining, improves efficacy prediction and patient stratification, and delivers clinical and industrial value. Moreover, it explores challenges regarding data standardization, cross-center generalization, interpretability, and regulatory governance, and envisions DIP evolving, through multimodal and trustworthy artificial intelligence, into a sustainable "diagnosis‒treatment‒Response" loop. For the development of national biopharmaceutical and health systems in China, DIP is positioned as a "core hub for translational medical engineering," steadily advancing pathology from a diagnostic science to a therapeutic engine. This suggests that DIP could become an important link between clinical decision-making and drug R&D, bolstering original innovation and R&D efficiency of China's biopharmaceutical industry.
The advancement of intelligent manufacturing is a critical benchmark for the core competitiveness of the industrial sector. Embodied intelligence provides crucial support for the strategic upgrading of manufacturing systems, driving a paradigm shift in industrial entities from passive execution to active perception, adaptation, and interaction. Distinct from traditional reviews focused on underlying algorithmic principles, this study targets engineering implementation. To address the inherent challenge that current foundation models struggle to bridge the engineering gap of high precision, rapid cycle times, and strong robustness in industrial settings, we propose an analytical framework that derives technological requirements by reverse-engineering scenario-specific pain points. Intelligent manufacturing of China is currently advancing rapidly toward flexibility, autonomy, and integration. The complex working conditions and stringent process indicators in industrial scenarios have raised higher requirements for the implementation of intelligent technologies, and the disconnection between general intelligent models and practical industrial demands has become increasingly prominent. Therefore, this study examines the technological empowerment of industrial embodied intelligence across five representative industrial application domains: human‒robot collaboration, precision assembly, intelligent operation and maintenance, logistics and warehousing, and digital twin. A four-tier technical architecture is established, comprising multimodal perception and state characterization, cognitive reasoning and task decision-making, dexterous execution and motion control, and digital twin-enabled closed-loop optimization. Furthermore, the engineering adaptation and evolutionary pathways of embodied intelligence are delineated, thereby achieving bidirectional alignment between application scenarios and technological capabilities. To advance the development of industrial embodied intelligence, this study recommends the following strategic directions: strengthening frontier research to overcome critical technical bottlenecks, cultivating a robust industrial ecosystem to accelerate large-scale deployment, and enhancing policy support to optimize the developmental landscape. These efforts are expected to facilitate the transition of embodied intelligence from laboratory validation to controllable and scalable industrial implementation, thereby exploring novel paradigms for intelligent manufacturing.
Domain-specific agents emerge from the deep integration of large language models (LLMs) with domain knowledge, data systems, and business tools. They are evolving from auxiliary applications for knowledge question answering into execution systems for complex tasks. As sources of domain data diversify, business rules grow more complex, and external tools proliferate, LLMs alone struggle to maintain semantic consistency across task understanding, resource invocation, and result assessment. Ontology enhancement formally represents domain objects and their relations and incorporates business conditions, operational capabilities, and runtime states into a unified semantic framework. It thus provides domain-specific agents with a stable knowledge foundation and a verifiable basis for task execution. This study first defines the concept and scope of ontology-enhanced domain-specific agents, develops an ontology semantic model and a layered technical framework, and reviews key technologies for ontology construction, knowledge acquisition, tool invocation, task planning, and result verification. It then conducts an application validation from a large manufacturing enterprise. The findings indicate that ontology enhancement can reduce semantic deviations in multi-constraint queries by explicitly representing business objects, relation paths, metric definitions, temporal conditions, and state rules, thereby providing verifiable semantic support for domain-specific agents accessing heterogeneous business data.
To address the needs of robotic manipulators in complex environments for autonomous decision-making, long-horizon execution, and multi-constraint interactions, the integration of intelligent agents with robotic manipulator systems is increasingly recognized as a key technical pathway for embodied intelligence. To support research review and engineering implementation in this direction, this study seeks to establish a structured technical framework for agent‒manipulator integration, clarifying the methodological landscape, data-and-evaluation context, and key challenges. First, the study reviews the technical foundations of manipulator intelligence and the evolution of control paradigms. Then, it summarizes representative methods in terms of system architectures, decision-making mechanisms, and typical applications from three levels: reinforcement-learning-based agents, planning and cognition-oriented agents, and large-model-driven embodied agents. Furthermore, we examine representative public datasets and provide key considerations for dataset selection. Our analysis shows that existing agent‒manipulator integration research has formed three major technical routes centered on learning-based policies, planning-and-reasoning mechanisms, and large-model-driven closed-loop frameworks, and these routes differ in task generalization capability, adaptability to interaction complexity, and deployment cost. Meanwhile, insufficient coverage of data distributions and non-unified evaluation protocols hinder the reproducibility and comparability of reported findings. Toward the development of general-purpose manipulator agents, future research should prioritize expanding data and task coverage, improving cross-domain transfer and robustness, developing safety and reliability mechanisms, and establishing engineering-oriented deployment pipelines, thereby supporting the transition of relevant methods from prototype validation to scalable application.
Large language models (LLMs) learn the grammar, semantics, and inherent patterns of human language through training on massive text corpora, thereby acquiring the core capabilities to comprehend and generate natural language text. Due to the incorporation of false, erroneous, and even maliciously manipulated data samples into their training data, LLMs suffer from data contamination, which is specifically manifested in two forms: model controllability contamination and content credibility contamination. The root causes of this issue can be attributed to four interrelated dimensions: data source risks arising from the open data ecosystem and insufficient data cleaning capabilities; inherent model defects stemming from inadequate statistical generation and secure generalization capabilities; the diffusion and secondary propagation of contamination under trust endorsement in application and interaction scenarios; and the dual dilemma of lagging defense measures and disjointed regulatory frameworks. To address the aforementioned challenges, it is imperative to implement targeted countermeasures across four key levels: at the data level, fortifying the source line of defense and establish a clean, controllable training data system; at the model level, optimizing generation logic and alignment capabilities to enhance model robustness; at the application and interaction level, dispelling users' blind trust to block the propagation of contamination and scenario-specific risks; at the defense and regulation level, adopting a dual-wheel drive model of law-based empowerment and technological collaboration to build a prevention and control system covering the full life cycle of data.
With the proliferation of generative artificial intelligence (AI) technologies, particularly large language models (LLMs), their associated safety risks have grown increasingly pressing. As a proactive safety methodology for discovering and evaluating potential threats, red teaming has attracted significant attention in academia, industry, and among policymakers. Distinct from existing reviews that mainly focus on a single model type and emphasize the categorization of attack methods, this study expands the scope to encompass multiple system types, including LLMs, multimodal models, agents, and retrieval-augmented generation (RAG). It establishes a novel classification framework centered on the dual dimensions of “attack generation” and “effect evaluation,” while systematically reviewing representative industrial practices. First, we clarify the core concepts and risk taxonomy of generative AI red teaming. Subsequently, from the perspectives of single-turn and multi-turn interactions, we survey automated red teaming attack techniques targeting LLMs, and categorize the corresponding attack effectiveness detection methods into three major classes: discriminative, generative, and edge-side detection. Furthermore, we explore the novel safety threats and innovative testing paradigms introduced by multimodal models, AI agents, and RAG systems. Finally, we examine the persistent challenges in this field, including the ongoing evolution of attack and defense methods and the imperative for multi-stakeholder governance, and outline directions for future research. This study aims to provide technical references and practical insights for China in constructing a safety assessment system for generative AI and enhancing proactive risk defense capabilities.
New-type electrochemical energy storage are key enabling technologies for building a new power system and achieving the carbon peaking and carbon neutralization goals. The performance of key materials directly determines the energy density, safety, lifespan, and cost-effectiveness of energy storage devices, and they hold irreplaceable strategic values for ensuring national energy security. This study focuses on three major new-type electrochemical energy storage systems (i.e., lithium-ion batteries, sodium-ion batteries, and flow batteries), and reviews the research progress on electrode, electrolyte, and separator materials, while proposing phased optimization roadmaps for the development of these different materials. Meanwhile, the study further focuses on common key issues in the field of material research and development (R&D). Given the complex and diverse nature of energy storage material systems, it analyzes bottlenecks in research framework planning, resource and environmental constraints, R&D models, and infrastructure platforms for the commercialization of research outcomes. The study proposes shifting from a decentralized approach to a systematic one by strengthening cross-system fundamental research and collaboration across the entire value chain; shifting from a performance-oriented approach to a full-lifecycle, green, and sustainable approach, incorporating resource availability, supply chain security, and environmental impact as core evaluation dimensions in the early stages of material R&D; transitioning from trial-and-error to data- and intelligence-driven approaches to build an independent "artificial intelligence + energy storage materials" R&D model; and moving from isolated laboratory innovation to end-to-end integration by establishing a technology commercialization mechanism that synergistically advances pilot-scale validation and standards testing, thereby promoting the high-quality development of key energy storage materials.
High‑purity synthetic quartz is a pivotal foundational material underpinning semiconductors, optical communications, aerospace and other sectors. It circumvents resource limitations of high‑quality natural quartz ores and provides a critical pathway for China to realize industrial self‑reliance in high‑purity quartz materials. This study summarizes the current status and trends in the global technology and industry development of high‑purity synthetic quartz, and analyzes the development status and challenges faced by China. China has achieved mass production of mid‑to‑high‑end synthetic quartz products; however, top‑tier scenarios including extreme ultraviolet lithography and ultra‑high‑precision optics remain monopolized by overseas giants. Moreover, multiple bottlenecks still exist. First, there is a lack of in-depth understanding of extreme‑regime process mechanisms, coupled with insufficient capabilities to modulate multi‑physics‑field coupling, micro‑defects, and internal stress, leading to unsatisfactory uniformity and yield of high‑end products. Second, core deposition equipment and precision control systems are heavily reliant on imports, restricting domestic iterative upgrading. Third, there is a shortage of validation platforms for extreme operating conditions together with imperfect evaluation standards, resulting in prolonged qualification cycles and high entry thresholds for high‑end products. Furthermore, this study proposes the following recommendations: strengthening national strategic guidance and sustained long-term investment, to assess industry-specific technological barriers and define roadmaps for key technological breakthroughs; fostering an integrated innovation ecosystem driven by major application demands while focusing on the coordinated advancement of core equipment and engineering process within extreme conditional windows; cultivating interdisciplinary talents spanning materials, equipment, and engineering, and deepening high-level international collaboration; establishing national testing platforms for extreme operating conditions together with indigenous evaluation standards systems, while improving upstream-downstream collaborative innovation and risk-sharing mechanisms, thus to integrate the entire chain spanning fundamental research, engineering scale-up, and application validation. These measures are expected to drive the self-reliant and high-quality development of the high-end synthetic quartz industry.
Polyimide is a strategically important functional material for extreme-environment applications, playing an irreplaceable role in key fields such as aviation, aerospace, and semiconductors. In recent years, the field of polyimide materials in China has made significant progress, highlighting its supportive and leading significance in the advanced industrial field. However, high-end polyimide products still rely mainly on imports, posing risks to the security of the industrial chain. Therefore, it is urgently necessary to address capability gaps, consolidate competitive strengths, and achieve greater industrial self-reliance and self-sufficiency industrial self-reliance in the polyimide material industry. Considering the urgency of strategic demand, representativeness of industrial scale, and necessity of independent supply assurance, this study focuses on four typical polyimide materials in applications: matrix resins for composites, molding and thermoplastic resins, films, as well as photosensitive materials. Moreover, the study reviews their current state of development both internationally and domestically, and identifies the major challenges currently faced. In response to the prominent structural contradictions and the challenge of being large yet not strong, it distills the common issues that the industry imperatively needs to address from three aspects: core materials, high-end products, and key equipment. Furthermore, the study analyzes the development trends and proposes strategic recommendations for achieving industrial self-reliance and self-strengthening, aiming to promote the high-quality development of the polyimide materials industry in China.
Advanced ceramic materials are the foundation for the safe operation of hot-end components and dynamic energy impact protection components of rotorcraft, as well as an important material basis for the electrification and intelligent development of rotorcraft. Grasping the technical direction of advanced ceramic materials oriented to rotorcraft applications and envisioning future development will provide key support for the development of China's advanced ceramic material industry and the improvement in its rotorcraft equipment capacity. The study explores the application of advanced materials driven by the demand of rotorcraft for high survivability, lightweight structures, all-domain/all-weather service, and intelligence, expounds on the technical characteristics of advanced ceramic materials and the adaptability of these materials for rotorcraft, and reviews the research progress of advanced ceramic materials for rotorcraft in China. Moreover, it summarizes the development trends in this field. Specifically, lightweight yet high-performance structural ceramic materials underpin the safe operation of specific structural parts; the composite toughening of advanced ceramic materials has become the key support for reliability improvement; multifunctional piezoelectric ceramics are the material basis for intelligent development; and the flexibility of information functional ceramics and its fusion with artificial intelligence are important driving forces for digital transformation of physical systems. The study proposes that future research and application of advanced ceramic materials for rotorcraft should prioritize bulletproof ceramic materials for high survivability, ceramic composite coatings for all-domain service, piezoelectric ceramic materials for all-weather service, flexible piezoelectric devices for intelligent integration, and advanced ceramic digital qualification technology for rapid research and development. These efforts aim to accelerate the technological progress of advanced ceramic materials and improve the comprehensive application efficiency of rotorcraft.
The deep processing technology for high-purity raw materials of critical metals serves as a vital technical link that connects mineral resources and high-end manufacturing industries, bearing great significance for safeguarding the security of national strategic resources. This study reviews the progress of high-purity processing technologies for typical critical metals including gallium, indium, germanium, titanium, tungsten, molybdenum, vanadium, tantalum, niobium and scandium, and conducts a comparative analysis of their industrial application status, technical maturity, international benchmark levels, and existing technical shortcomings. The results indicate that China has established a complete industrial system and large-scale production capacities for conventional critical metal products with a purity of 4N‒5N grade. However, there remain prominent gaps between China and international advanced levels in terms of preparation stability of ultra-high-purity products above 6N grade, precision removal of trace impurities, and independent research and development of high-end core equipment. Based on a technical maturity evaluation, it is projected that future deep processing technologies for high-purity raw materials of critical metals will exhibit four major development trends: an extreme elevation in product purity, diversified and complex development of material components, nanocrystallization and compounding of powder materials, and an evolution of separation mechanisms toward multi-field coupling. Oriented toward 2040, a medium- and long-term development roadmap for the industry is proposed, focusing on five key development directions: ultra-high-purity preparation, complex system separation, nano and composite powder preparation, multi-field-coupled targeted separation, and intelligent green metallurgy. Corresponding countermeasures and suggestions are proposed from five dimensions: consolidating industrial foundation, optimizing resource allocation, strengthening basic research, overcoming equipment technical bottlenecks, and advancing artificial intelligence empowerment. This study can provide scientific support for the independent development and strategic planning of China's critical metal industrial chain.