节点文献

具身智能的发展路径:从本体智能、群体智能到网算协同智能

The Development Path of Embodied Intelligence: From Ontological Intelligence, Swarm Intelligence to Networked Computing Collaborative Intelligence

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 王宗宇夏长清李栋曾鹏

【机构】 中国科学院沈阳自动化研究所

【摘要】 本文系统阐述了具身智能的概念、演进历程、关键技术及应用前景。具身智能作为人工智能的前沿方向,其核心在于智能体通过物理实体(如机器人、智能驾驶汽车)与环境交互,实现自主决策与行为执行。具身智能的演进形成了两条互补路径:一是聚焦个体能力的“本体智能”,它依赖多模态感知、运动控制算法及大模型任务规划,被应用于工业制造、医疗手术与家庭服务;二是强调协作的“群体智能”,它通过通信协议、协调算法和分布式决策机制,实现多智能体协同作业,典型场景包括仓储管理、灾害救援与智能交通。为进一步支撑群体智能,本文认为“网算协同智能”有助于群体智能进行动态资源调度、任务卸载与分布式控制技术,可以解决资源冲突、提升系统弹性。针对硬件成本高昂、数据稀缺性与算法可靠性不足等未来挑战,需通过跨学科融合、开源数据集建设与真实场景迭代来突破。具身智能有望在工业、医疗等复杂场景中规模化应用,成为推动社会进步的关键生产力。

【Abstract】 This article systematically elaborates on the concept, evolution, key technologies, and application prospects of embodied intelligence. As a frontier direction in artificial intelligence, its core lies in enabling intelligent agents, through physical entities(such as robots and smart driving vehicles), to interact with the environment, achieving autonomous decision-making and action execution. The evolution of embodied intelligence has formed two complementary paths: the first focuses on individual capabilities, known as "Embodied Intelligence" or "Ontological Intelligence," which relies on multi-modal perception, motion control algorithms, and task planning by large models, finding applications in industrial manufacturing, medical surgery, and home services; the second emphasizes collaborative "Swarm Intelligence," which utilizes communication protocols, coordination algorithms, and distributed decision-making mechanisms to enable multi-agent collaborative operations, with typical scenarios including warehouse management, disaster rescue, and intelligent transportation. To further support collective intelligence, the paper suggests that "network-computation collaborative intelligence" can contribute through dynamic resource scheduling, task offloading, and distributed control technologies, thereby resolving resource conflicts and enhancing system resilience. Future challenges include high hardware costs, data scarcity, and insufficient algorithm reliability, requiring breakthroughs through interdisciplinary integration, the construction of open-source datasets, and iterative testing in real-world scenarios. Embodied intelligence is expected to achieve large-scale application in complex scenarios like industry and healthcare, potentially becoming a key productive force driving social progress.

  • 【文献出处】 自动化博览 ,Automation Panorama , 编辑部邮箱 ,2026年01期
  • 【分类号】TP18
  • 【下载频次】69
节点文献中: 

本文链接的文献网络图示:

本文的引文网络