摘要：本文分析了现有比特导向通信范式在人机物融合中的局限，指出任务导向通信范式是满足人机物融合通信需求的可行方案。首先，梳理了任务导向通信中的任务特征联合编码、弹性组网和动态资源分配关键技术。其次，归纳总结了任务导向通信技术在边缘推理、多智能体协同和遥操作等人机物融合典型场景中的应用。最后，探讨了任务导向通信在任务特征表征、计算开销、通用框架、可靠性和可解释性等方面的技术挑战。相关分析可为支撑人机物融合的任务导向通信方案设计提供参考。

关键词：人机物融合;任务导向通信；任务特征联合编码；弹性组网；动态资源分配

Abstract: This paper analyzes the limitations of existing bit-oriented communication paradigms in human-cyber-physical applications. The task-oriented communication paradigm is identified as a feasible approach to meeting human-cyber-physical communication requirements. Key technologies for task-oriented communication are reviewed, including task-feature joint source-channel coding, elastic networking and dynamic resource allocation. Applications of these technologies in typical human-cyber-physical scenarios, including edge inference, multi-agent collaboration and teleoperation, are summarized. Finally, technical challenges related to task-feature representation, computational overhead, general-purpose frameworks, reliability and interpretability are discussed. The analysis provides a reference for designing task-oriented communication schemes to support human-cyber-physical applications.

Keywords: human-cyber-physical; task-oriented communication; task-feature joint source-channel coding; elastic networking; dynamic resource allocation