• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2026, Vol. 62 ›› Issue (17): 65-80.doi: 10.3901/JME.260660

• 特邀专栏:人形机器人与具身智能 • 上一篇    

扫码分享

人形机器人具身智能触觉感知技术:进展、挑战与展望

杨赓1,2,3,4, 梁冀凯1,2,4, 陶永5, 吕鸿昊1,2,3,4   

  1. 1. 浙江大学机械工程学院 杭州 310058;
    2. 浙江大学流体动力基础件与机电系统全国重点实验室 杭州 310027;
    3. 浙江大学高端装备研究院 杭州 311100;
    4. 电力装备制造及智能运维机器人浙江省工程研究中心 杭州 310058;
    5. 北京航空航天大学机械工程及自动化学院 北京 100191
  • 收稿日期:2025-12-31 修回日期:2026-05-20 发布日期:2026-11-20
  • 作者简介:杨赓,男,1980年出生,研究员。主要研究方向为智能感知与人机交互。E-mail:yanggeng@zju.edu.cn;吕鸿昊(通信作者),男,1996年出生,平台百人计划研究员。主要研究方向为机器人安全交互及控制。E-mail:lvhonghao@zju.edu.cn
  • 基金资助:
    浙江省杰出青年科学基金(延续)(No.LRG25E050001)、国家自然科学基金(52375031、52405038)、浙江省科技厅“尖兵领雁+X”研发攻关计划(2024C04057(CSJ)、2025C01012、2026C02A1208)和国家科技重大专项(2026ZD1609002)资助项目。

Embodied Tactile Perception Technologies for Humanoid Robots: Progress, Challenges, and Perspectives

YANG Geng1,2,3,4, LIANG Jikai1,2,4, TAO Yong5, LÜ Honghao1,2,3,4   

  1. 1. School of Mechanical Engineering, Zhejiang University, Hangzhou 310058;
    2. State Key Laboratory of Fluid Power and Mechatronic Systems, Zhejiang University, Hangzhou 310027;
    3. Institute of Advanced Machines, Zhejiang University, Hangzhou 311100;
    4. Zhejiang Engineering Research Center of Robotics in Electric Equipment Manufacturing and Intelligent Operation-Maintenance, Hangzhou 310058;
    5. School of Mechanical Engineering and Automation, Beihang University, Beijing 100191
  • Received:2025-12-31 Revised:2026-05-20 Published:2026-11-20

摘要: 人形机器人作为具身智能的物理载体,要实现自主执行、安全交互与协同作业,需要具备稳定可靠的触觉感知能力。本文围绕精细操作、人机交互、安全防护与探索学习等典型应用场景,分析场景驱动下的触觉感知在覆盖范围、响应特性与鲁棒性等方面的现实需求。在此基础上,对比不同触觉传感器的工作原理与性能差异,总结触觉感知在典型应用场景下的实现方式及关键技术挑战。围绕人形机器人系统框架,分析人工智能增强的触觉感知技术在具身智能感知-决策-执行闭环中的关键作用。在系统梳理相关研究进展的基础上,结合代表性工程实践,对大面积可扩展触觉皮肤在全身触觉感知与人机交互中的实际应用进行了分析。最后,从规模扩展、参与决策、具身学习等方面进行展望,为人形机器人触觉系统构建与工程应用提供方向性参考。

关键词: 触觉感知, 具身智能, 人形机器人, 人工智能

Abstract: Humanoid robots, as physical embodiments of embodied intelligence, require stable and reliable tactile perception to achieve autonomous execution, safe interaction, and collaborative operation. Focusing on representative applications scenarios such as dexterous manipulation, human–robot interaction, safety protection, and exploratory learning, scenario-driven tactile perception requirements are analysed in terms of sensing coverage, dynamic response characteristics, and robustness. On this basis, the working mechanisms and performance distinctions of various tactile sensors are examined, and typical implementation approaches of tactile perception across application contexts, together with the associated technical challenges, are summarized. Within the framework of humanoid robot systems, the roles of artificial intelligence–enhanced tactile technologies in the perception–decision–action loop of embodied intelligent systems are further analysed. Following a systematic review of related work, representative engineering examples are discussed to illustrate the application of large-area, scalable tactile skin in full-body tactile perception and human–robot interaction. Finally, future directions are discussed from the perspectives of system scalability, the involvement of tactile information in decision-making, and embodied learning, providing directional guidance for the development and engineering application of tactile systems for humanoid robots.

Key words: tactile perception, embodied intelligence, humanoid robots, artificial intelligence

中图分类号: