• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2026, Vol. 62 ›› Issue (17): 156-167.doi: 10.3901/JME.260823

• 特邀专栏:人形机器人与具身智能 • 上一篇    

扫码分享

基于约束辨识的油箱盖操作分层轨迹规划研究

张皓晨1, 魏洪兴1, 王殿君2, 李泽宇1   

  1. 1. 北京航空航天大学机械工程及自动化学院 北京 100191;
    2. 北京石油化工学院机械工程学院 北京 102617
  • 收稿日期:2026-02-17 修回日期:2026-07-23 发布日期:2026-11-20
  • 作者简介:张皓晨,男,1995年出生,博士研究生。主要研究方向机器人灵巧操作。E-mail:zhang_haochen@buaa.edu.cn;魏洪兴(通信作者),男,1974年出生,博士,教授,博士研究生导师。主要研究方向为机器人控制。E-mail:weihongxing@buaa.edu.cn;王殿君,男,1973年出生,博士,教授。主要研究方向为机器人技术及应用。E-mail:wdjkevin@126.com;李泽宇,男,1995年出生,博士,讲师。主要研究方向为机器人控制。E-mail:lizeyubuaa@buaa.edu.cn

Constraint-identification-based Hierarchical Planning for Fuel Tank Cap Operation

ZHANG Haochen1, WEI Hongxing1, WANG Dianjun2, LI Zeyu1   

  1. 1. School of Mechanical Engineering and Automation, Beihang University, Beijing 100191;
    2. Mechanical Engineering Academy, Beijing Institute of Petro-Chemical Technology, Beijing 102617
  • Received:2026-02-17 Revised:2026-07-23 Published:2026-11-20

摘要: 传统加油机器人在环境信息不全的操作场景中,需要具备自主或在人类协助下感知被操作对象的能力。为解决预定离线轨迹导致机器人自主能力受限的问题,提出一种约束辨识分层规划方法(Constraint-identification-based hierarchical planning, CIHP)。首先,根据真实场景的实际约束形式建立机器人与目标物的参数化受约束运动学模型;其次,基于人机交互的具身引导方法获取任务先验数据,辨识模型参数,然后利用主任务空间与冗余的约束零空间之间相互独立的性质,分层求解最优化油箱盖开启轨迹,同时满足工具姿态保持等次级优化目标。实验表明,相较于利用主成分分析(Principal component analysis, PCA)拟合轨迹规划方法,提出的方法任务时间更短、能耗更低、接触力收敛更快,验证了提出的策略求解方法具有良好的规划能力。CIHP方法增强了加油机器人对任务约束的具身感知,为实现复杂环境信息不全操作任务下的自主规划提供了理论与工程基础。

关键词: 加油机器人, 参数化建模, 约束辨识, 分层规划, 轨迹规划

Abstract: Conventional refueling robots operating in environments with incomplete information need the capability to perceive and understand the manipulated objects either autonomously or with human assistance. To overcome the limitation of robot autonomy imposed by pre-defined offline trajectories, a Constraint-Identification-based Hierarchical Planning (CIHP) method is proposed. First, according to the actual constraint observed in real scenarios, a parameterized constrained kinematic model of the robot-object system is established. Then, task prior data are acquired through a human-robot interaction embodied guidance approach, and the model parameters are subsequently identified. The optimal opening trajectory of the fuel tank cap is hierarchically solved by utilizing the mutual independence between the primary task space and the redundant constrained null space, while satisfying secondary optimization objectives such as tool posture keeping. Experimental results demonstrate that, compared with the principal component analysis (PCA)-based trajectory fitting method, the proposed CIHP method achieves shorter task duration, lower energy consumption, and faster convergence of contact force, verifying its superior planning performance. The CIHP method enhances the robot’s embodied perception of task constraints, providing both theoretical and engineering bases for autonomous planning in complex non-rigid contact manipulation tasks.

Key words: refueling robot, parametric modeling, constraint identification, hierarchical planning, trajectory planning

中图分类号: