| 刘海军*,王广慧*,吴善强**,周坤**,郑小飞**,冯伟博**.仿尺蠖爬壁机器人的曲面步态规划与CPG网络构建[J].高技术通讯(中文),2026,36(6):632~642 |
| 仿尺蠖爬壁机器人的曲面步态规划与CPG网络构建 |
| Curved surface gait planning and CPG network construction for an inchworm-inspired wall-climbing robot |
| |
| DOI：10. 3772 / j. issn. 1002 - 0470. 2026. 06. 009 |
| 中文关键词: 爬壁机器人； 中枢模式发生器网络； 步态分析； 运动规划； 吸附条件 |
| 英文关键词: wall-climbing robots, central pattern generator network, gait analysis, motion planning, adsorption conditions |
| 基金项目: |
| 作者 | 单位 | | 刘海军* | （*神华准格尔能源有限责任公司鄂尔多斯 010300）
（**中国计量大学机电工程学院杭州 310018） | | 王广慧* | | | 吴善强** | | | 周坤** | | | 郑小飞** | | | 冯伟博** | |
|
| 摘要点击次数: 82 |
| 全文下载次数: 80 |
| 中文摘要: |
| 针对爬壁机器人在复杂曲面的运动规划难题，提出了一种基于自适应中央模式发生器（central pattern generator, CPG）的仿生步态规划方法，并以自然界尺蠖运动为启发，设计了一种5自由度双足爬壁机器人。首先，分析了机器人双足吸盘稳定吸附的条件，通过基于有监督学习方法和多个可调节Hopf振荡器的耦合，构建了机器人3个关节的CPG网络模型；通过学习机器人在平面翻转步态的角度与振幅初值以获得曲面翻转的CPG网络参数，再对曲面翻转步态的CPG网络映射的关节角度和幅值进行线上修改，以实现曲面翻转步态的规划；然后，利用Matlab/Simulink与Adams建立仿真模型，验证机器人能够适应不同曲率曲面上的翻转步态；最后，将机器人在真实曲面环境中进行翻转步态实验，实验与仿真结果一致，表明在机器人学习平面步态后，CPG网络可将平面步态转换为曲面步态，验证了CPG在线调节步态规划方法的有效性。 |
| 英文摘要: |
| To address the motion planning challenges of wall-climbing robots on complex curved surfaces, this study proposes a bio-inspired gait planning method based on an adaptive central pattern generator (CPG). Inspired by the locomotion of inchworms in nature, a five-degree-of-freedom bipedal wall-climbing robot is designed. First, the conditions for stable adhesion of the robot’s bipedal suction cups are analyzed. A CPG network model for the robot’s three joints is constructed by coupling multiple tunable Hopf oscillators using a supervised learning method. The CPG network parameters for curved-surface flipping are obtained by learning the joint angles, amplitudes, and other parameters of the robot’s T-joint during planar flipping gaits. Online modifications are then applied to the joint angles and amplitudes mapped by the CPG network for curved-surface flipping to achieve gait planning. Subsequently, a simulation model is established using Matlab/Simulink and Adams to verify the robot’s ability of adapting to flipping gaits on surfaces with varying curvatures. Finally, flipping gait experiments are conducted on real curved surfaces, with experimental results consistent with simulations. The findings demonstrate that after learning planar gaits, the CPG network can transform them into curved-surface gaits, validating the effectiveness of the online adaptive gait planning method based on CPG. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
