新型磁吸附轮式机器人定位、规划与控制算法

doi:10.3901/JME.2025.21.060

摘要/Abstract

摘要： 核电站的安全壳作为最后一道安全屏障，具有至关重要的保护作用。传统的人工焊缝巡检效率低且具有一定的危险性，因此，机器人替代人工进行焊缝巡检具有重要的现实意义。以磁吸附轮式机器人为研究对象，针对其在复杂壁面环境中的定位、路径规划与控制展开了系统研究。首先，介绍磁吸附轮式机器人悬架结构，并提出了一种基于机器人内部传感器的定位算法，并结合焊缝分布地图与驱动力矩指标，设计了一种改进的遗传算法用于焊缝遍历路径规划。该方法优于传统的Fleury算法，能够提高路径的平稳性和经济性。针对恒定磁吸附力的局限性，提出了基于改进的深度确定性策略梯度(DDPG)算法的动态磁吸附力控制策略。使用MuJoCo、Gym和PyTorch搭建了机器人仿真环境，对不同磁吸附力条件下的机器人运动进行了模拟。通过改进的DDPG算法设计了动态磁吸附力控制策略，确保在足够磁吸附力的前提下，能有效降低能耗并提高稳定性。提出了基于YOLOv8的焊缝识别算法，并在机器人样机中进行了整机实验验证。最后，选择YOLOv8作为焊缝识别算法，并将训练后的模型部署至机器人中。研究成果为核电站、大型储罐、船舶舷侧等铁磁性结构的安全运维提供了一种高效、低风险的智能化解决方案，对推动我国核能、石化、船舶等行业的机器人化检测与维护具有重要示范意义和产业化推广价值。

关键词: 磁吸附机器人, 遍历路径规划, 深度确定性策略梯度算法, 磁吸附力控制, 焊缝识别

Abstract: The containment vessel of a nuclear power plant is regarded as the last line of defense and plays a crucial role in safety. Traditional manual inspection of weld seams is considered inefficient and poses certain risks. Therefore, the replacement of manual weld seam inspection with robots is endowed with significant practical significance. A systematic study is conducted on the positioning, path planning, and control of magnetically adsorbed wheeled robots in complex wall environments. Firstly, based on the structural configuration of the magnetically adsorbed wheeled robot, a positioning algorithm utilizing the robot's internal sensors is proposed. An enhanced genetic algorithm is designed for weld seam traversal path planning by integrating weld distribution mapping and driving torque metrics, demonstrating superior performance in path smoothness and energy efficiency compared to the conventional Fleury algorithm. To address limitations of fixed magnetic adsorption forces, a dynamic adsorption force control strategy is developed through an improved deep deterministic policy gradient (DDPG) algorithm. A simulation environment is constructed using MuJoCo, Gym, and PyTorch frameworks, where motor feedback control algorithms are validated and robot movements under varying adsorption forces are simulated. The optimized DDPG algorithm is employed to formulate a dynamic magnetic force regulation strategy that ensures adequate adhesion while minimizing energy consumption and enhancing stability. A weld seam recognition algorithm based on the YOLOv8 architecture is proposed and experimentally validated through prototype testing. Finally, YOLOv8’s superior performance is confirmed, leading to its deployment on the robotic system after model training. Finally, YOLOv8 was selected as the weld-seam detection algorithm and the trained model was deployed on the robot. The research results provide an efficient, low-risk, and intelligent solution for the safe operation and maintenance of ferromagnetic structures such as nuclear power plant containments, large storage tanks, and ship hulls. The work serves as an important demonstration for advancing robotic inspection and maintenance in China’s nuclear, petrochemical, and shipbuilding industries, and holds significant potential for wide industrial adoption.

Key words: magnetic adhesion robot, ergodic path planning, deep deterministic policy gradient algorithm, magnetic adhesion force control, weld identification

中图分类号:

TH112

李通甲, 臧鹏翔, 郭为忠. 新型磁吸附轮式机器人定位、规划与控制算法[J]. 机械工程学报, 2025, 61(21): 60-74.

LI Tongjia, ZANG Pengxiang, GUO Weizhong. Location, Planning and Control Algorithm of a Novel Magnetic Adhesion Wheel Robot[J]. Journal of Mechanical Engineering, 2025, 61(21): 60-74.

参考文献

[1] 周洪宇. 第三次工业革命的新能源革命[J]. 决策与信息，2016(3)：10-18. ZHOU Hongyu. The new energy revolution of the third industrial revolution[J]. Decision and Information，2016(3)：10-18.
[2] 王轶君，余功铭. 全球核电现状及发展前景展望[J]. 世界石油工业，2022，29(4)：78-80. WANG Yijun，YU Gongming. Current situation and development prospect of global nuclear power[J]. World Petroleum Industry，2022，29(4)：78-80.
[3] ROMAO J C，TAVAKOLI M，VIEGAS C，et al. Inchworm climber：A light-weight biped climbing robot with a switchable magnet adhesion unit[C]//Proceedings of the 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems. Piscataway，NJ：IEEE，2015：3320-3325.
[4] TAVAKOLI M，MARQUES L，ALMEIDA A T. OmniClimber：An omnidirectional light weight climbing robot with flexibility to adapt to non-flat surfaces[C]//Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems. Piscataway，NJ：IEEE，2012：4583-4588.
[5] PARK C，BAE J，RYU S，et al. R-Track：Separable modular climbing robot design for wall-to-wall transition[J]. IEEE Robotics and Automation Letters，2021，6(2)：1036-1042.
[6] KELLEY L，OSTOVARI S，BURMEISTER A B，et al. Design and experimental validation of a simple controller for a multi-segment magnetic crawler robot[C]//Unmanned Systems Technology XVII. Bellingham，WA：SPIE，2015：94680N.
[7] HONG S，UM Y，PARK J，et al. Agile and versatile climbing on ferromagnetic surfaces with a quadrupedal robot[J]. Science Robotics，2022，7(73)：eadd1017.
[8] SONG W，JIANG H，WANG T，et al. Design of permanent magnetic wheel-type adhesion-locomotion system for water-jetting wall-climbing robot[J]. Advances in Mechanical Engineering，2018，10(7)：1-11.
[9] WU M，GAO X，YAN W X，et al. New mechanism to pass obstacles for magnetic climbing robots with high payload，using only one motor for force‐changing and wheel‐lifting[J]. Industrial Robot：An International Journal，2011，38(4)：372-380.
[10] WU M，PAN G，ZHANG T，et al. Design and optimal research of a non-contact adjustable magnetic adhesion mechanism for a wall-climbing welding robot[J]. International Journal of Advanced Robotic Systems，2013，10(1)：63.S.
[11] KITAI S，TSURU K，HIROSE S. The proposal of swarm type wall climbing robot system “Anchor Climber”[C]//Proceedings of the 2005 IEEE/RSJ International Conference on Intelligent Robots and Systems. Edmonton，AB，Canada：IEEE，2005：475-480.
[12] WEN Jing，DUN Xiangming，MIAO Songhua，et al. Structural design and weld climbing characteristics of variable magnetic adsorption wall-climbing robot[J]. Robot，2011，33(4)：7.
[13] LI Z，XIAO C，LIU Z，GUO X. Intelligent robotics and applications[M]. Singapore：Springer International Publishing，2022.
[14] LI P，DING X，REN W. Research on path planning of cloud robot in dynamic environment based on improved DDPG algorithm[C]//Proceedings of the 2021 China Automation Congress (CAC). Beijing，China：IEEE，2021：3561-3566.
[15] WANG Xiaozheng. Research on control of permanent magnet adsorption robot based on improved DDPG algorithm[D]. Daqing：Northeast Petroleum University，2022.
[16] ELARBI F. Genetic algorithm for the design of a class of fuzzy controllers：An alternative approach[J]. IEEE Transactions on Fuzzy Systems，2007，8(4)：398-405.
[17] ZHOU Yi. Research on path planning of mobile robot based on genetic algorithm[D]. Tianjin：Hebei University of Technology，2015.
[18] WEN S，CHEN J，WANG S，et al. Path planning of humanoid arm based on deep deterministic policy gradient[C]//Proceedings of the 2018 IEEE International Conference on Robotics and Biomimetics (ROBIO). Kuala Lumpur，Malaysia：IEEE，2018：1755-1760.
[19] LIU Yunping，JIANG Changsheng，ZHANG Tingting，et al. Finite time ring formation control of multiple UAVs considering internal collision avoidance[J]. Chinese Journal of Mechanical Engineering，2022，58(1)：61-68.
[20] SAMPEDRO C，RODRIGUEZ-RAMOS A，GIL I，et al. Image-based visual servoing controller for multirotor aerial robots using deep reinforcement learning[C]//Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Madrid，Spain：IEEE，2018：979-986.
[21] ZHANG B，PENG J. Uncertain programming model for chinese postman problem with uncertain weights[J]. Industrial Engineering and Management Systems，2012，11(1)：18-25.
[22] UHLENBECK G E，ORNSTEIN L S. On the theory of the brownian motion[J]. Physical Review，1930，36(5)：823-831.
[23] REDMON J，DIVVALA S，GIRSHICK R B，et al. You only look once：Unified，realtime object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston，MA，USA： IEEE，2015：779-788.
[24] REDMON J，FARHADI A. YOLO9000：Better，faster，stronger[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu，HI，USA：IEEE，2017： 6517-6525.
[25] REDMON J，FARHADI A. YOLOv3：An incremental improvement[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Salt Lake City，UT，USA：IEEE，2018：89-95.
[26] GIRSHICK R，DONAHUE J，DARRELL T，et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Columbus，OH，USA：IEEE，2014： 580-587.
[27] GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision (ICCV). Santiago，Chile：IEEE，2015：1440-1448.
[28] REN S Q，HE K M，GIRSHICK R，et al. Faster R-CNN： towards real-time object detection with region proposal networks[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems (NeurIPS). Montreal，Canada：MIT Press，2015： 91-99.
[29] XU H. Design and analysis of a novel magnetic adhesion robot with passive suspension[C]//Intelligent Robotics and Applications：15th International Conference，ICIRA 2022，Harbin，China，August 1-3，2022，Proceedings，Part II. Cham，Switzerland：Springer，2022：369-380.
[30] XU Hao，GUO Weizhong. Wheeled robots：Innovation design and experiments[J]. Journal of Integration Technology，2022，11(4)：3-18.
[31] CHEN Z，WANG Y，CHEN H，et al. A magnetic multi-layer soft robot for on-demand targeted adhesion[J]. Nature Communications，2024，15：644.
[32] CHRISTEN S，NOGUEIRA M G，HILLIGES O. HiDe： Hierarchical reinforcement learning for dexterous manipulation[C]//Proceedings of the IEEE International Conference on Robotics and Automation (ICRA). Singapore：IEEE，2023：7654-7660.
[33] DU D，HAN S，QI N，et al. Reinforcement learning for safe robot control using control Lyapunov-barrier functions[J]. IEEE Transactions on Robotics，2024，40(2)：1050-1065.
[34] RODWELL C，TALLAPRAGADA P. Physics-informed reinforcement learning for motion control of a fish-like swimming robot[J]. Scientific Reports，2023，13：11427.
[35] RADAIDEH M I，WOLVERTON I，JOSEPH J，et al. Physics-informed reinforcement learning optimization of nuclear assembly design[J]. Nuclear Engineering and Design，2023，405：112134.
[36] TAYAL M，SINGH A，KOLATHAYA S，et al. A physics-informed machine learning framework for safe and optimal control of autonomous systems[J]. IEEE Robotics and Automation Letters，2025，10(1)：210-217.
[37] RODWELL C，TALLAPRAGADA P. Physics-informed reinforcement learning for motion control of a fish-like swimming robot[J]. Scientific Reports，2023，13： 11427.
[38] LIU Q，JI Z，XU W，et al. Knowledge-guided robot learning on compliance control for robotic assembly task with predictive model[J]. Expert Systems with Applications，2023，234：121037.
[39] ALGARÍN-PINTO J A，GARZA-CASTAÑÓN L E，VARGAS-MARTÍNEZ A，et al. Intelligent motion control to enhance the swimming performance of a biomimetic underwater vehicle using reinforcement learning approach[J]. IEEE Access，2025，13：3972-3983.