摘要
随着工业自动化与智能化的发展,巡检机器人在复杂和动态环境中的应用需求日益增长。然而,传统巡检机器人在面对多变环境时,常面临适应性差、执行效率低和协同作业困难等挑战,亟需依托前沿技术进行突破。针对这一问题,本文提出了一种基于具身大模型的智能巡检机器人系统框架,旨在通过多模态数据融合技术、边缘计算与云计算协同机制,提高巡检机器人的智能化水平和应用效能。研究在民航机场的飞机停机坪与维修机库环境中进行实验,采用多种传感器和设备,如高清RGB相机、超声探伤仪、红外热成像仪和激光雷达,通过实时数据融合实现飞机蒙皮损伤的精确检测。系统通过部署轻量化本地模型与云端大模型协同工作,结合材料特性与环境变化,实现了对裂纹、腐蚀和变形等异常的快速识别与精准定位。多机器人协同作业框架通过智能任务分配优化,显著提高了任务分配效率和执行精度。实验结果验证了该系统在提升巡检效率、保障安全和降低人工成本方面的显著优势。综合来看,本研究不仅提出了具身大模型与多机器人协同的新型系统框架,还通过技术融合和系统优化,突破了传统巡检机器人的感知、决策和执行瓶颈,具有重要的理论价值与实际应用前景。
With the advancement of industrial automation and intelligence,the demand for inspection robots in complex and dynamic environments is growing rapidly.However,traditional inspection robots often face challenges such as poor adaptability,low execution efficiency,and difficulties in collaborative operations when confronted with changing environments,necessitating breakthroughs through cutting-edge technologies.To address these issues,this paper proposes a framework for an embodied large language modelbased intelligent inspection robot system,aiming to enhance the intelligence level and application efficiency of inspection robots through multimodal data fusion technology,edge computing,and cloud computing collaboration mechanisms.The experiment is conducted in the aircraft apron and maintenance hangars of Nanjing Lukou International Airport,utilizing various sensors and devices,such as high-definition RGB cameras,ultrasonic flaw detectors,infrared thermal imagers,and LiDAR,to achieve precise detection of aircraft skin damage through real-time data fusion.The system employs a lightweight local model and a cloud-based large language model working in coordination,incorporating material characteristics and environmental changes to quickly identify and accurately locate anomalies such as cracks,corrosion,and deformation.A multi-robot collaborative framework,optimized through intelligent task allocation,significantly improves task distribution efficiency and execution accuracy.Experimental results validate the system's remarkable advantages in improving inspection efficiency,ensuring safety,and reducing labor costs.In conclusion,this study not only proposes a novel system framework based on embodied large language models and multi-robot collaboration,but also breaks through the sensory,decision-making,and execution bottlenecks of traditional inspection robots through technological integration and system optimization,offering substantial theoretical value and promising practical application prospects.
作者
李玉峰
宗国庆
张东豪
LI Yufeng;ZONG Guoqing;ZHANG Donghao(Nanjing Lukou International Airport Technology Co.,Ltd.,Nanjing 211100,China;Nanjing Xuance Intelligent Technology Co.,Ltd.,Nanjing 210023,China)
出处
《智能计算机与应用》
2025年第5期37-43,共7页
Intelligent Computer and Applications
关键词
具身大模型
智能巡检机器人
多模态数据融合
云边协同工作
系统集成
embodied large language models
intelligent inspection robot
multimodal data fusion
cloud-edge collaboration
system integration
