伴随化工生产设备运维工作量的增大以及人工巡检成本的不断升高，利用巡检机器人代替人工对化工车间关键设备进行定期巡检和实时监测，具有自动化、灵活性和安全性方面的独特优势。针对某化工车间巡检面积大、目标点位多、巡检导航任务多样化、导航定位精度要求高的特点，论文建立了化工车间巡检机器人定位导航软硬件系统平台，开展了建图、定位和导航等关键算法研究，并完成了工程应用验证。 根据化工车间巡检机器人实际工程应用需求，本文从硬件结构、软件系统、数据交互三个方面完成了兼具激光SLAM和视觉辅助定位功能的巡检机器人定位导航系统设计，建立了巡检机器人的运动控制模型和里程计观测模型。 针对化工车间为室内环境以及作业面积大的特点，本文在真实环境实测对比分析四种主流激光SLAM建图算法基础上，设计了一种基于EKF融合里程计和分段多级阈值重采样的改进Gmapping激光SLAM建图算法，实现了化工车间高精度地图构建，建图相对误差低于1.5%。 针对化工车间巡检点位多、巡检导航任务多样化的特点，本文设计了一种基于Dijkstra算法思想和目标点邻接矩阵的包含中间点的全局路径规划算法，实现巡检机器人多模式巡检导航任务，目标点导航平均精度小于30mm，满足用户需求。 针对个别目标点导航误差偏大的情况，本文以充电桩目标点为例设计了一种全局ORB特征匹配和局部SIFT特征匹配相结合的目标点位置视觉定位算法，成功获取机器人实际到达位置与标准目标点位置的位置差，从而实现机器人在目标点位置处的位姿修正。测试表明该算法在相关目标点平均定位误差小于10mm。
With the increasing workload of chemical production equipment operation and maintenance, as well as the increasing cost of manual inspection, using inspection robots to replace manual inspection and real-time monitoring of key equipment in the chemical workshop has unique advantages in automation, flexibility, and safety. In response to the characteristics of a certain chemical workshop with a large inspection area, multiple target points, diversified inspection navigation tasks, and high requirements for navigation and positioning accuracy, the paper establishs a software and hardware system platform for the positioning and navigation of the chemical workshop inspection robot, studies the key algorithms such as mapping, positioning, and navigation, and completes engineering application verification. Based on the actual engineering application requirements of the inspection robot in the chemical workshop, this paper completes the design of a positioning and navigation system for the inspection robot that combines laser SLAM and visual assistance positioning functions from three aspects: hardware structure, software system, and data interaction. The motion control model and odometer observation model of the inspection robot are established. In response to the indoor environment and large working area of the chemical workshop, this paper proposes an improved Gmapping laser SLAM mapping algorithm based on EKF fusion odometer and segmented multi-level threshold resampling, based on the comparative analysis of four mainstream laser SLAM mapping algorithms in real environment measurements. This algorithm achieves high-precision map construction of the chemical workshop with a relative error of less than 1.5%. In view of the characteristics of the chemical workshop with many inspection points and diversified inspection navigation tasks, this paper designs a global path planning algorithm containing intermediate points based on Dijkstra algorithm idea and target point adjacency matrix, which realizes the inspection robot multi-mode inspection navigation tasks. The average accuracy of target point navigation is less than 30mm, which meets user needs. In response to the large navigation error of individual target points, this paper takes the target points of Charging station as an example to design a target point position vision positioning algorithm combining the global ORB feature matching and local SIFT feature matching, which successfully obtains the position difference between the actual arrival position of the robot and the standard target point position, so as to realize the pose correction of the robot at the target point position. Tests have shown that the algorithm has an average positioning error of less than 10mm at relevant target points.
