摘要
针对具有很少甚至没有地标和缺乏良好照明条件的大型液化石油气球形储罐环境下,仅依靠里程计以及惯性测量单元的爬壁机器人定位精度较低,且存在较为明显的累积误差的问题,提出了一种新的解决方案,通过跟踪球形储罐表面焊缝相对于机器人的运动来改善机器人定位精度。首先,通过配备辅助光源,利用安装在机器人两侧的CCD相机进行图像采集,并实时对图像进行二值化处理,识别出焊缝特征区域并输出相应的检测信号,通过相邻时刻检测信号增量计算来估计机器人的相对位置;接着,提出了一种改进加权融合算法,用以实现各传感器之间的数据融合,该算法结合了自适应加权融合算法以及基于最小二乘原理的加权融合算法的优点,以对各加权因子重新分配权值的思想来对各个传感器数据进行最终融合,提高了测量精度以及动态适应性。最后,在实际环境以及虚拟环境下对该方法进行了实验评估,验证了该方法的有效性。实验结果表明,在没有良好照明条件及地标的球形储罐环境下,该方法可以保证爬壁机器人定位的准确性。
Aiming at the wall-climbing robot location problem in a large Liquefied Petroleum Gas(LPG)spherical tank with few or no landmark and good lighting condition that the location accuracy of the wall-climbing robot only rely on the odometry and the Inertial Measurement Unit(IMU)was low,and there were obvious cumulative errors,a new solution was proposed to improve the location accuracy of the robot by tracking the movement of the surface weld of the spherical tank relative to the robot.Firstly,carrying out image acquisition by using a CCD camera arranged on two sides of a robot with an auxiliary light source,and binarization processing was carried out on the image in real time to identify the welding seam characteristic area and output the corresponding detection signal.The relative position of the robot was estimated by incrementally calculating the detection signal at adjacent moments.Then,an improved weighted fusion algorithm was proposed to fuse the each sensor data.The algorithm combined the advantages of adaptive weighted fusion algorithm and weighted fusion algorithm based on least square principle,and finally integrated the data of each sensor with the idea of redistributing weight degree of each weight factor,so as to improve the measurement accuracy and dynamic adaptability.Finally,the effectiveness of the method was verified by the experimental evaluation in real and virtual environments.Experimental results show that the proposed method can ensure the accuracy of robot positioning in the spherical tank environment without good lighting condition and landmark.
作者
王凯
孙磊
何宏
Wang Kai;Sun Lei;He Hong(College of Electrical and Electronic Engineering,Tianjin University of Technology,Tianjin 300380,China)
出处
《现代制造工程》
CSCD
北大核心
2020年第5期15-24,共10页
Modern Manufacturing Engineering
基金
国家自然科学基金项目(61871173)
天津市科技计划项目(18YFZCSF00600,18ZXJMTG00160,18ZXZNSY00270)。
