基于BP神经网络-模糊控制的机器人无标定视觉伺服技术

针对机器人无标定视觉伺服技术中图像雅可比矩阵在线估计存在计算复杂的问题，提出了一种结合BP神经网络和模糊控制策略的机器人控制技术。本文以多自由度智能调节系统为例，提出其视觉伺服控制架构，根据工业场景数据集训练BP神经网络，采用本文所提算法进行法兰对中实验，帮助解决核电站蒸汽发生器人孔螺栓咬死问题。在方法层面，首先，利用BP神经网络建立图像特征信息与机器人多自由度运动之间的映射关系，之后，提出模糊控制方法根据图像特征偏差进行机器人位姿的精确调整。实验结果表明，本文提出的算法能够有效应用于无标定视觉伺服控制，最终法兰平均对中误差在±1mm内，平均耗时43秒，满足应用需求，具有较高的工作效率。

Research on Uncalibrated Visual Servo Based on BP Neural Network and Fuzzy Control for Robotic Application

Aiming at the computational complexity of online estimation of image Jacobian matrix in robot uncalibrated visual servo technology, a robot control technology combining BP neural network and fuzzy control strategy is proposed. This paper takes the multi degree of freedom intelligent regulation system as an example, proposes its visual servo control architecture, trains the BP neural network according to the industrial scene data set, and uses the algorithm proposed in this paper to carry out flange alignment experiments to help solve the problem of the steam generator manhole bolt seizure in nuclear power plants. At the method level, first, the mapping relationship between image feature information and robot multi degree of freedom motion is established by using BP neural network. Then, a fuzzy control method is proposed to precisely adjust the robot''s position and orientation according to the image feature deviation. The experimental results show that the algorithm proposed in this paper can be effectively applied to uncalibrated visual servo control. The average flange alignment error is within ±1mm, and the average time is 43 seconds, which meets the application requirements and has high work efficiency.