基于BP神经网络的机械臂视觉伺服控制

Visual servo control of manipulator arm based on BP neural network

针对机械臂视觉伺服系统中求解雅可比矩阵计算复杂导致控制实时性差的问题,利用神经网络来辨识机械臂末端执行器位置与各关节角度之间的关系以得到控制器模型,并在保证控制性能和精度的情况下,将神经网络控制器分为全局与局部两部分,通过切换控制器进一步提高伺服控制精度,同时降低神经网络训练成本。在NAO机器人平台对该方法进行实验,在实验环境中利用视觉对目标进行定位并通过神经网络控制完成抓取,从实验结果和与BP(Back Propagation)神经网络的对比实验证明该方法有效地提高了控制精度。

Aiming at the issue that the Jacobian matrix is poor in real-time because of high computational complexity, a neural network was used to identify the relationship between the actuator position and the joint angle of the manipulator to obtain the controller model. The neural network controller was divided into global and local parts, and the switching controller was used to further improve the servo control accuracy, while reducing the training cost of the neural network. In this paper,the experiment was carried out on the NAO robot platform, and the target was positioned in the experimental environment. The experimental results show that, compared with the BP neural network, the proposed method is more effective and accurate for servo control.