Neural Network Based Visual Servo Control Under the Condition of Heavy Loading

This paper proposes an image-based visual servo (IBVS) control system for hoist positioning under the condition of heavy loading. Various advanced visual servo methods for positioning have been proposed. However, the visual servo scheme under heavy load condition remains the following problems: 1) The deformation and oscillation of the end-effector under the condition of heavy loading will lead to coordinate deviation of end-effector and visual frame, which makes the system unstable and brings more errors. 2) The vibration velocity and offset caused by deformation and oscillation are difficult to measure and model. To address these concerns, we first model the system error based on visual error and differentiate it. We obtain that the basic vector of the system error derivative contains the vector of vibration velocity, then we use the basic vector of derivative of the system error to build up the relationship between system error and vibration velocity. Moreover, we construct a double parallel feed forward neural network to estimate the basic vectors which contain vibration velocity and offset, and compensate for them. Also, we give the proof of the boundedness of the parameters based on Lyapunov theory. Simulation results depict the superior performance of the proposed method against the other advanced visual servo methods under the condition of heavy loading. Furthermore, experiments are conducted on the hoisting platform in our laboratory, which verifies the effectiveness of the proposed method.