Adaptive neural network control of manipulators with input deadband and field-of-view constraints

Abstract

There exist input dead zone and field-of-view constraints in visual dynamic manipulator systems due to image information loss and uncertainty in actuator physical characteristics, which lead to reduced control accuracy, boundary-induced oscillations, motion instability, and high control energy consumption. In this study, a novel zone Barrier Lyapunov Functions based adaptive neural network visual controller is proposed to address these problems. First, the system model is transformed into an image-based second-order dynamic model, and for the unknown unmeasurable states in the model, a high-gain observer is utilized to achieve accurate estimation. Second, to improve the flexibility of the image trajectory and velocity in the constraint area, a new zone Barrier Lyapunov Functions function is introduced to the visual servoing controller. Additionally, an adaptive neural network approximates the input dead zone model and the unknown part of the system dynamics. Then, the stability of the system and the boundedness of all error signals are theoretically proven using the Lyapunov technique to ensure that the field-of-view constraints are not violated. Finally, comparative simulation and experimental results verify the effectiveness and superiority of the proposed method. The results show that our method can keep the image feature point error around 5 pixels without violating the constraints, and the overall energy consumption of the system is reduced by 83.58% compared to the BLF method and 31.58% compared to the tangent barrier Lyapunov function method, which has significant potential applications in manipulator vision systems.