Quantum-Interference Artificial Neural Network With Application to Space Manipulator Control

Abstract

Since advanced controllers for space manipulators rely heavily on the accuracy of the dynamic model and state feedback signals, neural-network-based adaptation approaches are often adopted to provide the robustness of the control system. In order to improve the performance of existing neural-network-based adaptive controllers, a new artificial neural network framework is presented based on the quantum-interference principle. A new activation function is established by quantum interference to fulfill the requirement of being a universal approximator. Driven by this new activation function, the classic Delta training method is replaced by an optimal on-line training rule to ensure better performance at a higher training rate (TR), which makes the new neural network more capable of tracking high-frequency noises. The quantum-interference neural network is then integrated into the space manipulator adaptive controller to track the estimation error of the model parameters and disturbances. The advantage of the new neural network at a high TR is validated by simulations, which shows a promising solution to the error tracking and compensation control for space manipulators.