Neural Network Based Learning and Control Co‐Design for Stewart Platform Control

This chapter closes the gap between the two research areas of model‐free recurrent neural networks for machine learning and model‐based dual neural networks for accurate model control. The proposed networks interact with the learning and control parts. An excitation noise is added to avoid the learning degradation. There are two coordinate systems associated with a Stewart platform, namely a base coordinate system, which is fixed as a global system, and a moving coordinate system of the platform. The chapter introduces a numerical and nonlinear gradient decent optimization method to resolve the real kinematic parameters of the Stewart platform. It demonstrates the efficiency and effectiveness of the proposed model‐free neural network approach applied to redundancy resolution of the Stewart platform by implementing it in MATLAB. Theoretical results are presented to verify the stability of the proposed model‐free neural network approach.