Application of a Narrow Neural Network Algorithm to Dynamic Analysis Results of a One-Link Planar Robotic Arm

The kinematics, dynamics, control, and optimal design of robotic mechanisms have been widely investigated using various analytical and computational approaches. In recent years, artificial neural network (ANN) algorithms have emerged as an additional tool in mechanism and robotics research. Experimental and theoretical results are commonly analyzed using supervised and unsupervised neural network algorithms. However, many existing studies do not explicitly examine the effects of input–output variations within the employed datasets. In this study, the dynamic modeling and simulation of a one-link planar robotic arm are first presented. Based on the simulation results, datasets are generated and analyzed using a neural network algorithm. One of the primary objectives of this study is to demonstrate the importance of appropriate input–output (predictor–response) selection in neural network applications for robotic mechanisms. A further objective is to contribute ongoing scientific investigations about robotic modeling which use neural network algorithms. The results show that different choices of input and output variables can lead to significantly different prediction performances, highlighting the critical role of dataset formulation in neural network–based robotic analysis.