Deep Regression on Quaternion: A Forward Kinematics Neural Network for Study Six-DoF Pose

Solving the forward kinematics problem of robots with high degrees of freedom using neural networks is a really innovative field of study. However, conventional neural network has not been adequate to address this mission requirements of the robot under it’s large-scale data with significant unpredictability and low coupling. Thus, we have analyzed and discussed this methodology, and have proposed three solutions, using empirical theory of visual tasks. The results of the experiments demonstrate that our proposed model was effective.