Neuro-fuzzy system for 3-dof parallel robot manipulator

Planar Parallel manipulators (PPMs) are widely used these days, as they have many advantages compared to their serial counterparts. However, their inverse and direct kinematics are hard to obtain, due to the complexity of the manipulators' behavior. Therefore, this paper provides a comparative analysis for two methods that were used to obtain the inverse kinematics of a 3-RRR manipulator. Instead of the conventional algebraic and graphical methods used for attaining the mathematical models for such manipulators, an adaptive neuro-fuzzy inference structure (AFNIS) model was alternatively employed. It is then compared with a traditional neural network (NN) model for the same manipulator in order to ascertain which model is better in angles prediction, training time and overall performance. The data points used for both training the models and testing their performance are acquired from motion studies in SolidWorks. © 2019 IEEE.