Deep Learning Based Kinematic Modeling of 3-RRR Parallel Manipulator

Abstract

This paper presents a novel low cost design for a 3-RRR Planar Parallel Manipulator (PPM). These manipulators proved their superiority over serial manipulators due to their speed, precision and smaller work space where the work space area is accounted for in the design to ensure that the robot is performing its task in a smooth and simple way without getting into any singularity points. The challenge with PPM is to obtain the kinematic constraint equations of the manipulator due to their complex non-linear behavior. Screw theory is a new approach that is used to compute the direct and inverse kinematics based on the relation between each link and its’ predecessor. The design is then inserted into ADAMS to study its dynamical behavior and to obtain a data set that would be used in analyzing the system in MATLAB. A Neuro-Fuzzy Inference System (NFIS) model was constructed in order to predict the end-effector position inside the work space and it is tuned with Particle swarm optimization (PSO) and Genetic algorithm (GA).