Forward Kinematics Development of a 3-RRS Parallel Manipulator for Real-Time Control

In control of parallel robots, forward kinematics calculation is mainly required. In some studies, obtained equations are solved numerically with massive calculations that is not acceptable for control implementations. This paper introduces a new method to extract forward kinematics for a 3- RRS robot. This method calculates the kinematics of passive joints with two separated equations, which can be solved to decrease the total computation demands. Also, a new weight compensation mechanism is proposed to improve the dynamic response by a linear pneumatic actuator placed between the ground and the end-effector. On the other hand, to use various control algorithms, the position of the rolling ball on the end- effector (moving platform) must be measured. This is performed using Convolutional Neural Networks (CNNs) as image processing algorithms. Since the system must work quickly for a high-frequency task, the proposed image processing algorithm can provide the desired fast feedback. Experimental results reveal the merits of the above-described proposed modules.