Simulation of a two DOF pneumatic manipulator robot using control based on back propagation neural network

Although pneumatic actuators have a good power-weight ratio, there is a reduced number of works that consider them to be applied in manipulator robots, mainly because they present difficulties to be controlled, and nonlinearities that make them complicated to obtain a mathematical model which describes their behavior in specific conditions. Control techniques can be implemented in this kind of actuators, and they can be complemented with path tracking to follow trajectories established by the user. This paper proposes software implementation of a back propagation artificial neural network to control a two DOF pneumatic manipulator and compares this method against PID control decreasing processing time and oscillations by self tuning parameters of the controller. Path tracking is realized using an algorithm of cubic splines interpoation and the movement of the pneumatic robot is presented in a graphic simulator developed in C++ using OpenGL libraries.