Controller design of a modeled AdeptThree robot arm

An AdeptThree robots is SCARA robot that has widest working range in its class. The robot has 4 joints, a central processing unit to process input and output data, and an operating system that include its programming language to program the robots. Neural networks are widely used in human application such as in pattern recognizing, forecasting, and scheduling, filtering, and adaptive control. In robot applications, many researchers reported that neural networks are also good controllers for handling the dynamics of robot manipulators. This paper introduces neural networks to control a modeled AdeptThree robot arm. PD controllers were chosen to be the trainer of neural networks in offline mode. The P and D gains of the PD controller were set to meet the control objectives and to be fixed to the time of each joint movement. The result shows that in all cases of the project simulations the neural networks have the least time consuming compared with PD controller and PD controller with saturation.