Artificial Neural Network Technique to Predict The Sun Position in Cloudy State with Tetrahedron Based Sensor

This work presents an improvement of a dual-axis tetrahedron-based sensor prototype. One of the functions of the tetrahedron sensor is as a sensor tracking the position of the sun to get a more accurate and precise value of the sun's position. This prototype uses three light sensors in the form of a Light Dependent Sensor (LDR) as input and two stepper motors as output. Stepper motor works to control the tilt position on two axes. This concept uses a combination of artificial neural networks (ANN) and PID control. PID control works to control the movement based on the tetrahedron sensor on two axes, while the ANN works as an optimization by tuning the Kp, Ki, and Kd parameters. Thus, the PID parameter is a variable. The network architecture used in ANN is Multilayer Feedforward Network with a backpropagation learning method. PID parameter tuning to control motor movement is performed by ANN. The ANN has two outputs, one each for controlling steppers. The experimental results show that the mean absolute error of reading the sensor is 1.65 for the first stepper output position and 1.15 for the second stepper position.