Desain Perilaku Home Position Pada Robot Soccer Dan Kursi Bergerak Otonom Menggunakan Algoritma Q-Learning

Abstract

Robot soccer is a type of wheeled car robot that has complexity in the design of the robot. The complexity in the design of the robot makes research development in the field of robotics increasing. This study discusses home position behavior (back to its original point) using the Q-learning algorithm which is implemented on robot soccer. Behavior system is implemented, the robot can move randomly in tracing the color box area to reach the target point. On the other side, using a chair as an object that will be randomly controlled with the Q-Learning algorithm combined with a webcam camera as a sensor to detect the color of the robot's movement area. Color acts as an initial identification so that the robot easily recognizes and knows the target point and area of ​​its movement.  From the results of tests that have been done, the optimal behavior of robots to reach the target point is purple-yellow-blue with a value of Q = 5.13 and the time response value is time peak (tp) = iteration 12, time rise (tr) = iteration 8, time settling (ts) = iteration 9, time delay = iteration 6, and the value of the steady state error = 0, while in an autonomous moving chair with a value of Q = 5.13, the number of iterations 9, the number of steps is small and the response time response as follows: time rise (tr) = 9 iterations, time peak (tp) = 7 iteration, time settling (ts) = 9 iterations, delay time (td) = 6 iterations and the steady state error value is zero (0) on the 9th iteration.