A SHOOTING STRATEGY WHEN MOVING ON HUMANOID ROBOTS USING INVERSE KINEMATICS AND Q-LEARNING

Artificial intelligence and robotics are two fields of study, currently experiencing formidable growth. The RoboCup initiative seeks ways to foster these developments through a series of experiments with new ideas and approaches. Nowadays, the capability of increasing the number of shoots provided by soccer robots is the common goal of all the teams in the league of the RoboCup3D soccer. This article attempts to develop a humanoid soccer robotic shooting strategy in the RoboCup3D using inverse kinematics (IK) and Q-learning while the robot is walking. The vision preceptor on the RoboCup3D soccer simulation has noise and a small calibration error. Accordingly, the robot's moving parameters such as the angle and speed are dynamically optimized more accurately by Q-learning. Finally, if the robot was in the apt position according to the ball and the goal, it triggers the IK to perform the shooting strategy. The simulation results show the superiority of the proposed algorithm compared to most competitors in leagues of Iran's Open RoboCup3D and RoboCup soccer.