Overhead Ball Throwing for Humanoid Robot Soccer

Abstract

This research is a handout to the RoboCup, a football competition for robots that aims to create more intelligent machines. Our task is the design and implementation of legal throw-in from sidelines. Objective is to formulate and test a technique that can be used for the above mentioned task. In football, legal throw-ins are carried out by using both hands, while keeping both the feet on the ground throughout the process. We have designed and implemented this technique on NAO. NAO has three fingers in each hand and to grasp the ball properly throughout the process is a daunting task. Grasp force was important since it is important to keep the ball in hands while lifting the ball behind the head and following the generated trajectory. For this purpose, we initially designed Static Grasping Model and Dynamic Grasping Model. After the calculation of sufficient force, the trajectory generation was planned and executed. We generated circular and elliptical trajectories and tested both separately. After the calculation of the maximum velocity on both the above mentioned trajectories, we planned the release points. At the time of release, we varied the throw-in force equip our robot to choose the destination player accordingly. We designed different methods to maximize the stability. Final results were closer to the expectations. Future works include the use of computer vision algorithms to variate the destination points.