Two-agent collaborative shooting method using BP neural network

Abstract

A two-agent collaborative shooting method based on BP neural network was proposed to improve the shooting success rate. In this method, the two players, an assistant player and a shooter, perform a shooting behavior collaboratively using the position relationships among the assistant player, the shooter and the goal. The three shooting parameters, the assistant's passing speed and passing angle and the shooter's spinning rate, are predicted using a BP neural network model, and the relative distances and angles between the assistant passer and the shooter at the beginning of shooting and those between the shooter and the mid-point of goal are used as the input of this BP model. The experiments of simulation match were conducted on the Robot Soccer v1.5a platform. Experimental results show that the collaborative shooting method using BP network has a higher shooting efficiency.