Opponent-Aware Ball-Manipulation Skills for an Autonomous Soccer Robot

Autonomous robot soccer requires effective multi-agent planning and execution, which ultimately relies on successful skill execution of individual team members. This paper addresses the problem of ball-manipulation for an individual robot already in possession of the ball. Given a planned pass or shoot objective, the robot must intelligently move the ball to its target destination, while keeping it away from opponents. We present and compare complementary ball-manipulation skills that are part of our CMDragons team, champion of the 2015 RoboCup Small Size League. We also present an approach for selecting the appropriate skill given the state of the world. To support the efficacy of the approach, we first show its impact in real games through statistics from the RoboCup tournament. For further evaluation, we experimentally demonstrate the advantages of each introduced skill in different sub-domains of robot soccer.