An application in RoboCup combining Q-learning with adversarial planning

RoboCup is a standard problem so that various theories, algorithms and architectures can be evaluated. Behavior learning for complex tasks is also an important research area in RoboCup. In this paper, we present a new approach to solve the kick problem in RoboCup Simulation, which combines Q-learning with online adversarial planning. This method not only achieves satisfactory learning effect, but also solves the adversary kick problem to some extends.