Role allocation tactics of soccer robots on RoboCup3D simulation platform

A Markov Decision Process (MDP) is employed to optimize the role allocation of soccer robots on RoboCup3D simulation platform in this paper. Role allocation tactics of soccer robots are modeled based on Markov Decision Process (MDPs) considering various factors of five different types of robot in simulation environment including distance, orientation, falling, speed and other factors. Then a linear function approximation based on Sarsa(λ) learning algorithm is applied to solve the action value function of the MDP model. The role of search for the optimal allocation scheme improves the efficiency of the entire team's offensive defense. The experimental results establish the effectiveness of our methods.