Research on the Decision System Model of Robocup3D Robotics Team
|Course||Computer Software and Theory|
|Keywords||robotic football decision system role assignment matrix adjustment Border rules partial order sequence voting algorithm|
Humanoid robotics is a research direction of a comprehensive witch combines many discipline. It is the explorer for more intelligent robots. The multi-agent system problem (Multi Agent System) in the field of Artificial Intelligence has been the focus research of many scholars. As intelligent bodies, robots often coordinate and cooperate with each other to solve a unified task. We need divide the large task into multiple small tasks and assign role location to each robot. This leads to the problem that how to assign role location to each robot of a football team and the method to solve team decision-making problems.In this paper, several problems are been solved on the background of Robocup3D simulate robot soccer match, as follows:Firstly, An overview is displayed about the historical development of humanoid robot development and Robocup project development. Focus on Robocup3D robot soccer simulation project, the3D simulation platform architecture, rules of the project and the way and rules of the robot and server communication are described in detail.Secondly, in order to reduce the decision time of robots, speed up the formation convergence as well as unify the individual decision and the team decision, a hierarchical model of team decision system is built based on the latest RoboCup3D simulation platform. It supports a specific framework to realize the lineup control, role assignment and cooperation. Based on the method of adjusting matrix minimally each time, an global optimal role assignment algorithm, which is provided for robot individual decision and costlier than previous algorithms, is implemented. By comparative experiments, the final results show that the decision time of robots is much reduced, the unification of team decision and the convergence speed of formation are all improved, the collisions between the robots are also reduced, and finally the combat capability of robot team is significantly improved.Thirdly, because of the bandwidth limitations of the broadcast information, how to find the optimal candidate in case of partial order sort, rather than total robot preference ranks of candidate. Therefore, a voting algorithm based on partial order Border rules is proposed to solve the partial order optimal selecting. It provides a good voting algorithm for team’s robot to deicide.Ideas and algorithms presented in this thesis has been applied to our actual team, the team’s overall combat effectiveness has been greatly improved.