Research On The Decision-making Subsystem Of MiroSot Soccer Robot

Robotic soccer is a new interdisciplinary research area, which encompasses diverse fields such as robotics, artificial intelligence, intelligent control, computer vision and so on. Soccer robot system is a typical multi-agent cooperative system. This system provides the theoretical research of robotics, multi-agent system and relative fields a standard test-bed.MiroSot of robot soccer system consists of four components: vision subsystem, decision-making subsystem, communication subsystem and robot car subsystem. In order to conquer opponent in soccer robot game, each parts of soccer robot system must be excellent capability and cooperative work reliably. The victory in the robot soccer is based on the excellent capability and cooperation between parts of the robot system. This thesis only emphasizes our research on some relative technologies in decision-making subsystem, from which we have succeeded in accomplishing the stage works in the system.Firstly, this thesis sums up the system structure of MiroSot soccer robot, introduces the consequence model of decision-making subsystem. Considering the characteristic of MiroSot, chooses the relatively mature three-levels decision-making model and divides the subsystem into coordination level, motion planning level and basic action level.Secondly, focusing on the basic action level, the differential driven mobile robot kinematics model and dynamic model are set up and common used basic actions are introduced. To conquer the defect of motion control mode in existence, a new motion controlling method, PDD~2 controlling method, is developed, which considers the dynamical restriction of robot. The experiment data demonstrate that this method allows a general differential driven mobile robot to reach the target point and given posture rapidly, accurately and steadily.Thirdly, introduces the content of motion planning level and different actions in each action levels. Then expounds the characteristic of path planning in the robot soccer games, introduces typical path planning method and analyses their advantages and disadvantages. Considering the dynamic characters of robot, exposes the conception of autonomous potential field based on the conventional artificial potential field, and designs a new path planning method, the path planning method based on autonomous potential field. The experiment result shows that the new method has good practicality, overcomes the drawbacks of the conventional potential field methods, and adapts to the complex and variational environment in the robot soccer games well.After introduced the constituent of coordination level and discussed the strategy choosing and the role assignment, introduces the structure and flow of the decision-making subsystem software. Finally, introduces the development of modules, which accords with the real-time character, beneath the former decision-making framework.