The Research On Attack And Defense Decision Of Robocup Mid-size Group Robot

The soccer robot's research was obtaining the fast development in recent years with more and more widespread application of robot. The soccer robot system contains visual processing, wireless communication, multi-robot cooperation and so on many domain technologies, decision system is the core in robot system, it has contained the logic reasoning, the real-time plan, the machine learning and so on. This article takes the object of study by the soccer robot of RoboCup Mid-size Group, mainly analysis and research to the offense and defense decision system of robot systematically.The main attack decision system profited from conceptual model of the plan to design the hierarchical task network, the task decomposition for many sub-tasks, used the SHORE C2 model determine the sub-tasks; The hierarchical task network as the main decision-making of robot, with other decision systems such as path planning, role assignment forms separates and coupling type decision-making, and could add sub-tasks in every nets of the hierarchical task network, it has well portability and improved efficiency of decision-making.Goal-keeper defense strategy of traditional Mid-size Group are used to use single geometric algorithms that difficult to adapt to field movement and ball trajectories and state in competition. Goal-keeper use dynamic decision defend basis on districted field according to dangerous rank and each area characteristic, and joins in returns to the goal state guarantees the goalkeeper can defend nearby own goal in the competition, as for does not have the spatial gate condition.The role assignment system has based on optimized calculation and the simplified fuzzy control, using the Frank-Wolfe arithmetic optimization calculate parameter of overshoot, Select key factor most superior influence time which robot gained ball and use the reduced fuzzy control model after determines the weight of each influence factor for role assignment; introduces the introduction of the role of feedback system in the design, use the broadcast reply type design and the role followed error judges the h role assignment start condition, reduced assigned roles wrongly and has solved role conflict questions and so on. The main attacker and the goalkeeper decision system used the BP neural network to forecast the ball station. The main attacker may run ahead of ball and gain it according to the ball forecasting result in the attack process, and the goalkeeper carried on the pre-defense according to the ball trajectory and the forecasting result, enhanced attack and the defensive rate of accuracy in the actual competition.