基于短期预测的追捕方法

针对具有安全区的追捕逃避问题,抛开参与人是否理性的假设,提出了一种基于短期预测的追捕方法.由参与人短期内状态空间有限和伪蒙特卡罗方法原理阐述了进行短期预测的可行性.追捕方法的流程为:追捕方先根据物理条件限制预测双方的短期状态;根据反映对抗态势的特征计算出对手处于某状态时自身某状态的收益,并由此进一步获得针对同时刻对手各个状态的自身某状态的综合收益;估算出自身各个预测状态的综合收益后,选取某个状态作为短期目标,执行能到达该状态的行为.结合足球机器人一对一对抗的实践进行了仿真,结果表明该方法有效.介绍了在自主式足球机器人上应用基于短期预测的追捕方法的初步实验,证实该方法具有一定的实际应用价值.

A Pursuing Method Based on Short-Term Prediction

For the pursuit-evasion games with safety zone,this paper put away the assumption of whether game players are rational or not,and proposed a short-term prediction based pursuing method.With the fact that players' possible states are limited in short-term due to physical limits and the principle of quasi-Monte Carlo,the feasibility of short-term prediction was analyzed.The decision-making procedure is: According to physical limits,at first the pursuer predicts its own and evader's possible short-term future states;then based on features reflecting game situation,the pursuer evaluates the single cost of certain state VS certain evader's state;after that,the pursuer computes the aggregate cost of certain state VS all evader's states at the same time;after all aggregate costs of all states have been gained,the pursuer selects a proper state as its short-term goal and executes the action leading to the goal.The simulation result illustrates that the short-term prediction based pursuing method is effective.It also presented an experiment with middle size soccer robots.The experiment reveals that the method can be applied in real robots effectively.