一种基于短期预测的追捕方法的分析与仿真

针对具有安全区的追捕-逃避问题,提出了一种基于短期预测的追捕方法,其算法流程为:追捕方先根据物理条件限制,预测双方可能的短期未来状态,然后评估自身在各种未来状态下的收益,获得一棵决策树,最后根据决策树决定并执行某种行为.讨论了如何确定合理预测时间长度和预测步数等问题,分析表明,预测时间并非越长越好.结合自主式足球机器人一对一对抗的研究实践,进行了仿真实验,仿真结果验证了相关理论分析.

Analysis and simulation of a short-term prediction based pursuing method

A pursuing method based on short-term prediction for the pursuit-evasion problem with safety zone is presented.The algorithmic process is as follows: at first,the pursuer predicts its own possible short-term future states and the evader's;then according to physical limits, the pursuer evaluates its all states with certain reward function,at last, the pursuer selects a state as its short-term target state and executes the corresponding action.The problems of selecting a suitable prediction time and predicting the step number are discussed.A simulation was carried out based on the practice of middle size soccer robots.The result of simulation illustrates the ideas.