一种基于短期预测的追捕方法的分析与仿真  

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

作　　者：罗真[1] 曹其新[1] 

机构地区：[1]上海交通大学机器人研究所,上海200030

出　　处：《高技术通讯》2006年第9期909-914,共6页Chinese High Technology Letters

基　　金：国家自然科学基金（60443007）资助项目.

摘　　要：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.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.

关 键 词：追捕-逃避 短期预测 收益函数 伪蒙特卡罗方法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]