障碍环境下基于生物捕食逃逸行为的多运动体边界防御博弈方法  

Multi-agent border defense differential game based on biological predation and escape behavior in an obstacle environment

作　　者：仝秉达 段海滨 夏洁 刘小峰 Bingda TONG;Haibin DUAN;Jie XIA;Xiaofeng LIU(Bio-inspired Autonomous Flight Systems Research Group,Science and Technology on Aircraft Control Labora-tory,School of Autormation Science and Electrical Engineering,Beihang University,Beijing 100191,China;Peng Cheng Laboratory,Shenzhen 518055,China;College of IoT Engineering,Hohai University,Changzhou 213022,China)

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,飞行器控制一体化技术重点实验室,仿生自主飞行系统研究组,北京100191 [2]鹏城实验室,深圳518055 [3]河海大学物联网工程学院,常州213022

出　　处：《中国科学：信息科学》2022年第12期2213-2224,共12页Scientia Sinica(Informationis)

基　　金：科技创新2030—“新一代人工智能”重大项目(批准号:2018AAA0100803);国家自然科学基金(批准号:U20B2071,91948204,T2121003,U1913602,U19B2033)资助项目。

摘　　要：基于微分对策的追逃博弈和疆土防御问题是多智能体对抗博弈控制的关键问题之一.本文研究了含障碍物的有界区域中多运动体边界防御博弈方法.首先通过对自然界中生物的捕食逃逸行为进行分析,对多运动体边界防御博弈系统进行了建模,得到不同情况下博弈终止条件和价值函数.其次,本文对含障碍环境下博弈双方的主导区域和界栅面进行了分析,并与无障碍情况进行了对比.最后,数值仿真验证了本文提出的多运动体边界防御方法满足微分对策中的鞍点策略条件和有效性.The pursuit-evasion and reach-avoid problems based on differential games are the key problems of multi-agent confrontation control. This paper studies a multi-agent border defense differential game in a bounded area with a line obstacle. First, by analyzing the predation and escape behavior of creatures in nature, the multi-agent border defense differential game system is modeled, and the game termination conditions and value functions are obtained in different situations. Second, this paper analyzes the dominance regions and barrier surfaces of the game in the presence of a line obstacle and compares it with the obstacle-free situation. Finally,numerical simulations show that the proposed method satisfies the conditions of the saddle-point strategy in the differential game and verify the effectiveness of the proposed method.

关 键 词：捕食逃逸 协同控制 边界防御 微分对策 鞍点策略 

分 类 号：Q141[生物学—生态学] O225[生物学—普通生物学]