基于POMDP的多机无源传感器协同任务规划  

Mission Planning of Airborne Passive Sensors for Cooperative Target Tracking Based on POMDP

作　　者：马玲 左燕[1] 彭冬亮[1] 任金磊 MA Ling;ZUO Yan;PENG Dongliang;REN Jinlei(School of Automation,Hangzhou Dianzi University,Hangzhou 310018,China;China Academy of Launch Vehicle Technology,Beijing 100076,China)

机构地区：[1]杭州电子科技大学自动化学院,浙江杭州310018 [2]中国运载火箭技术研究院,北京100076

出　　处：《无线电工程》2022年第7期1260-1265,共6页Radio Engineering

基　　金：国家自然科学基金(61673146,61771028,61973102);电子信息控制重点实验室基金(6142105200110)。

摘　　要：针对多机无源传感器协同跟踪任务规划问题,提出了一种基于部分可观察马尔可夫决策过程(Partially Observable Markov Decision Process,POMDP)的多无人机无源传感器调度算法。在POMDP框架下建立了多无人机协同跟踪规划模型。考虑量测噪声方差距离相关特性,建立了广义克拉美-罗下界(Generalized Cramer-Rao Lower Bound,GCRLB)的目标跟踪长期代价指标。为满足在线规划的实时性,提出了一种基于分布式自主决策算法,仿真验证了所提算法的有效性。An optimal path planning problem is investigated for cooperative target tracking using Unmanned Aerial Vehicle(UAV)equipped with passive sensors.The Partially Observable Markov Decision Process(POMDP)for multiple airborne passive sensors is established to improve the effectiveness of cooperative target tracking.Then,a Generalized Carmer-Rao Lower Bound(GCRLB)metric is designed to adapt to distance-dependent variance of TOA measurement noises and derive the long-time tracking costs.Further,the distributed decision algorithm is developed to solve the scheduling problem.At last,the effectiveness of the proposed algorithm is verified by simulation results.

关 键 词：机载无源传感器 部分可观察马尔可夫决策 广义克拉美-罗下界 分布式决策 任务规划 

分 类 号：TN958.97[电子电信—信号与信息处理]