基于潜博弈的多传感器协同跟踪分布式优化算法  

Distributed Optimal Algorithm Based on Potential Game for Multi-sensor Collaborative Tracking

作　　者：张淯铧 林庆宝 左燕[1] 彭冬亮[1] ZHANG Yuhua;LIN Qingbao;ZUO Yan;PENG Dongliang(School of Automation,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学自动化学院,浙江杭州310018

出　　处：《无线电工程》2025年第3期672-678,共7页Radio Engineering

基　　金：国家自然科学基金(61673146);浙江省自然科学基金重点项目(LZ23F030002)。

摘　　要：针对部分可观多传感器多目标协同跟踪问题,提出了一种基于潜博弈的分布式优化算法。以传感器为博弈方,选择广义Fisher信息矩阵(Generalized Fisher Information Matrix,GFIM)为跟踪收益函数,将探测和通信约束下的多传感器多目标分配问题描述为一个局部信息博弈模型,证明了该模型是一个潜博弈模型,至少存在一个可行的纯策略纳什均衡(Nash Equilibrium,NE)。为了提高计算效率,设计了一种改进并行最佳响应动态(Modified Parallel Best Response Dynamic,MPBRD)的分布式决策算法,分析了算法的复杂度。仿真结果显示,在小规模场景下,基于潜博弈的分布式优化算法能够达到集中式全枚举优化算法的跟踪性能,计算时间大大缩短。在大规模场景下,基于潜博弈的分布式优化算法具有较好的收敛性,满足大规模传感器决策的实时性需求。A distributed optimal assignment algorithm based on potential game theory is proposed to solve the partially observable multi-sensor collaborative tracking problem.The sensor is selected as the game player and the multi-sensor multi-target assignment problem under detection and communication constrains is constructed as a game model based on local information with the target tracking reward function of Generalized Fisher Information Matrix(GFIM)metric.The model is proved to be a potential game model with at least one feasible pure strategy Nash Equilibrium(NE)point.A Modified Parallel Best Response Dynamic(MPBRD)algorithm is developed to solve the game problem efficiently.Then its computation complexity is analyzed.Simulation results show that in small-scale scenario,the proposed algorithm can achieve the tracking performance of the full enumeration algorithm in greatly reduced time,and in large-scale scenario,it has a convergence good enough to meet the real-time decision requirements of large-scale sensor network.

关 键 词：多传感器协同跟踪 潜博弈 纳什均衡 改进并行最佳响应动态算法 

分 类 号：TN958.97[电子电信—信号与信息处理]