|||
网络协同“一致性”研究如火如荼十余载,源起复杂网络控制,后发展至分布式滤波和跟踪。但是现有所有的分布式滤波基本都是“滤波-融合-再滤波-再融合”这种“你方唱罢我登台,轮番上阵”的串行方式,或者说“鸡生蛋,蛋孵鸡”这种相互依赖迭代模式:所融合的对象就是滤波结果,而下一轮滤波的先验就是融合结果。
0-1的突破:
提出了 “一边滤波一边通讯融合”并行模式(即网络通讯融合与节点滤波计算同步进行,Parallel Filtering-Communication mode),难点在于:没有鸡(滤波)怎么来的蛋(融合)?没有蛋(融合)又怎么来的鸡(滤波)?这就是我们Engineer们 施展拳脚 了。。。
团队近期两篇论文(全世界独此两篇?):
第1篇基于粒子滤波,借助Importance Sampling方法,巧妙实现了高度Parallel Filtering-Communication ;
第2篇基于高斯混合GM滤波,Parallel Filtering-Communication 实现难度较大,所实现的融合对象仅仅是目标数cardinality的估计,融合层次教浅有待进一步研究。
两个工作分别对最具代表性的两类滤波后验近似形式进行了“滤波与通信并行”机制设计,可以推广到诸多以粒子滤波和高斯混合为基底的其他多传感实时滤波器设计。
网络通讯与节点滤波计算同步进行的优势自不必说,甚至有些场景下是唯一选择!比如局部节点的滤波计算占据了整个传感器扫描周期,根本没有剩余时间去搞通讯和信息融合 ---- 而这将伴随着传感器扫描周期越来越快变得普遍。。即使有些剩余时间,因为平行通讯可以完成更多周期的信息交换,传播更远,网络收益更大。。
(Submitted on 17 Dec 2017 (v1), last revised 20 Dec 2018 (this version, v2))
We propose a particle-based distributed PHD filter for tracking an unknown, time-varying number of targets. To reduce communication, the local PHD filters at neighboring sensors communicate Gaussian mixture (GM) parameters. In contrast to most existing distributed PHD filters, our filter employs an `arithmetic average' fusion. For particles--GM conversion, we use a method that avoids particle clustering and enables a significance-based pruning of the GM components. For GM--particles conversion, we develop an importance sampling based method that enables a parallelization of filtering and dissemination/fusion operations. The proposed distributed particle-PHD filter is able to integrate GM-based local PHD filters. Simulations demonstrate the excellent performance and small communication and computation requirements of our filter.
Comments: | 13 pages, codes available upon e-mail request |
Subjects: | Systems and Control (cs.SY); Distributed, Parallel, and Cluster Computing (cs.DC) |
Cite as: | arXiv:1712.06128 [cs.SY] (or arXiv:1712.06128v2 [cs.SY] for this version) |
submission | Information Fusion,under revision |
2.
Abstract:
This paper proposes a new cardinality consensus (CC) approach called “prior-CC” to distributed probability hypothesis density (PHD) filtering based on a decentralized sensor network. In our approach, network-wide average consensus is sought with respect to the prior cardinality estimate. Unlike existing serial filtering-communication approaches, the prior-CC scheme allows the internode communication to be performed in parallel with the local filter calculation and requires only a small amount of interfacing fusion calculation and communication. This enables real-time filtering that minimizes data delay and is of great significance in realistic tracking systems. We provide details of the Gaussian mixture implementation of the proposed prior-CC-PHD filter based on a diffuse target birth model and analyze the filtering-communication parallelization. In addition, we evaluate the gain of the prior-CC scheme with respect to the filtering accuracy in comparison with the standard CC scheme via simulations using a stationary linear sensor network and a dynamic nonlinear sensor network, respectively.
Published in: IEEE Sensors Journal ( Early Access )
Date of Publication: 22 June 2020
DOI: 10.1109/JSEN.2020.3004068
相关博文链接:
分布式网络信息共享:Many Could Be Better Than All
Archiver|手机版|科学网 ( 京ICP备07017567号-12 )
GMT+8, 2024-11-25 13:57
Powered by ScienceNet.cn
Copyright © 2007- 中国科学报社