| This Paper:Browse 1670 Download 1255 |
码上扫一扫！ |
| 基于动态最小生成树的分布式协同目标跟踪算法 |
| 肖林,陆科林,王梦琦,姜博文,冉华明,周礼亮 |
|
|
| (1.中国航空工业集团公司沈阳飞机设计研究所，沈阳110035;2.东南大学 自动化学院,南京 210096;3.空军装备部驻成都地区第三军事代表室,成都 610036;4.中国电子科技集团公司航空电子信息系统技术重点实验室,成都 610036) |
|
| 摘要: |
| 针对多连接通信拓扑传感器网络因可能存在信息冗余和未知相关性而造成融合估计偏差较大的问题，提出了一种具有通信层和融合层的基于最小生成树的分布式协同目标跟踪算法。在通信层，基于Bellman-Ford-Moore算法对多传感器网络进行动态配置，生成具有最小生成树结构的通信拓扑，消除分布式传感器网络中的冗余连接。在融合层，各传感器节点基于树形连接拓扑的通道滤波算法，对其存储的局部数据和从相邻节点接收到的数据进行分布式数据融合，实现协同目标跟踪。同时，以融合估计的动态均方差为性能评价标准设计了一种通用的数据融合评价方法，对该目标跟踪算法的性能进行了分析。该算法可以推广至任意类型的通信拓扑，并可消除冗余通信连接。仿真结果表明，该算法在理论边界范围内的估计误差的比例大于95%，具有良好的估计精度、跟踪性能和鲁棒性。 |
| 关键词: 传感器网络 分布式协同 目标跟踪 最小生成树 通道滤波 |
| DOI：10.20079/j.issn.1001-893x.230817006 |
|
| 基金项目: |
|
| A Distributed Collaborative Target Tracking Algorithm Based on Dynamic Minimum Spanning Tree |
| XIAO Lin,LU Kelin,WANG Mengqi,JIANG Bowen,RAN Huaming,ZHOU Liliang |
| (1.Shenyang Aircraft Design and Research Institute of AVIC,Shenyang 110035,China;2.School of Automation,Southeast University,Nanjing 210096,China;3.The 3rd Military Representative Office of Air Force Equipment Department in Chengdu Region,Chengdu 610036,China;4.CETC Key Laboratory of Avionic Information System Technology,Chengdu 610036,China) |
| Abstract: |
| For the issue of information redundancy and unknown correlation in multi-sensor network with multi-connection communication topology,resulting in significant fusion estimation bias,a distributed collaborative target tracking algorithm based on dynamic minimum spanning tree with communication and fusion layers is proposed.At the communication layer,multi-sensor network is dynamically configured based on the Bellman-Ford-Moore algorithm to generate a communication topology with a minimum spanning tree structure and eliminate redundant connections in the distributed sensor network.At the fusion layer,each sensor node uses a channel filtering algorithm based on a tree connected topology to perform distributed data fusion on its stored local data and data received from adjacent nodes,achieving collaborative target tracking.Using the dynamic mean square error of fusion estimation as the performance evaluation criterion,a universal data fusion evaluation method is designed,and the performance of the target tracking algorithm is analyzed.This algorithm can be extended to any type of communication topology and can eliminate redundant communication connections.The simulation results show that the proportion of estimation errors within the theoretical boundary range of this algorithm is greater than 95%,and this algorithm has fine estimation accuracy,tracking performance,and robustness. |
| Key words: multi-sensor network distributed collaborative target tracking minimum spanning tree channel filtering |
