多阵列水下多目标跟踪的分布式算法研究

Study on Distributed Algorithm for Multi-Array Underwater Multi-Target Tracking

针对水下单水听器阵列探测范围受限、目标定位和跟踪性能不足问题,论文提出一种基于匹配场定位量测模型的分布式多目标联合定位与跟踪方法。在各阵列节点选取匹配场模糊函数大于设定阈值的峰对应的坐标作为量测,结合势平衡多伯努利(Cardinality Balanced Multi-Bernoulli,CBMB)滤波,滤除噪声干扰,解决常规匹配场定位方法低信噪比导致的跟踪精度下降的影响。在分布式网络架构下,利用广义协方差交集(Generalized Covariance Intersection,GCI)融合法则,序贯融合每个阵列节点与其邻近节点各自滤波后的多目标后验概率密度,以充分利用不同阵列节点的量测信息,提高水下多目标的跟踪精度。由于融合的是多目标后验概率密度而非量测集本身,改善了集中式融合处理的高通信负担问题。仿真实验结果证明,与单水听器阵列目标跟踪算法相比,经分布式融合后,多次蒙特卡洛实验下的平均最优子模式分配(Optimal Subpattern Assignment,OSPA)距离显著下降,多目标的状态和数目的跟踪精度有所提升。在系统通信负担和单节点计算负担大幅降低的情况下,可达到与集中式融合处理相当的跟踪精度。

The paper proposes a distributed multi-target joint localization and tracking method based on the matched field localization measurement model to address the issues of limited detection range and insufficient target localization and tracking performance of underwater single hydrophone arrays.The proposed method aims to address the limitation of con-ventional matched field localization method,which relies solely on selecting coordinates that match the number of targets of interest.To mitigate the adverse impact of mismatched acoustic field parameters and the actual environment on localiza-tion accuracy,the proposed method selects coordinates corresponding to peaks in the ambiguous functions generated after matched field localization at each array node that exceed a set threshold as measurements.The measurements are then sub-jected to the cardinality balanced multi-Bernoulli filtering algorithm to filter out noise interference.Under a distributed net-work architecture,to fully utilize the received information of different array nodes in the distributed fusion structure and thus improve the tracking accuracy of multiple underwater targets,the multi-target posterior probability density obtained from each array node is sequentially fused with its neighboring array nodes through the Generalized Covariance Intersection fusion law.Due to fusing the posterior probability density of multiple targets instead of the measurement set itself,the method improves the high communication burden problem of centralized fusion processing.The simulation results demon-strate that distributed fusion leads to a significant decrease in the average Optimal Subpattern Assignment(OSPA)dis-tance,and a clear improvement in the tracking accuracy of both the state and cardinality of multiple targets compared to the single hydrophone array multi-target tracking algorithm from multiple Monte Carlo experiments.Under the condition of significantly reduced communication and computation burdens on the system and individual nodes,it is possible to achieve tracking accuracy comparable to centralized fusion processing.