分布式传感器网络下弹道目标跟踪算法研究

This project is devoted to state estimation of ballistic targets with distributed sensor network. Some algorithms will be proposed using intelligent recognition method for nonlinear filtering and queuing theory for data fusion in accordance with the characteristics of ballistic target motion and network coverage ratio. There are four main themes within the proposed research. The first is to construct the model of ballistic target tracking system involving the motion characteristics of ballistic target and coverage ratio of network. The second theme is to explore the existence and robustness of optimum track-to-track fusion strategies in the target tracking system. In this direction the proposed research involves optimum queuing theory approach and thus requires a detailed study of the corresponding optimization problems: opportune fusion moment and fusion time to fulfill the desired accuracy of target state estimation. The third theme is to explore the function relationship between the parameters and the accuracy of particle filtering. The research in this direction leads towards novel particle filtering, which involves intelligent recognition method for any switching nonlinear system. The last theme is to confirm the feasibility of the induced tracking algorithms based on the above strategies and filtering. In this direction the proposed research incorporates computer simulation and the experiment of some mobile robot tracking with wireless sensor network. These research findings are beneficial not only to perfection of target tracking theory, but also to design of some practical applications system including distributed satellite, missile defense and so on.

本项目以提高对弹道目标跟踪的实时性和精确性为目标，针对传感器网络覆盖特点及弹道目标运动特性，提出分别将随机服务系统理论和智能识别理论与目标跟踪理论相结合，研究分布式传感器网络条件下弹道目标跟踪的状态估计及航迹融合算法。主要内容包括：（1）建立携带目标不同阶段运动特征及传感器网络覆盖特点的弹道目标跟踪系统模型；（2）研究传感器网络数据处理系统中最佳融合时机和时长问题，提出基于随机服务系统理论的最优航迹关联算法，并分析算法的鲁棒性；（3）研究粒子滤波中各待定参数与估计精度的函数关系，提出基于智能识别理论的粒子滤波算法；（4）对所提算法进行仿真，并在无线传感器网络移动机器人平台上进行物理验证。本项目的研究不仅能丰富目标跟踪理论，而且能为分布式卫星，导弹防御等实际系统提供设计方案，具有重要的科学意义和应用价值

自本项目启动以来，项目组严格按照计划任务书中的要求积极开展研究工作，针对传感器网络覆盖特点及弹道目标运动特性，提出分别将投影理论和并行化理论与目标跟踪理论相结合，研究分布式传感器条件下目标跟踪的状态估计及算法并行化。针对杂波环境中非机动扩展目标跟踪，提出了最优集整数规划算法；针对目标出现机动的情形，将最优集整数规划算法与经典的交互式多模型算法相结合，提出基于最优集整数规划的交互式多模型算法；针对传感器量测统计信息未知的情形，将 滤波引入到最优集整数规划的交互式多模型算法中，提出更具鲁棒性的机动性扩展目标跟踪算法；对于在轨跟踪的目标，提出了充分利用轨道信息的滤波算法，并给出了在各种轨道情形下获得目标状态估计值的充分条件；针对传感器位置布设与目标间的相对运动，提出将传感器位置与目标位置共同作为联合状态向量进行估计预测的对角型交互式多模型算法；针对交互型多模型算法在各维度的切换同规则性，提出将交互式多模型算法进行并行化处理，在不影响估计精度的情况下节约计算时效。针对硬件实现中所遇到的图像去雾、图像风格化等问题给出了一些初步的基础性研究成果。