分布式多传感器组网协同跟踪方法研究

发布时间：2018-11-26 17:44

【摘要】：随着计算机、传感器、通讯技术的飞速发展,各种面向复杂应用的多传感器系统大量涌现,将多个传感器信息进行有效的融合是提高目标探测性能的有效途径。有限的传感器能力和不确定目标环境使得传感网无法利用所有传感器跟踪同一个目标。如何选择有限的传感器,通过不同传感器之间的通信和协同完成跟踪任务是传感器组网协同跟踪的核心。随着传感器数目和目标数目的增多,协同跟踪的复杂程度随之增加。在大规模传感网络中,受容量和处理能力限制,采用分布式处理可以避免单传感器故障对整个网络性能的影响,同时降低通信要求。本文在前人工作的基础上,从不同的分布式融合框架以及传感器资源管理优化方法的角度研究了分布式多传感器组网协同跟踪问题,主要研究的内容如下:1.针对大规模传感网融合中心处理能力限制,考虑传感网内各传感器采样异步情况下,提出了一种分布式异步传感网协同跟踪算法。首先,对监视区内布署的传感器进行聚类,构建有限个子网;其次,基于最大信息增量的原则进行传感器选择,确定下一时刻各子网内参与协同跟踪的传感器集,并对参与协同跟踪的传感器进行异步顺序融合跟踪;随后采用多跳方式进行传感器之间信息传递,确定最优通信传输路径,并进行全局数据融合;最后,仿真验证了算法的可行性。2.在监视区域存在多个目标情况下,考虑坐标变换造成传感器量测误差时变影响,提出了一种时变量测方差下基于PCRLB的多传感器多目标协同跟踪算法。首先,给出了时变量测误差的描述,分析了时变量测误差和跟踪性能的影响,建立了时变量测误差下PCRLB指标。随后基于PCRLB优化指标进行多传感器多目标分配,确定下一时刻用于跟踪目标的传感器集。同时在滤波跟踪过程中,考虑坐标转化带来的传感器量测方差变化,通过转换卡尔曼滤波实现时变量测方差下的多传感器协同跟踪,最后,仿真验证了算法的可行性。3.针对大规模传感网中传感器之间通信传输的限制,提出了一种PCRLB的分散式传感器网协同跟踪算法。采用树拓扑结构的分散式传感器网络结构,设计分散PCRLB指标。基于分散PCRLB指标进行传感器选择,通过并行滤波的方式进行局部融合中心的估计。仿真通过与集中式融合算法比较,分散式多传感器协同跟踪算法具有一定的优势。
[Abstract]:With the rapid development of computer, sensor and communication technology, a large number of multi-sensor systems for complex applications have emerged. The effective fusion of multi-sensor information is an effective way to improve the performance of target detection. The limited sensor capacity and uncertain target environment make it impossible for the sensor network to track the same target using all sensors. How to select limited sensors, communicate among different sensors and cooperate to complete tracking task is the core of cooperative tracking in sensor network. As the number of sensors and targets increases, the complexity of cooperative tracking increases. In large scale sensor networks, due to the limitation of capacity and processing capacity, distributed processing can avoid the effect of single sensor failure on the performance of the whole network and reduce the communication requirements. Based on the previous work, this paper studies the cooperative tracking problem of distributed multi-sensor network from different distributed fusion frameworks and sensor resource management optimization methods. The main research contents are as follows: 1. In view of the limited processing capacity of the fusion center of large-scale sensor networks, a distributed cooperative tracking algorithm for asynchronous sensor networks is proposed, considering the asynchronous sampling of each sensor in the sensor network. Firstly, the sensors deployed in the surveillance area are clustered and the finite subnet is constructed. Secondly, based on the principle of maximum information increment, sensor selection is carried out to determine the set of sensors involved in cooperative tracking in each subnet at the next moment, and the sensors involved in cooperative tracking are tracked by asynchronous sequential fusion. Then the multi-hop method is used to transfer information between sensors, to determine the optimal communication transmission path, and to perform global data fusion. Finally, the simulation verifies the feasibility of the algorithm. 2. A multi-sensor multi-target cooperative tracking algorithm based on PCRLB is proposed under the condition of multiple targets in the monitoring region and considering the time-varying influence of coordinate transformation on the measurement error of the sensor. Firstly, the description of time-varying measurement error is given, the influence of time-varying measurement error and tracking performance is analyzed, and the PCRLB index under time-varying measurement error is established. Then the multi-sensor multi-target assignment based on the PCRLB optimization index is carried out to determine the sensor set for tracking the target at the next time. At the same time, in the process of filtering and tracking, considering the change of sensor measurement variance brought by coordinate transformation, the multi-sensor cooperative tracking under time-varying measurement variance is realized by transforming Kalman filter. Finally, the feasibility of the algorithm is verified by simulation. 3. Aiming at the limitation of communication and transmission between sensors in large scale sensor networks, a distributed tracking algorithm for sensor networks based on PCRLB is proposed. The decentralized PCRLB index is designed by using the distributed sensor network structure based on tree topology. The sensor is selected based on decentralized PCRLB index, and the local fusion center is estimated by parallel filtering. The simulation results show that the decentralized multi-sensor cooperative tracking algorithm has some advantages compared with the centralized fusion algorithm.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212

【参考文献】