【作者】 王东；

【导师】 汪晋宽；

【作者基本信息】 东北大学 ， 通信与信息系统， 2009， 硕士

【摘要】 作为一种新兴的IT热点技术,无线传感器网络在军事与民用诸多领域有着广阔的应用前景。在军事领域,基于无线传感器网络的目标跟踪技术可以应用于导弹防御、海陆空防御和作战监视等；在民用领域,可以用于交通管制、导航及其机器人的道路规划和避碍等,因此对目标跟踪进行研究具有重大的理论和实际意义。由于在传感器网络中传感器节点处理能力有限并且携带能量有限,因此设计出简单高效的目标跟踪算法对于延长传感器网络的生命期以及增强传感器网络的健壮性有着非常重要的意义。目标跟踪的极端复杂性与无线传感器网络节点能力相对不足形成矛盾,分布式是解决这对矛盾的有力武器。本文从跟踪的能量消耗、跟踪精度、跟踪的鲁棒性和跟踪反应时间四个指标出发,提出了一种由中心计算机、移动目标和无线传感器网络组成的分布式目标跟踪算法,通过邻域节点间的协同工作,克服了单个节点的能力不足,满足了目标跟踪的应用需求。在算法中提出的退避定时机制用于首领节点的选取及传输序贯的确定；改进的扩展卡尔曼滤波算法提高了滤波的精度和稳定度。为了实现实时而准确地运用已知无线传感器网络对目标进行定位跟踪,本文结合粒子滤波和卡尔曼滤波的各自特点提出了JPEKF (Jointed Particle and Extended-Kalman Filter)算法。改进的滤波算法利用粒子滤波对初始误差不敏感的特点获取目标比较准确的初始位置估计,克服了扩展卡尔曼由于初始误差较大而引起的发散问题；然后使用扩展卡尔曼滤波进行目标的跟踪保持从而保证了系统的实时性。当目标出现丢失的情况时,系统将启动重定位机制,JPEKF滤波算法也将重新初始化,这种机制可以迅速定位丢失的目标保证了系统的稳定性。仿真结果表明提出的这种滤波算法具有很高的精度和稳定性以及较低的计算复杂度,利用节点间的协同工作,能有效地对目标进行跟踪。更多还原

【Abstract】 As a new attractive IT technology, wireless sensor network (WSN) is very promising in many military and civil applications. In the military field, the target tracking technolodge based on wireless sensor network can be applied to missile defense, air defense and surveillance operations, etc. In the civilian areas, it can be used for traffic control, navigation and robot path planning and obstacle avoidance, etc. Therefore the study of target tracking is of great theoretical and practical significance.Since the limited energy of the sensors and the weak processing capability, it is significant to design an alogorithm with low complexity as well as high efficiency. There is a contradiction between the extreme complexity of target tracking and the low capacity of each sensor node. Distributed is an essential characteristic of WSN, which is also a powerful measure to solve this contradiction. Through the collaborative work among all neighboring nodes, the lack of powerful capcity for a single node is overcome in order to satisfy the application requirements.In order to realize real-time and accurate localization and tracking of mobile target on known wireless sensor network, we propose a novel algorithm:JPEKF (Jointed Particle and Extended-Kalman Filter). This alogrthm has two modules:it first applies insensitiveness of Particle Filter to initial estimation error to catch the target, which avoids the divergence due to the big error in initial estimation when using Extended Kalman Filter (EKF), and then uses EKF to keep the tracking of target. When the target is missing, the system will start relocalization mechanism and JPEKF also will initialize.The simulation results show that JPEKF is a robust algorithm with high accuracy and computing efficency.更多还原