【作者】 黄中林；

【导师】 邓平；

【作者基本信息】 西南交通大学 ， 通信与信息系统， 2010， 硕士

【摘要】 定位是无线传感器网络(WSN)投入实际应用所需的关键技术之一,高精度、低复杂度的定位算法对无线传感器网络具有重要意义。传统的无线传感器网络定位算法仅适用于节点静止的情况,当节点移动时,会因为周期性重复定位消耗大量的能量,降低网络使用寿命。因此,迫切需要研究适合于移动场景的无线传感器网络定位、目标跟踪技术。为此,本文重点研究了基于序列蒙特卡洛方法(SMC)的移动无线传感器网络定位技术和基于粒子滤波(PF)的无线传感器网络目标跟踪技术。首先,在查阅大量相关文献的基础上,综述了国内外无线传感器网络定位的研究现状。分别从静态无线传感器网络定位和移动无线传感器网络定位两个方面进行了介绍,重点分析了其中的典型算法,并对典型算法的性能进行了比较,给出了移动无线传感器网络定位的性能评价标准。其次,作为移动无线传感器网络定位和目标跟踪的理论基础,详细介绍了适用于非线性、非高斯情况下的贝叶斯滤波和粒子滤波技术,并对扩展卡尔曼滤波(EKF)、无迹卡尔曼滤波(UKF)和标准粒子滤波(PF)的性能进行了仿真分析。结合无线传感器网络能量受限的特性,文中还讨论了粒子滤波在无线传感器网络中的五种运行模式,对基于分布式粒子滤波的无线传感器网络目标跟踪算法进行了仿真分析。再次,对序列蒙特卡洛定位(MCL)、蒙特卡罗盒定位(MCB)和多跳蒙特卡罗定位(MMCL)算法的特点、性能进行了详细的理论分析,并分别从定位精度、样本数、节点密度、锚节点密度、节点运动速度、定位覆盖率等方面进行了MATLAB仿真与性能分析。在此基础上,提出了一种基于自适应采样蒙特卡罗盒(AMCB)的移动传感器网络定位算法,仿真结果表明改进算法在保持定位精度的同时有效地减少了计算量。针对MMCL算法在锚节点密度低时定位精度高的特性,结合DRL动态洪泛锚节点信息节约通信量的优点,提出了一种跳数自适应的移动无线传感器网络蒙特卡罗定位(AMMCL)算法,仿真分析表明,AMMCL在提高定位精度的同时有效地减少了通信量。最后,本文对开展的研究工作进行了总结,并给出了进一步研究的内容和建议。更多还原

【Abstract】 Localization plays a crucial role in wireless sensor networks (WSN) as most of WSN applications need the awareness of the node’s location. Localization algorithms with high accuracy and low complexity are very important for WSN. Traditional wireless sensor networks localization algorithms only apply to a static situation, when the nodes move, lots of energy will be consumed by the periodically repeated localization. Therefore, it is urgent to study the mobile localization and tracking algorithm which is suited for mobile wireless sensor networks. This thesis focuses on mobile wireless sensor networks location techniques based on Sequential Monte Carlo (SMC) method and target tracking based on particle filter (PF).First of all, this thesis introduces an overview of wireless sensor networks localization algorithms at home and on abroad, and analyzes the classical localization algorithms of static wireless sensor networks and mobile wireless sensor networks, and then compares the performance of classical location algorithms. The criterion of performance evaluation of mobile wireless sensor networks localization is also summarized.Secondly, this thesis introduces the particle filter which is suit for non-linear, non-Gaussian, the Bayesian filtering theory and particle filter (PF) are described in detail. The properties of Extended Kalman filter (EKF), unscented Kalman filter (UKF) and standard particle filter are simulated and analyzed. Take the energy-constrained characteristics of the wireless sensor networks into account, this thesis introduces five operating mode of particle filter which are applied in wireless sensor networks, simulates and analyzes the target tracking algorithms of distributed particle filter in wireless sensor networks.Thirdly, this thesis analyzes the performance and characteristics of Monte Carlo localization (MCL) and its improved algorithm, such as Monte Carlo localization Boxed (MCB), Multi-hop-based Monte Carlo localization (MMCL), then compares their performance from the aspects of localization accuracy, sample number, node density, anchor node density, node velocity and localization ratio base on MATLAB simulation. Based on MCB algorithm we propose a sample Adaptive Monte Carlo Localization Boxed (AMCB) mobile localization algorithm, simulation results demonstrate that the proposed algorithm produces good localization accuracy as well as low computational cost compared with MCL and MCB. Aim at the speciality of the localization accuracy of MMCL algorithm is high when the anchor density is low, take the merit of DRL dynamically changing of anchor’s flooding-hop into account, we propose a Adaptive Multi-hop-based Monte Carlo Localization (AMMCL) algorithm. Simulation results show that the proposed algorithm produces better localization accuracy as well as low communication cost compared with MMCL.Finally, the research work of this paper is concluded and the future research topics are presented.更多还原