基于移动传感器网络的城市交通监测与估计

发布时间：2018-06-09 05:19

本文选题：交通监测 + 移动传感器网络　；参考：《上海交通大学》2014年硕士论文

【摘要】：随着移动终端、传感器技术和通信技术的迅猛发展，集信息感知与处理、信息传输与发布和移动控制于一体的移动传感器网络已逐步进入人们的视野。比起传统的传感器网络，移动传感器网络的移动能力提升了网络的覆盖性、连通性等性能，因此被广泛应用于环境监测的场景中，动态、实时地获取区域内的监测数据。本文将以城市的交通监测为背景，研究移动传感器网络中的数据感知与处理，数据估计以及移动控制的问题。在城市交通背景下，通常以城市车辆及智能手机作为移动传感器节点。前者可利用车载设备感知交通信息并借助车联网实现数据传输，但当前需要解决由于节点分布不均而引起的感知问题，包括如何通过控制少量节点的移动以提升网络的覆盖性能，以及如何估计未采样的数据；后者借助蜂窝网等网络实现数据传输，但当前需要解决如何通过机载传感器有效感知城市交通信息数据的问题。基于现有的VSN网络收集的车辆移动轨迹数据，本文进行了数据处理，形成表示交通信息的交通矩阵。同时我们发现由于参与交通感知的车辆有限，因此交通矩阵的采样率较低，因此提出了使用矩阵还原的方法对其中未采样的数据进行估计。由于交通数据具有非负性以及除高峰期外的时间缓变性，我们改变了原始的优化问题的构造，从而使得估计结果更加合理。位了提高现有解决矩阵还原优化问题的算法运算速度，我们提出了基于L1/2范数的迭代矩阵还原算法。为了进一步提高算法性能，我们通过数据分析，发现由于VSN中车辆分布的不均匀，导致了使用矩阵还原算法的效果与最理想情况下的效果存在一定的差距。因此我们考虑通过控制其中一部分可控车辆的轨迹从而改变交通矩阵的采样。我们从信息熵的角度分析了同意道路相邻时刻交通信息的相关性，并通过函数拟合，建立估计误差于信息熵之间的联系，将减少估计误差的优化问题变成最小化采样信息熵的优化问题，并提出所需满足的采样规律。基于该规律，我们提出了基于巡逻的移动控制策略。由于我们得到了估计误差与采样信息熵的关系，因此交通信息中心可以根据VSN采集的交通数据判断缺乏数据的道路，并鼓励位于这些道路上的手机进行交通感知。当手机用户在车上时，可以通过手机GPS和无线通信定位，计算车辆移动速度。为了自动判断用户是否在车上，我们提出使用手机上的陀螺仪、磁力计与加速度计，计算手机垂向加速度，，并根据垂向加速度和移动速度进行识别。在论文最后，我们对全文进行了总结并讨论了未来进一步的研究工作。
[Abstract]:With the rapid development of mobile terminal, sensor technology and communication technology, the mobile sensor network, which integrates information perception and processing, information transmission and distribution, and mobile control, has gradually entered the field of vision. Compared with the traditional sensor networks, the mobility of mobile sensor networks improves the network coverage, connectivity and other performance, so it is widely used in environmental monitoring scenarios, dynamic, real-time access to monitoring data in the region. In this paper, the problems of data perception and processing, data estimation and mobile control in mobile sensor networks are studied under the background of urban traffic monitoring. Urban vehicles and smart phones are usually used as mobile sensor nodes. The former can use vehicle-mounted equipment to perceive traffic information and realize data transmission by means of vehicle networking, but it is necessary to solve the problem of perception caused by uneven distribution of nodes at present. Including how to improve the coverage performance of the network by controlling the movement of a small number of nodes, and how to estimate unsampled data, which can be transmitted through networks such as cellular networks. However, it is necessary to solve the problem of how to effectively perceive the urban traffic information data through airborne sensors. Based on the data collected by the existing VSN network, this paper processes the data to form a traffic matrix to represent the traffic information. At the same time, we find that the sampling rate of the traffic matrix is low because of the limited number of vehicles involved in traffic perception, so we propose a matrix reduction method to estimate the unsampled data. Due to the non-negative traffic data and the slow change of time except the rush hour, we change the structure of the original optimization problem and make the estimation results more reasonable. In this paper, we propose an iterative matrix reduction algorithm based on L ~ (1 / 2) norm. In order to further improve the performance of the algorithm, we find that due to the uneven distribution of vehicles in the VSN, there is a certain gap between the effect of matrix reduction algorithm and that of the optimal case. Therefore, we consider changing the sampling of traffic matrix by controlling the trajectory of some of the controllable vehicles. From the point of view of information entropy, we analyze the correlation of traffic information at adjacent time, and establish the relationship between estimation error and information entropy through function fitting. The optimization problem of reducing the estimation error is transformed into the optimization problem of minimizing the entropy of sampling information, and the sampling law that needs to be satisfied is put forward. Based on this rule, we propose a mobile control strategy based on patrol. Because we can get the relationship between the estimation error and the entropy of sampling information, the traffic information center can judge the road that is lack of data based on the traffic data collected by VSN, and encourage the mobile phone located on these roads to carry out traffic perception. When the mobile phone user is in the car, the mobile speed can be calculated by GPS and wireless communication. In order to automatically judge whether the user is in the car, we propose to use gyroscopes, magnetometers and accelerometers on the mobile phone to calculate the vertical acceleration of the mobile phone, and identify it according to the vertical acceleration and the moving speed. We summarize the full text and discuss further research work in the future.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.116;TN929.5;TP212.9

【参考文献】