伴随着越来越多的传感器被集成到移动终端,群智感知逐渐成为可能.群智感知的第一步是传感任务的分发,由于移动终端具有移动性,且拥有多样化的硬件设备与感知能力,因而群智感知任务分发不同于普通的传感任务分发.考虑到感知任务的位置固定且节点移动往往基于道路,因而将子道路和任务绑定在一起,设计出基于矢量任务地图的任务分发方法.子道路可以是完整道路中无分叉的一小段道路,也可以由某几个坐标点组成,满足不同覆盖范围的任务.节点接收到任务地图后,由位于相同道路并能够测量该道路上所要求传感任务的移动节点组成传感组,彼此协调完成任务测量.交替进行两种方法实现矢量任务地图的渐进式分发:(1)提高地图分辨率;(2)增加子道路上坐标点数目.通过坐标点精度按位增加的方法提高地图分辨率;通过零阈值Douglas-Peucker算法增加子道路上坐标点的数目.同时,设计新增点坐标信息压缩算法节约矢量任务地图的数据量,其利用新增点在前一低分辨率地图中位于已有点连线上、在当前分辨率地图中位于已有点连线外这一特点,推测新增点的坐标信息.在MATLAB中的模拟实验表明,当矢量任务地图分发给与任务数量相同的移动节点时,任务分发的成功率达到60％以上;当分发给任务数量两倍的移动节点时,任务分发的成功率达到85％以上.同时,采用渐进式分发方法比直接分发更节约数据量且道路坐标点数目的增加与分辨率增加相适应.最后与Zoom进行比较,本文设计的矢量任务地图在1024×1024分辨率下错测量接近0％,且数据量比Zoom降低30％.%With integrated sensors increasing in mobile terminals,terminal users can participate in crowd-sensing as sensing units.The first step towards crowd-sensing is to allocate tasks to terminals.Task allocation in crowd-sensing differs from that in common sensor networks because mobile terminals move disorderly and hardware devices are varied to different terminals.Among the existed methods in crowd-sensing task allocation,sensing server needs to allocate task in real time or makes mobility model of terminals.A new system called Vector Task Map is proposed.This system can work without sensing server.Terminals can be divided into groups and sense environment autonomously after receiving the Vector Task Map.The Vector Task Map combines sub-roads with task information because position of sensing task is fixed and terminal users always move along the road.Most sub-roads are non-furcation part of normal road.Some sub-roads are made by a few of coordinate points if tasks can not combine any real roads.Therefore,Vector Task Map can set tasks over the geographic region in a fine-grained way.Two methods are implemented alternately to allocate Vector Task Map progressively.One is improving map resolution and the other is increasing the number of points on each sub-road.We allocated more precise point coordinate by bit to improve map resolution.We used'ZeroThreshold'DouglasPeucker algorithm to add points.In'ZeroThreshold'Douglas-Peucker algorithm,only coordinate point that is out of all existing lines can be added.Meanwhile,we proposed'Add-Point Coordinate Compression'algorithm to save map data.Because of add-point locating out of existing line in current resolution map and within line in previous low resolution,we can speculate the coordinate of add-point.This paper simulated terminals' movement by MATLAB.The allocation effect of the Vector Task Map is measured by the rate of sub-roads that mobile terminals locate in.The success rate of allocation is above 60％ when the number of mobile terminals is equal to that of tasks.The success rate is above 85％ when the number of mobile terminals is twice as many as tasks.Moreover,the success rate has a room for further improvement with increasing measured interval.Because of low map resolution or a handful of coordinate points on sub-roads,one coordinate may contain multiple tasks in Vector Task Map.This impacts task-group division.We defined this situation as an error rate.At resolution 1024× 1024,error rate of the Vector Task Map is nearly 0％.In terms of data size,progressive task allocation is smaller than direct task allocation.During progressive task allocation,the number of new coordinate points adapts to map resolution improving,too.Meanwhile,the time of decoding Vector Task Map is less than 0.5 second in terminal.Comparing with Zoom in same resolution,the data size of Vector Task Map is always smaller than 30 ％.
展开▼
