A Dynamic Task Distribution Approach For Mobile Crowd Sensing System

With the rapid development of the sensor network and the popularity of a variety of mobile intelligent devices,the mobile crowd sensing has stepped into the core stage of mobile computing.These intelligent mobile terminal devices have powerful sensing ability and data processing ability because they integrate a large number of sensor element and computing processing chips.Now,research in distribution and obtainment of task has become an important part.How to obtain a simple and low cost sensing task and how to get a task information correctly,should be a worthy problem to solve.Because a large quantity of data amount exists in some task distribution method such as Zoom and how does sensor aquire appropriate raster dimension according its need.An approach of creation of self-adaptive raster-vector mixed map based on quadtree is provided.In addition,there are errors in task map and sensor positioning,so that sensors can't obtain tasks correctly,to solve this problem,a method based on map-matching is proposed.The main work is as follows:First of all,this paper makes a deep analysis and discussion on the mobile crowd sensing network,shows structure and application of the network.We also analyze problems encountered in task distribution.Besides we also introduce the reasons of positioning error and map error.Moreover,we briefly list the existing map matching approaches.Secondly,we introduce an approach of creation of self-adaptive raster-vector mixed map according to requirement of sensors.Through interaction of sever and sensors,server employs quadtree to divide the original vector map according to sensing ability such as error tolerance of sensors,then server can get a raster dimension.Depending on compressibility of quadtree,we compress raster data and reserve key rasters.Finally,we create a self-adaptive raster-vector map by using vector quantization.Compared with Zoom and raster-vector mixed approach without compression,experimental results show that the method of this paper can reduce the amount of data by using the same raster dimension,so that sensors can download less data.Lastly,this paper explains how sensors obtain task information.As for displacement among sensor and task street,which is generated from error of map and sensor localization.So position of sensor can't match the street by formula ax+b=y.We employ different ways to match positions of sensors and different kind of roads.In wide area,we use projection method directly.While in crowded road net,we use method based on historical trail to assist selecting road.Thus,speediness of projection method is be used,historical trail method can select right road.Experimental results show that a comprehensive method can help sensor match the corresponding task street,then the sensor can get correct task instruction.