Research On Mobile Crowd Sensing Method Based On Task Collaboration

Mobile crowd sensing has become a hot topic in the field of pervasive computing,most of the sensing tasks are complex,which need the multi sensing ability of multiple ordinary users and the multiple sensors on the mobile intelligent devices they carry cooperate to perform the sensing tasks.In the process of mobile crowd sensing,a series of tasks need to be carried out,such as publishing perception tasks,selecting appropriate users to complete sensing tasks,uploading and collecting sensing data,etc.For mobile crowd sensing system,it is of great significance to select the right users through reasonable task allocation algorithm.Considering that most of tasks in the city are location-based,a large number of perceptual mobile users are all over the city.This may cause some users to upload false data of non task location and affect the perceived data quality;the task completion rate may not be guaranteed in the area with relatively small number of users;for each task on each location,if the user's perception value and perceived cost are not considered,it may lead to low task completion rate and high perceived cost.Therefore,this paper studies the prediction of user trajectory,the extraction of hot spots,and the calculation of user perception ability and perceived cost,the main work of this paper is as follows:1.Aiming at the problem that the traditional user trajectory prediction algorithm can't predict the user's long-term and regular moving track well,this paper proposes a user trajectory prediction method based on long short-term memory neural network.Firstly,outlier filtering method based on velocity threshold is used to remove noise points in trajectory sequence directly;secondly,the extended spatiotemporal consistency extension dwell point extraction algorithm and k-means clustering algorithm are used to process the dense moving trajectories of participants into a moving track sequence composed of dwell points;finally,the long short-term memory neural network is used to predict the location information and time information of users,which provides the basis for task allocation in the next stage.Based on the real data set Geolife and the simulation results show that the short-term memory neural network can effectively predict the location information and time information of users.2.Aiming at the problem that some users upload false data of non task location,which affects the quality of perceived data,and can not reduce the overall perceived cost and maximize the number of tasks completed,this paper proposes a collaborative task allocation method based on dynamic change of facial user trajectory.Firstly,the long short-term memory neural network is used to predict the user's location information and time information;secondly,the spatiotemporal correlation between tasks and users is quantitatively analyzed,and a bipartite graph network is established;finally,based on the relationship of nodes,the bipartite graph network flow model is used to perform multiple matching between tasks and users.The simulation results show that the collaborative task assignment method for dynamic change of user trajectory can reduce the perceived cost and improve the task completion rate.3.In order to solve the problem of high cost of task perception in non hot areas,and without considering the user's execution ability,the task completion rate can not be guaranteed,a collaborative task allocation method for hotspot discovery is proposed in this paper.Firstly,a multi-density clustering algorithm based on grid relative density is used to find the hotspots and match the task location to the hotspots;secondly,the location similarity and ability similarity between task and user are calculated respectively in the hot area,and the satisfaction matrix of bilateral matching between task and user is established;finally,from the perspective of bilateral satisfaction between tasks and users,a task allocation algorithm based on bilateral matching is proposed.Based on real data set Geolife and simulation results show that the hot spot detection algorithm can effectively extract hot spots,and the collaborative task allocation method based on hot spot detection can effectively reduce the perceived cost and improve the task completion rate.