Research On Resource Scheduling Problem In Edge-Cloud Architecture Based Participatory Crowdsensing

The frequent and fast communication between mobile devices and their powerful sensing capabilities have led to the emergence of a sensing paradigm called"crowdsensing".Compared with traditional sensing paradigm,crowdsensing is a"lightweight"perception method with three basic elements:users,tasks,and data.Specifically,crowdsensing collects data through a large number of mobile devices equipped with rich sensors in the hands of the crowd,completing large-scale and fine-grained perception tasks without the need to build and operate large specific devices.It can significantly reduce sensing costs while effectively completing tasks.According to the degree of participation of mobile users during the execution of crowdsensing tasks,crowdsensing can be divided into participatory crowdsensing and opportunistic crowdsensing.In opportunistic crowdsensing,mobile users’devices automatically collect data when specific conditions are met.Mobile users complete tasks unconsciously and do not actively participate in the process of executing tasks.In participatory crowdsensing,mobile users will consciously make actions to complete tasks,knowing that their actions are completing tasks published by the platform.Due to the conscious participation of users in completing tasks in participatory crowdsensing,the quality of task completion is relatively high.Users who complete crowdsensing tasks will upload a large amount of data.The traditional centralized crowdsensing architecture is responsible for data processing by the platform,which will cause huge computational storage pressure and costs to the platform,as well as high latency and privacy leakage risks.Mobile edge computing is a new technology that utilizes the computing power of edge devices.Compared with centralized computing architecture,mobile devices in mobile edge computing offload their data to multiple edge devices instead of a single cloud platform.In the scenario of participatory crowdsensing,edge devices are responsible for processing the data uploaded by users and transmitting the processing results to the platform,which reduces the computational pressure on the platform and each mobile user,reduces the time required for data processing,and thus reduces service latency.In edge-cloud architecture based participatory crowdsensing,how to assist the platform in resource scheduling within the scope of the platform’s power,thereby improving the experience of task requesters,is a key issue.For example,how to make tasks as many as possible selected by users,how to make tasks respond as quickly as possible,and how to reduce data processing time as much as possible.Aiming at the three basic elements of users,tasks,and data in crowdsensing,this article studies the resource scheduling problem in edge-cloud architecture based participatory crowdsensing scenarios to help the platform achieve its optimization goals and improve the experience of task requesters.The specific research content is as follows:（1）Aiming at the basic element of task,we studied the task bundling problem in edge-cloud architecture based participatory crowdsensing,where the platform has the authority to bundle tasks.After the platform bundles tasks,mobile users can choose the task bundle they want to complete within their travel budget and complete the tasks they choose.The goal of the platform is to design an appropriate task bundling plan to maximize the number of expected completed tasks.To address this issue,we propose an algorithm for bundling tasks based on the locations of tasks and trajectories of users.The experimental results show that,compared with other algorithms,our algorithm can achieve better performance in maximizing the number of expected completed tasks.（2）Aiming at the basic element of user,we investigated the user transportation problem in edge-cloud architecture based participatory crowdsensing,where each mobile user has its own target tasks.Due to distance reasons,mobile users need to take vehicles arranged by the platform to the task location.The platform has the authority to utilize vehicle scheduling users and can decide how to schedule users by designing user transportation plans.In this issue,the goal of the platform is to design a suitable user transportation scheme to minimize the average response time of all tasks selected by the users.We propose a heuristic algorithm called Similar Direction to solve the problem for situations where the vehicle capacity is sufficient to transport all passengers.The algorithm consists of three stages:Passenger Number Determination,User Assignment,and Delivery Order Determination.We have demonstrated that the approximation of the Delivery Order Determination algorithm is 2(8,where(8 is the passenger capacity of each vehicle.We propose three algorithms,Del Far,Unchanged,and Del Ran,to solve the problem for situations where the vehicle capacity is insufficient.We have demonstrated the effectiveness of the algorithm through experiments.（3）Aiming at the basic element of data,we studied the data offloading problem and the payment determination problem in edge-cloud architecture based participatory crowdsensing,where edge devices can help users process data,and will receive the payment paid by users after the data processing is completed.The goal of the platform for the data offloading problem is to develop a data offloading plan to maximize the reduced data processing time of all users with the help of edge devices.The goal of the payment determination problem is to help the platform develop a payment plan based on the data offloading plan to ensure that the bid submitted by each user is true.In order to simultaneously solve these two problems,we propose a data scheduling mechanism based on primal-dual technology and demonstrate that the data offloading plan generated by this mechanism has the approximate ratio of₁₊,and its payment plan can ensure the truthfulness of each user.Considering the feasibility of the data offloading plan,we have made slight modifications to the data scheduling mechanism to ensure that the output data offloading plan must be a feasible solution to the data offloading problem.We conducted a large number of experiments and verified the effectiveness of the proposed data scheduling mechanism.