The Collection Technology For Crowdsensing Big Data In Internet Of Things

With the rapid development of ICT,it makes the massive data become the new strategic resources and the driving force of economic and social development.With the implementation of the national big data strategy,more and more big data applications,such as intelligent city,will integrate and improve people's life and work.In order to ease the pressure caused by the growth of massive data requirements,internet of things(Io T)technology has been widely applied and developed.Nevertheless,the large-scale deployment of Io T devices based on traditional wireless sensor networks will lead to huge consumption of resource,installation,maintenance and upgrading.Meanwhile,intelligent terminal devices represented by smartphones have been widely developed,and gradually become an essential part of people's daily life.Currently,the capabilities of the intelligent devices are strong and powerful,the high performance built-in sensors,such as GPS,accelerometer,gyroscope,and camera,etc.,make intelligent devices have great power of sensing and computing.With the popularity of 4G,and the coming of 5G in the near future,high speed mobile networks also greatly enhance the communication capabilities of intelligent devices.In this critical period,crowdsensing emerges as the times require.As a novel collection method of Io T,mobile users collect and share the surrounding environmental data with their intelligent devices.Therefore,crowdsensing will not only reduce the cost of the data collection of Io T,but also make the data collection become flexible,and increase the type of data collected and enrich people's life.However,crowdsensing also faces lots of challenges.Such as in the crowdsensing systems,mobile users need to consume resources(power,memory,CPU,and etc.),to complete the sensing tasks,and the participation of mobile users will have the risk of privacy disclosure.Thus,without rewards return,mobile users will not participate in the crowdsensing systems.How to design the efficient incentive mechanisms,and encourage the mobile users to participate in the mobile crowdsensing is a primary challenge that we are facing at the moment.This dissertation addressed this challenge,and the main contents and results are summarized as follows:1.We investigate the incentive mechanism for the scenario which has only one task and multiple mobile users in one sensing slot,and the intelligent devices have resource constraints and their owners also have stochastic uncertainty demands.We first give the economic models of the crowdsensing system,then we analyze the interaction between the sensing platform and the mobile users by using Nash bargaining theory.More specially,we formulate the interaction between sensing platform and mobile users as a one-to-many bargaining problem,then we study the bargaining solutions under ordered bargaining and simultaneous bargaining systematically.Finally,we design a distributed algorithm based on a dual decomposition method which can not only keep the participators' privacy,but also reduce the sensing platform's computation load.Extensive numerical simulations have been implemented to verify the efficiency of our incentive mechanism.2.In the previous work,one-to-many bargaining method is based on that the sensing platform and mobile users are cooperative to complete the sensing task,this cooperation relationship is only applicable in some special occasion,and it is hard to reach cooperation in most applications.For the non-cooperative scenario,we first give the optimization goals of the sensing platform and mobile users,respectively,and formulate the interaction between sensing platform and mobile users as a Stackelberg game,where the sensing-platform is the leader and the mobile users are the followers.We then analyze the game and prove that there exits a unique Stackerlberg equilibrium,and give the way to compute this unique equilibrium in the static game.Moreover,a dynamic game is investigated,in which the private parameters of the mobile users are unknown to the sensing platform.A deep reinforcement learning discriminated pricing strategy is developed for the sensing platform to determine the Stackelberg equilibrium.Finally,the extensive simulations are conducted to evaluate and validate the performance of the proposed mechanism.3.In the previous two works,we assume that there is only one sensing task in one time slot,and do not consider the quality of sensing data.In this part,we study a critical problem of the payoff maximization in crowdsensing system with incentive mechanism.Due to the influences of various factors(sensor quality,noise,and etc.),the quality of the sensing data contributed by individual users varies significantly.Obtaining the high quality sensing data with less expense is the ideal of the sensing platforms.Therefore,we take the quality of individuals which is determined by the sensing platforms into incentive mechanism design.We propose to maximize the social welfare of the whole system,due to that the private information of the mobile users are unknown to the sensing platforms.It is impossible to solve the problem in a central manner.Then,a dual decomposition method is employed to divide the social welfare problem into sensing platforms' local optimization problems and mobile users' local optimization problems.Finally,distributed algorithms based on an iterative gradient descent method are designed to achieve the close-to-optimal solution.Extensive simulations demonstrate the effectiveness of the proposed incentive mechanism.4.The first three parts of our work are about tasks allocation problem in crowdsensing,and still do not study the data collection problem,while data collection is a key operation.Due to that the resources of intelligent devices for complex computing and sensing tasks are very limited,the design of data collection mechanism is a severe challenge.Based on this situation,we proposed the time sensitive data collection with an incentive aware mechanism.Since the mobile users are selfish,without the rewards return,the relay users will not participate in the data collection.For the sensing data carriers and relay users,their goal is to maximize their own rewards.We formulate the data carriers and mobile users as a cooperative game,and use the asymmetry Nash bargaining model to solve the problem of transfer price decision.We propose two data collection mechanism,one is based on the online D2 D data collection mechanism.Based on the online mechanism,we propose the data collection mechanism which takes the offline mechanism into consideration.For each mechanism,the experiments based on real mobile opportunistic trace data set are implemented to validate the efficiency of our data collection mechanism.At the end of the dissertation,the main results are concluded and the problems which need to be solved in the future are presented.