Research On Mobile Crowdsensing Based On Urban Transportation Network

With the development of social productivity,the process of urbanization is also accelerating.However,due to the improper initial planning,there is always a series of "urban diseases".In order to effectively solve these problems from the very beginning,the concept of"smart city" was born.The primary goal of a "smart city" is to solve problems that are closely related to people’s lives,such as traffic congestion,environmental pollution and urban governance.When solving these problems,a lot of urban environmental information is often needed to provide data support for intelligent solutions.Therefore,collecting urban environmental information has become the basis of building a"smart city".In order to perform a better collection of urban environmental data,this paper has studied the mobile crowdsensing based on the urban transportation network.First of all,this paper compares mobile crowdsensing with traditional wireless sensor network,and obtains the characteristics of mobile crowdsensing and its advantages in the scenes of sensing urban environmental data.Secondly,in order to maximize the collection of sensed urban environmental data,this paper builds a model of mobile crowdsensing based on the urban transportation network,and draws a mathematical definaiton of spatial temperol coverage of the city.And point out the key problem of the research is to choose the best vehicle set to participate in the sensing tasks for maximizing its spatial temporal coverage,and to satisfy the limitation of the sensing budget.Then,by analyzing the interaction between the initiator of the sensing tasks and the vehicle,we modeled it as a markov decision process,so that the selection of optimal vehicle set becomes a decision making problem,and the problem description is converted as a bellman equation Finally,considering that in the vehicle selection problem,the state space is too large to be solved by conventional tabular search methods,this paper uses a deep reinforcement learning algorithm to overcome the problem of huge state space,and proposes a deep reinforcement learning based vehicle selection algorithm whose performance on the spatial temporal coverage is verified through simulation.