Research On Crowdsensing Transfer And Its Application In The Smart Parking Lot

In recent years,a new round of technologic and industrial revolution is developing towards the smart city which everything is interconnection.The rapid popularization of molibe portable devices,the high-speed network in the 5G era,and the rapid improvement of cloud computing capabilities have made the smart cloud and the big data analysis becoming future trend to make full use of computing resources and transmit real-time results more effectively and accurately.Mobile Crowdsensing(MCS)is a real-time prediction system that combines large-scale crowd perception and model computing resources to give target task results.Now,MCS is commonly researched and applied in the field of smart cities,traffic management,environmental monitoring,social event analysis and so on.This article analyzes the quality of questions answered by the crowd,discusses how to choose the number of control questions and the number of available parking spaces,studies the number of available parking spaces and retention time,proposes the crowd incentive algorithms and built parking information system.The main contents of this paper summarized as follows:1)An algorithm based on Crowdsensing for selecting the least number of workers with the highest quality is proposed.Firstly,we obtain the parking space information uploaded by the workers and generate the corresponding reliability judgement questions based on the uploading informations.Then,according to the correct rate about the answers which is set to test the reliability of workers,we can use it as the first value of the worker’s reliability.In next,the EM algorithm is used to update the reliabilities of worker.At last,we can use joint estimation algorithm to predict the available parking spaces and related informations.2)An algorithm for predicting parking space retention time based on transfer learning is proposed.First,the historical parking data and the first spatial feature corresponding to the historical parking data are acquired.The geographic location of the target parking space is acquired as the second spatial feature.Then,we should obtain the common space features between the first spatial feature and the second spatial feature.Based on the common features and the historical parking data,a transfer learning model is trained.Finally,we can use the trained model to predict parking space occupancy rate at some time in the future.Through analysis the occupancy rate,the parking space retention time under this period will be obtained.3)A marginal incentive algorithm based on game will be adopted.Firstly,the feature function of the worker’s ability is obtained from the worker reliability calculated above,and the reward distribution scheme is calculated based on the feature function.Then,the ability of workers is related to their own efforts.Finally,the connection between the amount of reward and the margimal effort will be built.And the superiority of the incentive algorithm is judged by maximizing fairness.4)A mobile crowdsensing system whose task is to search real-time available parking spaces is obtained.We design multi-user smart applications that run on mobile phones and algorithm models as well as background manager running on cloud servers.Therefore,workers and people who want to park can access,upload and receive data information through the app at the same time.When the management system not only conforms to the concept of smart cities and big data transportation,it can effectively facilitate parking and alleviate traffic congestion,but also for increasing the additional income of low-income people and promoting its application in other directions such as air quality collection and social behavior analysis and so on.