Research And Implementation Of Parking Space Recommendation System Based On Crowd Sensing

At present,"difficulty in parking" has increasingly become one of the major urban problems faced by major cities in China.Under the urban pattern that has been formed,the lack of parking space resources has led to an urgent need for real-time availability information of parking spaces.This paper proposes a parking space recommendation system based on crowd sensing,which takes advantage of the flexibility and convenience of crowd perception technology to collect on-street parking space information.Realize a kind of “Everyone for me,I am for everyone" design concept.In the process of integrating crowd sensing technology with parking space status sensing tasks,how to motivate more users to participate in sensing tasks,how to verify the authenticity of collected information,and how to make the processing results accurate and usable have always been the focus of research.Therefore,in order to solve these problems,this paper designs and implements a parking space recommendation system based on group perception on the basis of analyzing the parking needs of drivers based on the existing technology.The key research work in this process includes:First,in order to encourage more users to join the perception task,this paper designs an incentive mechanism.At the same time,in order to ensure the authenticity of the perception information,an incentive mechanism optimization algorithm is designed on the basis of the traditional auction incentive model.The user’s uploaded information is verified through the hybrid verification model to ensure the authenticity of the updated information;the additional rewards obtained by the user are adjusted through the dynamically updated reputation value,thereby motivating the user to provide more high-quality perception data.Secondly,in order to strengthen the usability of the recommendation results so that the recommendation results are still accurate and available when the user arrives at the parking space,so as to avoid the influence of the interference of non-platform users on the recommendation results,the advantage of LSTM neural network in time series prediction is used to construct a model based on A one-step prediction model for LSTM neural networks.By optimizing the network structure and the hyperparameters used during training,the prediction results have higher accuracy and stability.Then,in order to enhance the personalization of the recommendation results,the multi-attribute decision-making method is used to design and implement the system recommendation module,which integrates the predicted value of parking space status,the distance from the target location,the difficulty of parking and other attributes,and uses the TOPSIS decision sorting method to generate parking A recommended list of places to recommend ideal parking spaces to users.Finally,the back-end implementation of the incentive module,prediction module and recommendation module of the parking space recommendation system based on group perception is implemented,and a client based on We Chat applet is developed to enable users to upload perception parking space information,obtain incentive points,and update reputation.Value,as well as functions such as free parking space query,and perform relevant interface display and functional testing of key modules.Simulation experiments are carried out on the proposed incentive model.The experimental results show that when the initial information error rate is 0.1,the detection rate of the system is as high as 80%,and the trusted user ratio can still be maintained at a high level in the long-term iterative evolution.Collecting real data to verify the availability of the system recommendation results,the correctness of the system recommendation is as high as93%,which has a good recommendation effect.