Design And Implementation Of Positioning System For Non-Intelligent Lock Bike Based On TensorFlow

With the rapid development of the Internet,the application of computer and Internet technology has been paid more and more attention by the industry.Especially in recent two years,the share economy develops in full swing.Uber and Airbnb,the originator of share economy,of beginning to go to China to carry out business inspired and driven the domestic share economy development.A large number of rookie internet companies in the name of share economy born in infrastructure areas.The typical examples include Yi Ersan,Huijia Chifan,Xiaozhu Duanzu,Didi Chuxing and so forth.In the field of travel,to drop the behavior of the Internet companies,including Shenzhou Zhuanche,Yidao Yongche,Dida Pinche and others have made a lot of innovative work,which gives the majority of users a great travel convenience.However,in the short-distance travel area above the enterprise has no business coverage.In order to solve the urgent demands of short trip which Didi Chuxing cannot cover,large numbers of sharing bike startup set up.Among them,ofo bicycle(the ancestor and leader in sharing bicycle area)and mobike are on the behalf of the first echelon in the market at present.Ofo bicycle is still mainly non-intelligent lock bike,which gives users a bad user experience,while increasing the company's offline operations management difficulties.So the hot issue for company is that how to position the so many non-GPS bikes with higher accuracy to improve the user experience by deep learning or machine learning techniques.For short-distance travel pain points and user needs,based on the traditional learning strategy of non-depth learning,this paper uses the machine learning computing framework of Tensor Flow,which is launched by Google brain team,to implement the function of non-intelligent lock bike positioning system.This paper uses the official version of Tensor Flow 1.0 just released in February this year by Google.The system implemented by this paper uses Python as the development language,Docker as the container carrying the development environment and the Numpy numerical analysis database,and designs and implements two modules and various sub-modules of data preprocessing and data learning.This paper first queries the user riding data information from the backend database,including the longitude and latitude(when start and end orders),the riding time,cycling speed,time interval,and position deviation(two orders),etc.By constructing a multi-layer neural network model,I set up a reasonable learning rate for the effective data and then optimize the loss function for data training and data learning.By solving the problem of over-fitting,I can get the real-time display bicycle position with high accuracy.In this paper,after the realization of the non-intelligent lock positioning system,the system has been fully functional testing and performance testing,and found some loopholes in the system implementation,and timely completion of the modification.The positioning system in the test later in some areas of a city into the operation of the actual production environment,according to the actual operating system's point of view,the implementation of the non-intelligent bicycle lock positioning system can better meet the needs of the company business positioning system.The use of the positioning system significantly improves the accuracy of the non-intelligent lock cycle display at the APP side,improves the user experience and brings a good reputation and image to the company.It also improves the efficiency of the company's operational level,Reducing the cost of offline operations and maintenance management.This paper has completed the expected work.