Design And Implementation Of Indoor Positioning Service System Based On Machine Learning

The current development process of smart cities is gradually accelerating.With the promotion of Internet of Things and cloud computing technologies,indoor positioning technology has been rapidly developed.It is widely used in smart manufacturing,commodity retailing,and security monitoring and other fields.effect.The current indoor positioning solution mainly uses the signal arrival time as the starting point for positioning calculation,and the traditional positioning algorithm will appear jitter in time measurement,which is caused by the crystal oscillation of the timing components,the object blocking in the positioning area and the external environment interference.The time measurement value fluctuates abnormally,causing sudden data changes during the positioning process and affecting the robustness of the system.Based on the above problems,this paper studies location service technology and machine learning algorithms,and designs and develops an indoor location service system based on machine learning.The main work of this article is as follows:(1)The paper uses machine learning algorithms to analyze the motion state,and proposes a distance prediction model based on a gated cyclic unit.The model corrects the distance difference data from the anchor point of the base station,which avoids sudden changes in coordinates caused by crystal oscillation and external occlusion,and improves the reliability of the downstream data of the model.In addition,on the basis of this model,a convolutional neural network positioning calculation model based on distance prediction is proposed for the final position calculation.After a lot of iterative training,the overall model has higher accuracy and lower system error.In addition,after offline training,the complexity of the online model is low,which can meet the requirements of real-time positioning scenarios with strict delay requirements.(2)The paper uses the proposed positioning algorithm to design and implement a set of matching location service application platforms.The thesis analyzes the general requirements of indoor positioning scenarios.Based on the reliability,security and ease of use of the location service application platform,the application platform architecture is designed into three parts: data processing,permission management and functional application,and implemented separately.The data processing part fully describes the data processing process from the acquisition of the original data,integration to the positioning model calculation,location data storage,and positioning data display.The authority management part guarantees the security of the system in terms of identity authentication and access authority.The functional application part includes multiple modules such as device management,gateway management,map management,real-time location display,historical data backtracking,fence warning and user identity management,and is the core of user interaction on the application platform.(3)The thesis carried out functional test and performance test on the actual location service application platform and the proposed indoor positioning algorithm respectively.The application platform has conducted a number of tests including positioning real-time display and historical tracing back,identity verification and access permission verification,device information management and fence warning and reminder.The test results show that the system is stable and reliable.The paper tests the calculation error and running time of the convolutional neural network positioning calculation model based on distance prediction,and analyzes the prediction error under different time window lengths.At the same time,the model proposed in the paper and the other five types of models are calculated from the accuracy And the operation time is compared.The experimental results show that the distance calculation model based on the distance prediction convolutional neural network proposed in this paper has a high comprehensive performance.