Research On Indoor Positioning Technology Based On Optimization PDR And Neural Network

With the rapid development of modern network information technology and the continuous enrichment of smart-phone APPs,people’s demand for accurate location services is becoming more and more urgent.Location based on indoor environment has become one of the research hotspots in recent years.At the same time,artificial neural network because of its powerful ability to learn in an environment,it has been widely used in indoor positioning and other fields.Aiming at achieving accurate indoor positioning,the data of global navigation satellite system(GNSS)and performed pedestrian dead reckoning(PDR)was collected in this paper in the outdoor walking stage,then determined the way to achieve indoor positioning by judging the amount of data accumulated in this process.When the accumulated data was enough,neural network could be used to assist PDR to realize positioning,so as to reduce the accumulated error of PDR at this time;If the data accumulation is insufficient,PDR was only used to realize indoor positioning.In order to improve the accuracy of PDR positioning at this time,PDR was optimized from two aspects:its step length and heading angle estimation.The main work and research results of this paper are as follows:(1)The method of indoor positioning based on PDR and GNSS data using neural network was studied.Firstly,collect GNSS data and calculate PDR when walking outdoors.Then,set the tolerance error for the data input into the neural network,the minimum sample size corresponding to the error was estimated by using the hypothesis test combined with the collected data,and whether the neural network can be used to assist indoor positioning was decide by comparing the sample size and the amount of data accumulated at this time.Finally,if the outdoor data accumulation meets the sample size requirements,use neural network to established the mapping relationship between GNSS and PDR.When GNSS disappeared,the predicted output of the network was obtained by inputting the results of PDR into the network model,the prediction output of the network cloud be obtained to improve the accuracy of indoor positioning.The experiment shows that in the process of 400m walking,the error could be reduced within 1.5m by using the above method,and good positioning accuracy was achieved.(2)The methods of improving PDR from two aspects:step estimation and heading angle acquisition were studied.In order to deal with the insufficient sample accumulation during outdoor walking and the inability to use neural network to correct the PDR,the PDR was optimized from the above two perspectives in order to reduce the PDR positioning error at this time.When optimizing the step calculation method,the height was calculated by combining the GNSS data collected in the outdoor walking stage,and then the step was calculated according to the height step model and the calculated real-time step frequency;When acquiring the heading angle,collect the values of gyroscope and direction sensor at the same time,use Kalman filter to fuse the values of gyroscope and direction sensor to obtain the heading angle,so as to reduce the heading angle noise and reduce the cumulative error at the same time.The experimental results show that the optimized PDR has better accuracy,and the walking error of 40m could be maintained at about 0.7m.(3)The analysis and development of indoor positioning system were realized.This paper analyzes the requirements and functions of the system,realized the design of the positioning system through the separation of mobile terminal and server terminal,realized the data collection and abnormal value processing required for indoor positioning,introduced the development of android app in detail,and finally expounded and analyzed the functional modules of the project by using the combination of mobile terminal and server terminal.