Investigation Of Android-based Gnss Positioning Models

Currently,smartphones are the predominant tool used by the mass market to obtain geographic location.Due to constraints in both software and hardware,smartphones can only attain a positioning accuracy of 5～10 meters,which,despite being the ubiquitous standard in navigation,is still inadequate for many everyday applications.In 2016,Google Inc,the developer of Android,allowed access to the Global Navigation Satellite System(GNSS)measurements obtained by Android devices;previously,users could only access the position result that the system computed from such measurements.Since then,researchers began to design and implement their own positioning models and algorithms for Android.As the third generation of Beidou satellites complete their deployment,the number of satellites and frequency bands available for navigation is at an all-time high,creating an opportunity to improve positioning accuracy across all GNSS use cases.In particular,research on Android-based positioning models can potentially improve the positioning accuracy of Android devices and,in turn,enhance the user experience for locationbased Android services.This thesis is focused on three areas: the raw GNSS parameters produced by the Android Application Programming Interface(API),GNSS measurements recovered from raw GNSS parameters,and smartphone-tailored improvements to GNSS positioning models.(1)The framework of Android location services and GNSS-related API interfaces were presented.Furthermore,all raw GNSS parameters in Gnss Clock and Gnss Measurement,the two classes in which these parameters are defined,were listed and described in detail.Analysis on parameters in Gnss Clock showed that Full Bias Nanos and Bias Nanos estimated the error of the smartphone’s internal hardware clock,while Drift Nanos Per Second depicted the clock error’s rate of change.Meanwhile,in Gnss Measurement,the uncertainty-related parameters showed a strong correlation with the corresponding Cn0 DBHz,the carrier-to-noise density ratio parameter of the signal measured.(2)The theory behind the recovery of GNSS measurements from smartphone’s raw GNSS parameters was outlined.If the parameters are stored with fewer significant digits than dictated by their data types,the recovery process will result in round-off errors in the GNSS measurements;these errors affect the receiver clock error estimate,but not the position solution.Furthermore,the noise figures of the recovered pseudorange,phase and Doppler measurements were characterized.For Pixel 4,the standard deviation of pseudorange noise for BDS and GPS were 1.88 and 7.05 m,respectively;for Galaxy S8,the corresponding standard deviations were 2.79 and 6.82 m.For both phones,the phase noise varied between-0.1 and 0.1 m,while the standard deviation of Doppler noise was on the order of several centimeters.Carrier-to-noise ratio(C/N0)was only weakly correlated with the satellite’s elevation angle;the lowest C/N0 observed was 15 d BHz.(3)A new single-point positioning model based on C/N0-weighted phase/Dopplersmoothed pseudoranges was introduced.The new model achieved higher positioning accuracy compared to traditional phase/Doppler-smoothing algorithms.Accuracy in North,East,and Up directions was improved on average by 30%,19%,18%respectively for Pixel 4,and 29%,72%,60% respectively for Galaxy S8.