Analyzing Android GNSS Raw Measurements Flags Detection .
Analyzing Android GNSS Raw Measurements FlagsDetection Mechanisms for Collaborative Positioning inUrban EnvironmentThomas Verheyde, Antoine Blais, Christophe Macabiau, François-XavierMarmetTo cite this version:Thomas Verheyde, Antoine Blais, Christophe Macabiau, François-Xavier Marmet. Analyzing AndroidGNSS Raw Measurements Flags Detection Mechanisms for Collaborative Positioning in Urban Environment. ICL-GNSS 2020 International Conference on Localization and GNSS, Jun 2020, Tampere,Finland. pp.1-6, 10.1109/ICL-GNSS49876.2020.9115564 . hal-02870213 HAL Id: hal-02870213Submitted on 17 Jun 2020HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.
Analyzing Android GNSS Raw Measurements FlagsDetection Mechanisms for Collaborative Positioning inUrban EnvironmentThomas Verheyde, Antoine Blais, Christophe Macabiau, François-XavierMarmetTo cite this version:Thomas Verheyde, Antoine Blais, Christophe Macabiau, François-Xavier Marmet. Analyzing AndroidGNSS Raw Measurements Flags Detection Mechanisms for Collaborative Positioning in Urban Environment. ICL-GNSS 2020 International Conference on Localization and GNSS, Jun 2020, Tampere,Finland. pp.1-6, 10.1109/ICL-GNSS49876.2020.9115564 . hal-02870213 HAL Id: hal-02870213Submitted on 17 Jun 2020HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.
Analyzing Android GNSS Raw MeasurementsFlags Detection Mechanisms for CollaborativePositioning in Urban EnvironmentThomas VERHEYDETéSA Research LaboratoryToulouse, FRANCEthomas.verheyde@recherche.enac.frAntoine BLAISENAC Research LaboratoryToulouse, FRANCEantoine.blais@enac.frChristophe MACABIAUENAC Research LaboratoryToulouse, FRANCEchristophe.macabiau@enac.frFrançois-Xavier MARMETCentre National d’Études SpatialesToulouse, FRANCEfrancois-xavier.marmet@cnes.frAbstract—The release of Android GNSS raw measurements, inlate 2016, unlocked the access of smartphones’ technologies foradvanced positioning applications. Recently, smartphones’ GNSScapabilities were optimized with the release of multi-constellationand multi-frequency GNSS chipsets. In the last few years, severalpapers studied the use of Android raw data measurements fordeveloping advanced positioning techniques such as PrecisePoint Positioning (PPP) or Real-Time Kinematic (RTK), andquantified t hose m easurements c ompare t o h igh-end commercialreceivers. However, characterizing different smartphone modelsand chipset manufacturers in urban environment remains anunaddressed challenge. In this paper, a thorough data analysiswill be conducted based on a data collection campaign thattook place in Toulouse city center. Collaborative scenarios havebeen put in place while navigating in deep urban canyons.Two vehicles were used for this experiment protocol, equippedwith high-end GNSS receivers for reference purposes, whileseven smartphones were tested. Android algorithms reliabilityof both the multipath and cycle slip flags w ere investigatedand evaluated as potential performance parameters. Our studysuggests that their processing may differ from one brand toanother, making their use as truthful quality indicators forcollaborative positioning yet open to debate.Index Terms—Android Raw Measurements, Cycle Slip Flag,Multipath Flag, Collaborative PositioningI. I NTRODUCTIONIn May 2016, Google announced that GNSS raw data measurements will be available on Android smartphone devices viatheir latest Android Application Programming Interface (API)called Android Nougat (7.0). This innovation allowed developers and the scientific c ommunity t o o btain a ccess t o GNSSmeasurements from embedded smartphones receiver. Code,phase, Doppler and C/N0 data can now be retrieved fromAndroid’s mass market receivers. Following this milestone,mobile chipset manufacturers started to develop innovativetechnology, including the newly announced Broadcom BCM47765 dual-frequency, multi-constellation chipset [1].978-1-7281-6455-7/20/ 31.00Commercial opportunities rose from this technologicalachievement, that led mobile manufacturers to compete inorder to obtain the world’s most precise smartphone. Multiplephone companies joined the race, releasing dozens modernsmartphones, equipped with various chipset models, that aremulti-frequency and multi-constellation ready. Those technological progress potentially unlocked the access of a widecrowdsourced and connected network of embedded smartphone receivers.In the last few years, several research studies explored thepossible implementation of advanced GNSS processing techniques (e.g. PPP, RTK) on Android mass market device [2][3]. Positioning performances were then compared to lowcost and high-end commercial receivers. Most of those worksfocused on one phone model in optimal conditions withoutcharacterizing other smartphones and chipset brands.Android based positioning is most of the time performedin constrained environments around urban areas. The mainchallenge associated with positioning in an urban environment,is signal degradation caused by disruptive multipath and NonLine of Sight (NLOS) signals reception. Apprehending thosedifficulties is even more challenging while using embeddedsmartphones’ linearly polarized antennas. Antenna design architecture limitations make them unoptimized for acquiringmulti-frequency GNSS signals.On the other hand, the Android positioning API providesdetection mechanisms in the form of flags in order to detectmultipath and cycle slip occurrences. However, their detectionalgorithms are unknown and coping with those flags couldbecome ambiguous.To overcome those issues, a thorough study has been conducted during a data collection campaign in Toulouse citycenter. In the interest of developing a global smartphone qualification method, an analysis was made on seven smartphonesin a constrained environment. An assessment of Android flags
will be presented as an introduction to smartphone performance parameters. In future work, the identified performanceparameters will be used and exchanged in a collaborativesmartphone network. This shared data would help network’susers to qualitatively and quantitatively assess their smartphone’s performances.This article will be articulated in three main sections. First, thedata collection campaign will be presented. Then, a detailedinvestigation on multipath and cycle slip flag algorithms isconducted. Finally, this paper will be concluded by a discussion on how to integrate performance parameters in acollaborative smartphone network.II. DATA C OLLECTION C AMPAIGNOur data collection campaign took place in August 2019,in Toulouse city center. The goal of this campaign was toaccurately depict urban conditions encountered by Androidusers. A fleet of two vehicles was used along a specifictrajectory as shown in figure 1. Collaborative scenarios wereestablished along the way. This data collection campaign lastedfor 2 hours and 10 minutes.A. Experimentation ProtocolsThe two vehicles were both equipped with high-end GNSSCommercial-Off-The-Shelf (COTS) equipments, a NovAtelSPAN receiver coupled with a high-grade IMU unit, a Septentrio PolaRX5 and a Ublox F9P for reference purposes. Table Ilists all smartphones analyzed during this data campaign. Eachmobile was securely placed on their assigned car’s rooftopfor the entire duration of the data collection. All smartphoneswere multi-constellation (Galileo, GPS, GLONASS & Beidoucompatible) and multi-frequency (L1/L5 & E1/E5a) except forthe Google Pixel 3 that was single-frequency. The smartphoneselection process was decided in function of their brand andmodel as shown in table I and were running Android Pie 9.0.GNSS Raw data measurements were recorded by each devicein a .log format using the GNSSlogger application. Finally,it was noticed that the Google Pixel 3 and the Xiaomi Mi9 did not record any phase measurements data for the entirecampaign whereas the other Android phones recorded themcorrectly.Fig. 1. Data Collection Campaign Vehicles’ Trajectory.TABLE IA NDROID S MARTPHONES A NALYZEDCar iaomiHonorSmartphonesModelChipsetMi 8Broadcom BCM 47755Mi 9Qualcomm Snapdragon 855Pixel 3Qualcomm Snapdragon 845View 20HiSilicon Kirin 980Mate 20XMi 8View 20HiSilicon Kirin 980Broadcom BCM 47755HiSilicon Kirin 980B. Collaborative ScenariosThroughout the collection campaign, collaborative scenarioswere implemented. Figure 1 shows the different cooperativescenarios created. Each scenario is represented by a letterand a picture, taken by our on-board camera, illustrating theenvironment condition. The first one, labeled A in figure1, represents a test case in nominal conditions (open-sky).Scenario B illustrates Android-based positioning in a deepurban environment. The third test case (C) defines a collaborative event between two users with one being in goodreception condition and the other, on the contrary, positionedin constrained environment. This scenario has been achievedby setting one car on the last level (in open-sky conditions) ofa five stories parking garage while the second car was roamingaround the streets of Toulouse. Finally, the last scenario namedD, is a static case in urban condition around Canal du Midi.Every collaborative scenario lasted between fifteen and twentyminutes. Outside of those specific test cases, the two cars werestrictly following each other throughout the data collectionprocess.III. A NALYZING A NDROID M ULTIPATHAND C YCLE S LIP F LAG A LGORITHMSSmartphones’ embedded GNSS receiver architecture ismostly similar to COTS GNSS receivers, from capturing thesignal to estimating its position. Android allowed their usersto have access to raw data measurements outputted by thebaseband signal processing unit of the chipset receiver. Rawdata measurements range from the most basic parameters(i.e code, phase, Doppler, C/N0) to more complex ones (i.eAutomatic Gain Control (AGC), signals states and indicators).A GNSS raw measurement task force group, created bythe European GNSS Agency (GSA), wrote a white paper[4] explaining in details Android’s location data service.Among these complex measurements we find the ’MultipathIndicator’ and the ’AccumulatedDeltaRangeState’ parameters.Few information are released by Android and/or by chipsetmanufacturers concerning their computation algorithms. In atime where smartphone GNSS receiver’s technology rapidlyadvances, it became crucial to understand and evaluate thoseflags reliability in order to better characterize smartphones’positioning performances.
A. Android Flags Detection ProcessAndroid raw data measurements are obtained through theuse of the ’Android.location’ API [5]. Within this API, a publicclass called GnssMeasurement contains GNSS data supposedlycoming directly from the embedded chipset. This class isdivided into two data groups. The first one, called ’Publicmethods’, regroups all GNSS raw data measurements. The second one, named ’Constant’, gathers information about receivedsignals characteristics. Within this second group, we find a’Multipath Indicator’ and an ’AccumulatedDeltaRangeState’that provide multipath and cycle slip flags detection mechanism to the user.1) Multipath Indicator: The Android multipath indicatorstate flag can take three different values. If the flag takes thevalue of 1, a multipath interference has been detected for thatmeasurement. On the other hand, when the indicator is setto the value 2 it signifies that multipath was not detected.Moreover, the indicator can also take the value of 0, meaningthat the presence or absence of multipath is unknown.In our study, the multipath detection mechanism is simplyactivated when the indicator shows a value of 1. It has to benoted that only Honor View 20 smartphones reported signalsbeing unaffected by multipath (i.e Multipath Indicator 2).2) Accumulated Delta Range State: Android phase measurement characterization is based on the combination value ofsix state indicators. Each indicator corresponds to a constantvalue, and the overall addition of those states is promptedto the user by the ’AccumulatedDeltaRangeState’ parameter.Those states are listed below: ADR STATE CYCLE SLIP: value 4 ADR STATE HALF CYCLE REPORTED: value 16 ADR STATE HALF CYCLE RESOLVED: value 8 ADR STATE RESET: value 2 ADR STATE UNKNOWN: value 0 ADR STATE VALID: value 1Processing cycle slip flag detection is done by identifying ADR STATE CYCLE SLIP and ADR STATE RESETconstants presence in the final AccumulatedDeltaRangeState value. Thus, to detect Android cycle slip, weset values that ’AccumulatedDeltaRangeState’ could take(V alid State [1, 8, 16, 9, 17, 24, 25]). If the current ’AccumulatedDeltaRangeState’ indicator value falls out of ourselection we then flag our current measurement to be impactedby a cycle slip.Even though multipath and cycle slip detection mechanismsare provided by the Android API, no information is yet tobe found about how the chipset computation process is madeand how they are transferred to the Android GnssMeasurementclass. We will now show the flags repeatability and efficiencyin an urban environment.B. Preliminary Measurements Analysis in an Urban CanyonAndroid based positioning in urban conditions was expectedto be difficult due to possible signal degradations and the useof an inefficient smartphone antenna. Nevertheless, it was seenFig. 2. Signal Analysis for Huawei Mate 20X - Car ID n 2that each smartphone tracked more than 30 signals (considering all frequencies and all constellations simultaneously) perepoch during our entire data collection. All our tested chipsetbrands (Broadcom, Qualcomm and Kirin) achieved the sametracking performance. Although, it has been noted that bothHonor View 20s under-performed compared to others units.Due to the rapid evolution of the user-to-satellite propagationchannel, we observed fast varying C/N0 values. The top graphof Figure 2 illustrates those C/N0 fast fluctuations observedover time. For each smartphone, the minimum, median andmaximum C/N0 value has been computed in function of eachindividual received signal (considering all frequencies and allconstellations simultaneously) for every epoch. For the HuaweiMate 20X, the median value of C/N0 range between 30 and35 dBHz. The bottom graph of figure 2 shows the percentageof signals where a multipath and/or a cycle slip detection hasbeen recorded in function of time. The percentage computationwas obtained by dividing the number of flags detected by thetotal number of received signals for that specific epoch. Thefirst observation made here is that cycle slip seems to be oftendetected by the receiver whereas multipath detections remainless frequent.The mark, labeled C, on figure 2 highlights the third collaborative scenario. During this time, the second car was parkedon the last floor of a parking garage in open-sky receptioncondition. C/N0 values of all signals improved, while bothcycle slip and multipath flag detection decreased as expected.On the other hand, uncorrelated situations between C/N0 andflags detection have been observed during multiple occasions.This situation can be observed on figure 2 between epoch400 and 1000, where median signal strength remains constantduring that time period whereas flags activation numberssuddenly decrease.Similar analysis has been performed for all tested smartphones.As stated before, the Google Pixel 3 and the Xiaomi Mi9 did not record any phase measurements data. It is then
safe to state that cycle slip detection is not possible forthose devices. Moreover, no multipath flags have been raisedduring our data campaign by either phone. This evidencesuggests that the ’Multipath Indicator’ algorithm is not a naivelinear correlation of C/N0 variation but exploits the phasemeasurement to detect multipath.Independently of those phones, it appeared that both XiaomiMi 8 and the Huawei Mate 20X exhibit similar behaviors.However, both Honor View 20 (equipped with the same chipsetas the Huawei Mate 20X) generated fewer cycle slip flags.C. Correlating Flags Detection MechanismsIn order to confirm the previously stated hypotheses, adetailed analysis of multipath and cycle slip flags has beenconducted. Multiple basic GNSS measurements have beentested through a series of correlation events. While processingthe preliminary analysis of our data samples, we stated thatmultipath and cycle flag detection algorithms were not solelylinearly correlated to C/N0. To validate this hypothesis, flagsdistributions in function of C/N0 and elevation were analyzed.Figure 3 represents the cycle slip flag detection distribution infunction of C/N0. Histograms and cumulative density functions (cdf) are drawn here.Cycle slip detection distribution seems to be quite uniformlydistributed over C/N0 values. Even though our tested smartphones are not equipped with the same chipset component,they tend to have similar detection behaviors (increased detection activity between 13 and 16 dBHz, before peaking aroundthe C/N0 value of 22 dBHz). However, Honor View 20s didnot detected as many cycle slip flags (i.e section III-B) as otherdevices and their distribution are surprisingly shifted towardhigh C/N0 value (around 35dBHz).Multipath flag detection distribution in function of C/N0 havesimilar characteristics has the one observed in the case of cycleslip flag distribution. Clear peaks are located at 16 dBHz, 25dBHz and 33 dBHz. While the peak around 16dBHz mostlikely corresponds to low-elevation satellites, the two otherpeaks can not be fully explained by physical phenomenon inFig. 4. Multipath Flag Distribution in Function of Elevationthe propagation channel. Once again, the distribution behaviorsof multipath flags are similar from phone to phone (includingHonor View 20s). A simple interpolation of C/N0 does notdescribe the Android multipath detection mechanism.The distribution of multipath and cycle slip flags have alsobeen studied in function of satellite elevation. Figure 4 showsthe distribution of multipath flag detection in function ofsatellite elevation. From this graph, it is clear that no directcorrelation between multipath detection technique and satelliteelevation can be established. Moreover, smartphones tend tofollow a similar distribution trend that could indicate that themultipath detection algorithm might not be directly outputtedby the chipset itself.Overall, Android flag detection systems are not naively onlyinterpolated from C/N0 and satellite elevation value of thecurrent signals. Detection mechanisms might be as complexas the one found in modern COTS GNSS receivers. Allsmartphone brands and models shown similar distributionpatterns making us believe that those estimation algorithmsmay be using the same detection techniques at the chipsetlevel and/or that these flags might be computed at a commonlow-level Android layer.D. Android Detection Algorithm EfficiencyFig. 3. Cycle Slip Flags Detection Distribution in Function of C/N0T
Android raw data measurements are obtained through the use of the ’Android.location’ API [5]. Within this API, a public class called GnssMeasurement contains GNSS data supposedly coming directly from the embedded chipset. This class is divided into two data groups. The first one, called ’Public methods’, regroups all GNSS raw data .
With the availability of GNSS raw data from the mobile phone, there is a growing interest in analyzing the quality of raw data and its feasibility for precise positioning. The research work focused in the paper is highly dedicated in analyzing the quality of GNSS raw data and its RTK performance analysis.
Oct 15, 2017 · GNSS 5 click - board with NEO-M8N GNSS receiver module from u-blox MikroElektronika Subject: Determine your current position with GNSS 5 click. It carries the NEO-M8N GNSS receiver module from u-blox
Unlike conventional inertial navigation systems, the RT uses GNSS to correct all its measurements. GNSS makes measurements of position, velocity and (for dual antenna systems) heading. Using these measurements, the RT is able to keep other measurements, such as roll and pitch, accurate. Tight coupling of the GNSS and inertial measurements
GSM/GPRS/GNSS HAT User Manual 1 / 32 GSM/GPRS/GNSS HAT User Manual OVERVIEW This is a handy, low power Raspberry Pi HAT which features multi communication functionalities: GSM, GPRS, GNSS and Blueto
CONTENTS xi 6.3.1 Basic Principles 466 6.3.2 GNSS-ROInstruments 472 6.3.3 GNSS-ROApplications 473 6.4 Global Navigation Satellite System-Reflectrometry 476 6.4.1 BasicPrinciples: GNSS-Ras aMultistatic Radar 478 6.4.2 GNSS-RParticularities 485 6.4.3 Thermal Noise, Speckle, and CoherenceTime 489 6.4.4 GNSS-RInstrume
Android Studio IDE Android SDK tool Latest Android API Platform - Android 6.0 (Marshmallow) Latest Android API emulator system image - Android 6.0 Android Studio is multi-platform Windows, MAC, Linux Advanced GUI preview panel See what your app looks like in different devices Development environment Android Studio 9
Figure 1: Single Point Positioning using GNSS raw data from mobile phone (Nexus 9, analyzed with the MuSNAT tool in SPP mode) At the Institute of Space Technology and Space Applications, we are analyzing the feasibility of RTK positioning using the GNSS raw data from the smartphone.
K-5 ELA Missouri Learning Standards: Grade-Level Expectations Missouri Department of Elementary and Secondary Education Spring 2016 . Reading 2 1 Develop and apply skills to the reading process. Grade K Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 A With assistance, develop and demonstrate reading skills in response to read-alouds by: a. predicting what might happen next in a text based on the .
