PHD in Environmental and Infrastructure Engineering XXIX cycle Low cost GNSS receivers: navigation and monitoring activities final dissertation Marco Negretti marco.negretti@polimi.it Supervisor: Ludovico Biagi Tutor: Giovanna Venuti 15-03-2017
Low cost GNSS receivers: navigation and monitoring activities Performances assessment of Low Cost GNSS receivers and positioning improvement Navigation: identify paths, POI correction using data from Continuously Operating Reference Stations (CORS) correction using cartography as reference Local Monitoring (landslide monitoring): verify if it is possible to do local monitoring with low cost receivers using geodetic techniques 2
Navigation Navigation Case study: MEP (Map for Easy Path) project Winner Polisocial Award 2014, the academic social responsibility programme launched by the Politecnico di Milano Get around the city for physically impaired people can be difficult different kinds of obstacles and dangers 3
Navigation Aim of the MEP projetc create a tool to map obstacles identify accessible paths feedback to identify problems and define methods to improve mobility data should be easily upgradeable continuously Three departments were involved DICA, DEIB, DESIGN DICA group: manage GNSS data 4
Navigation The mapping activity is expected to be carried out by volunteers mappers, with mass market devices (smartphones and tablets) Improve positioning accuracy to warrant the identification the right side of the road if possible, otherwise only the road Example: obstacle on sidewalk it is important identify the side of the street actual coordinates collected coordinates 5
Navigation: correction using data from CORS Correction using data from CORS Individualized corrections were compute epoch by epoch from CORS data Apply corrections to rover trajectory Typically a Rover observes only a subset of satellites viewed by CORS a software was developed to remove satellites not observed by rover from the CORS RINEX CORS CORS estimated coords epoch by epoch rover Rover estimated coords epoch by epoch 6
Navigation: correction using data from CORS First test: static positioning Test in Milan and Como one week for both sites Data elaboration to estimate CORS positions gogps, rtklib 7
Navigation: correction using data from CORS Example with Como Data day 15-130, north CORS, rover, rover corrected 8
CORS data: point positioning Como u-blox [m] u-blox corr. [m] (original RINEX) East North 0,07-0,40 u-blox corr. [m] (modified RINEX) East North 0,07-0,39 East 0,07 North 0,58 mean std 0,85 1,22 0,65 0,82 0,65 0,81 max std 0,91 1.39 0,71 0.88 0,69 0,86 mean Milan u-blox [m] u-blox corr. [m] (original RINEX) East North -0,65 0,19 u-blox corr. [m] (modified RINEX) East North -0,50 0,21 East -0,65 North 1,38 mean std 1,50 2,32 1,42 2,02 1,42 2,00 max std 1,63 2,71 1,48 2,15 1,47 2,14 mean 9
CORS data: point positioning Como mean mean std u-blox corr. [m] (original RINEX) East North 0,07-0,40 u-blox corr. [m] (modified RINEX) East North 0,07-0,39 East 0,07 North 0,58 0,85 1,22 0,65 0,82 0,65 0,81 0,91 1.39 0,71 0.88 0,69 0,86 Generally CORS corrections work max std u-blox [m] Few improvements using original RINEX compared modified RINEX Milan mean u-blox [m] rover in a East -0,65 u-blox corr. [m] (original RINEX) good position North East North 1,38-0,65 0,19 u-blox corr. [m] (modified RINEX) East North -0,50 0,21 mean std 1,50 2,32 1,42 2,02 1,42 2,00 max std 1,63 2,71 1,48 2,15 1,47 2,14 10
Navigation: correction using data from CORS Second test: kinematic survey Two test paths were defined to verify different conditions of satellite visibility urban canyons areas of intermediate visibility open areas Path 1 Path 2 OpenStreetMap contributors OpenStreetMap contributors 11
Navigation: correction using data from CORS Tablet, smartphone and u-blox as rovers A geodetic receiver as benchmark to evaluate the rovers survey data processed with double different method data as reference only if it is obtained a fixed phase solution: a good accuracy reached Path 1 16-139 N of points fixed/total 1886/2144 fixed solutions 88% 36% 16-207 1841/2017 91% 33% 16-208 1596/1811 88% 14-259 N of points fixed/total 1805/4038 fixed solutions 45% 14-267 1313/3634 14-268 1198/3547 Path 2 12
Navigation: correction using data from CORS Statistical errors before and after the correction: the accuracy after the CORS correction does not collected corrected improve Path 1 tablet Day [m] East North mod RINEX East North 14-259 Path2 [m] Day 16-139 mean std 16-207 mean std 16-208 mean std mean 0.26-0.22 0.07 std 2.44 2.52 2.46 14-267 mean -0.43-3.14 0.89 std 3.68 5.27 3.40 14-268 mean 1.05-0.27-4.12 std 3.93 5.00 4.00 smartphone u-blox collected corrected collected corrected mod RINEX mod RINEX East -0.83 1.97-1.36 3.95 0.93 4.88 North 1.84 2.38-2.30 4.36-0.88 3.98 East -0.89 1.97-1.28 4.03-0.89 4.87 North 2.07 2.41-2.27 4.54-0.83 4.00 East -0.64 0.86-0.11 1.42-0.21 1.19 North 0.48 1.31 0.57 2.44-0.20 1.11 East -0.71 1.11-0.11 1.37-0.25 1.23 0.71 2.49-3.41 5.22-2.24 5.14 North 0.64 1.36 0.33 2.77-0.13 1.16 13
Navigation: correction using data from CORS Local disturbances particular properties of the area local effects are prevalent also in open areas reference points corrected points collected points 14
Navigation: correction using cartography as reference Correction using cartography as reference Reference map: OpenStreetMap (OSM) Database: PostgreSQL/PostGIS points detected with mass market devices paths and buildings from the OSM maps in the neighborhood of the points detected script to implement the data correction procedures Different correction procedures were implemented 15
Navigation: correction using cartography - on segments Correction on segments Chose the segment OpenStreetMap contributors distance from the point collected to the candidate segment the bearing from the previous to the current point the bearing of the candidate segment the distance from the current corrected point to the previous corrected point Drawback: could correct too much 16
Navigation: correction using cartography - out of buildings Correction out of the buildings Customer request Avoid overlap between points collected and buildings For each point query to check if it is inside a building in case, the point is projected outside 17
Navigation: correction using cartography - out of buildings Example of correction effective 18
Navigation: correction using cartography - out of buildings Example of correction ineffective 19
Navigation: correction using cartography - on buffers Correction on buffers 1.Identification of a potential path (a) choice of a set of segments candidates to be part of the potential path 20
Navigation: correction using cartography - on buffers 1.Identification of a potential path (b) identification of the potential path: segments selected 21
Navigation: correction using cartography - on buffers 3. Around of the selected potential path is built the buffer buffer size fixed buffer size variable 4. The points outside the buffer are corrected 22
Navigation: correction using cartography - on buffers Buffer size fixed: 7 m urban two-lane road with the sidewalks 23
Navigation: correction using cartography - on buffers Buffer size variable dseg= distance between segment and the nearest building buffer size: 2 < dseg < 7 [m] 24
Navigation: correction using cartography - on buffers In case of very bad quality of signal a wrong segment can receive an high score and be selected as segment path 25
Navigation: correction using cartography - on buffers Reclassification: reduces the score for handling segments one vertex not shared with others candidate segments suspected to be erroneously selected (A) cover the holes: adds the segments not selected both the vertexes shared with others candidate segments suspected to be erroneously discarded (B) segments selected segments not selected 26
Navigation: correction using cartography - on buffers 27
Navigation: correction using cartography - results Comparing results Two tolerance areas were defined around the true path: 2.5 m and 5 m it has been checked how many points were inside them Total points: 5879 Proximity areas 5m 2.5 m collected out of the data buildings on buffers constant variable correction on variable segment with reclass. points inside 4030 4384 4641 4876 4928 5318 % 68.5 74.6 78.9 82.9 83.8 90.5 points inside 2614 2960 2944 3457 3496 2896 % 44.5 50.3 50.1 58.8 59.5 49.3 Day 14-259 28
Navigation: correction using cartography - results In general the choice of correction method is not easy The final results are very dependent on the site characteristic and from the cartography used as reference an incomplete or wrong reference datum can lead to consider a point position wrong when the problem can be the map itself it is important to have cartography updated and easily upgradeable, like OSM, that permits to add a new path as it becomes available 29
local monitoring Local Monitoring Research to assess precision and accuracy of low-cost GNSS receivers in monitoring displacements with low latency Controlled movements imposed to a low cost rover, estimated displacements compared with the imposed ones accuracy of low cost GNSS receivers how reliable are the results obtained or rather, how many false or missed alarms are expected per unit of time 30
local monitoring Two different analysis have been performed: Significance analysis based on Fisher test this procedure allows a separation of the results into two groups: coordinates significantly changed with consecutive sessions and coordinates unchanged compared to the actual values Congruence analysis the aim is to verify whether the estimated coordinates differences are congruent with the imposed displacements 31
local monitoring: pre-test Pre-test: points of survey Reference station Valleggio 1: COMO CORS (EUREF network) receiver: Topcon Odyssey RS antenna: TPSCR3 GGD Rover: ublox NEO 7P (evaluation kit) about 100 m from the reference station 32
local monitoring: pre-test Pre-test: points of survey 33
local monitoring: pre-test Simulation slide test two days of static acquisition to compute the reference start position the antenna was shifted only horizontally 5 mm for each session three session per day, three hour per session total displacement of 10 cm was aimed but, due to an instrumental problem, the test stopped at 5.5 cm Data processed with LGO and rtklib 34
local monitoring: pre-test Data processed with LGO and rtklib LGO rtklib Estimated displacement Imposed displacement 35
local monitoring: pre-test Congruence analysis: class 0 class 1 class 2 class 3 number of couples LGO rtklib 43.9 % 46.5 % 48.8 % 49.0 % 6.4 % 4.5 % 0.9 % 0% 36
local monitoring: test Test: points of survey: Reference station Valleggio 1: Reference station Valleggio 2: COMO CORS (EUREF network) receiver: Topcon Odyssey RS antenna: TPSCR3 GGD receiver: ublox NEO 7P (evaluation kit) antenna Tallysman TW31525* about 10 m from the reference station Rover: receiver: ublox NEO 7P (evaluation kit) about 130 m from the reference station * Many thanks to Geomatics Research & Development (GReD) 37
local monitoring: test Test: points of survey 38
local monitoring: test Test: points of survey N 45 Rover is installed on a device* (slide) that allows to apply controlled displacements E * Many thanks to the colleagues of the Politecnico di Torino - DIATI 39
local monitoring: test Initial position of the receiver was determined by a 24 hour session with a geodetic receiver and estimated respect COMO CORS The rover antenna was moved, alternately in horizontal and down in vertical every 2 hours 5 mm in horizontal (0 10 cm) 5 mm in vertical (0 10 cm) The data were processed with LGO software two times: once using Valleggio 1 as reference station, the other one using Valleggio 2 40
local monitoring: test Horizontal displacement estimated imposed 41
local monitoring: test Vertical displacements estimated imposed 42
local monitoring: test Statistics of the errors [cm] mean std E N vert. horiz. 0.1-0.2 0.3 0.0 0.4 0.6 1.3 0.4 Reference station: Valleggio 1 [cm] mean std E N vert. horiz. 0.0-0.2-0.6-0.1 0.7 0.9 1.8 0.7 Reference station: Valleggio 2 43
local monitoring: test Significance analysis 78 % 85 % 80 % Movements between couples 0 1 20 % 14 % 28 % 2 29 % 17 % 3 63 % 4 5 Movements between couples 0 2D UP 3D 2D UP 3D 88 % 84 % 92 % 1 14 % 14 % 14 % 46 % 2 23 % 18 % 21 % 28 % 68 % 3 31 % 27 % 31 % 88 % 31 % 90 % 4 64 % 33 % 61 % 99 % 38 % 100 % 5 84 % 36 % 88 % Reference station: Valleggio 1 Reference station: Valleggio 2 44
local monitoring: test Congruence analysis East North Vertical Horizontal class 0 16 % 14 % 2% 18 % class 1 29 % 28 % 4% 32 % class 2 26 % 24 % 5% 26 % over 29 % 34 % 89 % 24 % Reference station: Valleggio 1 East North Vertical Horizzontal class 0 18 % 15 % 2% 20 % class 1 30 % 24 % 4% 32 % class 2 24 % 21 % 4% 24 % over 27 % 39 % 89 % 22 % Reference station: Valleggio 2 45
local monitoring Generally the estimated displacements follow the trend of the imposed displacement Low cost receiver can be used, under specific constraints, in monitoring activities vertical component is not well recognized main displacement in the monitored area should be in the horizontal component It is possible to use a low-cost receiver also as a reference station enhanced antenna 46
Future work Future work Monitoring test with better antennas, but always low cost Navigation trajectory estimation using neural networks 47
Reference MEP - http://mep5x1000.wix.com/mepapp PostgreSQL - http://www.postgresql.org PostGIS - http://postgis.net gogps - http://www.gogps-project.org/ RTKLIB - http://www.rtklib.com/ Cartography OpenStreetMap - http://www.openstreetmap.org Orthophoto AGEA 2012 from WMS service of Regione Lombardia: http://www.cartografia.regione.lombardia.it/ http://creativecommons.org/licenses/by-nc-nd/4.0/ 48