Post on 29-Jan-2016
transcript
VRS Network The Magic Behind the Scene
Xiaoming Chen
Trimble Terrasat GmbH
Outline GNSS Positioning Error Sources General Introduction of Network RTK
– VRS– RTCM 3 Network Message
Use Network Correction Quality To Improve Rover Performance
Sparse Glonass Network Large Network Data Processing Summary
GNSS Positioning Error Sources
Orbir Error Satellite Clock Error
Îonosphere
Troposphere
Receiver Clock Error Multipath
GNSS Positioning Error Sources
Reference stations
Ionosphere
GNSS Network Models
Network RTK Utilize a reference station network to model
distance dependent errors in real-time Generate network corrected reference station
data/corrections and transmit to rover in real-time – VRS– FKP– RTCM Network Message
Rover use the network corrected data to achieve better performance over longer distance
Geometric Filter
Geometric Filter
Geometric Filter
Raw DataAnalysis
Synchronizer
Geometric Filter
Ionospheric Filters
Code-Carrier Filters
Ambiguity Search & Fix
Residual ManagementNetwork Model
IntegrityVRS/Net RTCM/FKP
Generation
Raw DataAnalysis
Raw DataAnalysis
Raw DataAnalysis
Geometric Filter
Network Processing Diagram
Virtual Reference Station (VRS)
Computes tropospheric, orbit and ionospheric models in real time.
Derives an optimized VRS correction stream derived from these models for each rover
Requires bi-directional communication, also works with rebroadcast/RTCM VRS module
Based on RTCM, CMR, CMRx. Low bandwidth required
Support GLONASS
RTCM Network Message RTCM 3.1 standard Broadcast solution Derive carrier ambiguities in network and generate
observations on one ambiguity level (no ambiguities in the Double Difference sense)
Master & Auxiliary station One master station Up to 31 auxiliary stations (ambiguity “free” observations) High bandwidth or lower rate for corrections
Network corrections are computed on the rover from a subset of the network
GPS only
Modeling Error SourcesServer Centric vs. Rover Centric
VRS = Server Centric Approach: Complex error models are used:– Ionospheric model– Tropospheric model
RTCM Network Message = Rover Centric Approach: – Interpolation in the rover
Modeling Error Sources: An Example for tropospheric modeling
AGNES Network, Switzerland on July 7, 2003, operated by Swisstopo with Trimble VRS Jungfraujoch as rover Nearest ref. station: Hohtenn
Station Height [m]
Jungfraujoch
3634
Hohtenn 985Sannen 1419Zimmerwald
956
Huttil 779Luzern 542Andermatt 2367Jungfraujoch
Modeling errors: VRS vs. RTCM Network Message
-0.14
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.3 0.5 0.7 0.9 1.1 1.3
GPS Hour
Ion
o. F
ree
re
sid
ua
ls [m
]
10
15
20
25
30
35
Ele
vatio
n [d
eg
ree
]
VRS
NetRTCM
Elevation
Iono-free Residuals for SV 05
Benefits with VRS
NetRTCM [mm]
VRS [mm]
Improv[%]
Mean[mm]
North -4.66 -3.30
East -4.29 -5.11
Height
-117.00 -41.34
Standard
Deviation
[mm]
Height
46.12 39.19 15.1
2D 30.83 26.67 13.9
3D 55.47 47.41 14.5
RMS[mm]
Height
125.76 56.96 54.7
2D 31.47 27.36 13.1
3D 129.64 63.19 51.3
Use Correction Quality to Improve Rover Performance
Network RTK correction considered as interpolated corrections between reference stations
Interpolation is not perfect depending on actual atmosphere conditions
RTK Network server process provides quality estimates for residual interpolation
Can be used by the RTK rover to optimize RTK performance
Sparse GLONASS networks with reduced GLONASS correction quality
Residual Error Description
RTK Network generates a description of the dispersive and non-dispersive error for each satellite
Consists of constant, distance and height dependent terms
2222 didici
220
220
20
20 hd hdc
Predicted Network Correction Quality (strong ionosphere)
Predicted Network Correction Quality (calm ionosphere)
Network used for Evaluation of Quality Information
24 h data (1Hz) 5 Stations 1 Rover (33 km
from 0272)
Ionospheric Residuals PRN 22
0
20
40
60
80
100
120
0 30 60 90 120 150 180 210 240 270 300
Time in Minutes
Ion
osp
her
ic e
ffec
t [m
m]
0
10
20
30
40
50
60
70
Ele
vati
on
[°]
Absolute residuals
Predicted sigmas
Elevation
55% of the DD residuals < predicted sigmas
Geometric Residuals PRN 22
47% of the DD residuals < predicted sigmas
0
5
10
15
20
25
30
35
40
45
50
0 30 60 90 120 150 180 210 240 270 300
Time in Minutes
Geo
met
ric
effe
ct [
mm
]
0
10
20
30
40
50
60
70
Ele
vati
on
[°]
Absolute residuals
Predicted sigmas
Elevation
0
20
40
60
80
100
120
140
0 30 60 90 120 150 180 210 240 270
Time in Minutes
Ion
osp
her
ic e
ffec
t [m
m]
0
10
20
30
40
50
60
70
Ele
vati
on
[°]
Absolute residuals
Predicted sigmas
Elevation
Ionospheric Residuals PRN 1
62% of the DD residuals < predicted sigmas
Ionospheric Residuals PRN 31
56% of the DD residuals < predicted sigmas
0
10
20
30
40
50
60
70
80
0 30 60 90 120 150 180 210 240 270
Time in Minutes
Ion
osp
her
ic e
ffec
t [m
m]
0
10
20
30
40
50
60
Ele
vati
on
[°]
Absolute residuals
Predicted sigmas
Elevation
Improving Rover Performance With Network Correction Quality
Predicted error statistics can help to improve positioning by– Better measurement weighting– Optimum combination of L1/L2 measurements
Helps to improve – Positioning accuracy – Ambiguity fixing
Positioning Error Comparison - East Error
Positioning Error Comparison – Height Error
Positioning Performance
average 3D-RMS (½ hour slots)
Sparse GLONASS Network
Increasing number of RTK network service providers introduce GLONASS only on selected stations
RTK Servers have to handle sparse GLONASS coverage in dense GPS networks
Provide high quality GPS correction and acceptable GLONASS correction
RTK Rover performance is better or equal to GPS only solution
GPS Network
GPS Only
GPS/GLN
GPS/Glonass Network
GPS Only
GPS/GLN
Partial GPS/GLONASS Network
GPS Only
GPS/GLN
Sparse Glonass Network
GPS Only
GPS/GLN
A Dense GPS/GLONASS Test Network
GPS&GLONASS
GPS only
A Sparse GPS/GLONASS Test Network
GPS&GLONASS
GPS only
Sparse GLONASS Test Results
Rover Initialization
Rover Positioning
Network Type 68%[sec] 95%[sec] No. Init.
GPS Only 14 18 2643
Dense GPS/GLN 12 15 2645
Sparse GPS/GLN 13 16 2644
Network Type RMS North
[mm]
RMS East
[mm]
RMS Height
[mm]
GPS Only 12 7 25
Dense GPS/GLN 12 6 23
Sparse GPS.GLN 12 6 23
Large GNSS Network Data Processing
Increasing Complexity and Demand… More Stations
– Tendency to increase networks to more than 100 stations – Challenge to process all data on one server in real-time (1Hz)
More Satellites– GPS– GLONASS– GALILEO
More Signals– L5– E5A, E5B
VRSNow Germany (145)
VRSNow Germany (Subnetwork)
VRSNow Germany (145)
Raw DataAnalysis
Synchronizer
Ionospheric Filters
Code-Carrier Filters
Ambiguity Search & Fix
Residual ManagementNetwork Model
IntegrityVRS/Net RTCM/FKP
Generation
Raw DataAnalysis
Raw DataAnalysis
Raw DataAnalysis
Geometric Filter
Geometric Filter
Geometric Filter
Geometric Filter
Geometric Filter
Network Processing Diagram
Centralized Geometry FilterProvide iono.-free ambiguity for network
ambiguity fixingProvide ZTD estimationAll states estimated in a big (centralized) filterTypical setup
ZTD per stationReceiver clock error per stationSatellite clock error per satelliteAmbiguity per station per satelliteOrbit error
Centralized Geometry filterNumber of States
1215
18
20
40
80
120
0
500
1000
1500
2000
2500
Centralized Geometry FilterNumber of multiplications
0
10
20
30
40
50
60
70
80
90
100
10 20 30 40 50 60 70 80 90 100 110 120
Number of stations
Num
of M
ultip
licat
ion
in B
illion
No. Multiplications
Cubic function
Principle of Federated Filter
A bank of local Kalman filters runs parallel. A central fusion processor computes an
optimal weighted least-square estimate of the common system states and their covariance
Then the result of the central fusion processor is fed back to each local filter
Parallel Computing
Simultaneuous use of multiple compute resources to solve a computational problem
Computation Time Comparison(4 Core Dell Precision 490)
0
10000
20000
30000
40000
50000
60000
0 1 2 3 4 5 6
Hour
CPU
Tim
e (m
sec)
Opt, No OMP
Fed, No OMP
Fed, OMP
Computation Time Comparison(4 Core Dell Precision 490)
CPU Time OMP vs No OMP
0
100
200
300
400
500
600
700
800
900
1000
0 50 100 150 200 250 300 350 400
CPU Time OMP (msec)
CPU
Tim
e No
OM
P (m
sec)
CPU Load (VRSNow Germany)
Summary
Quality measures for RTK network corrections significantly improve the rover performance Positioning improved by up to a factor of 2 Initialization time reduced by 30%
Sparse GLONASS network provides decent rover performance during low to medium iono activity
Large network processing provide seamless and homogeneous solution cross the whole network with balanced CPU loadReduce the complexity of network administration