RH INOS “Railway High Integrity Navigation Overlay System”
Project Overview and Main Achievements
Alessandro NERI
GSA H2020 RHINOS Project
Based on • international cooperation between EU and USA Objective • a positive step beyond the proliferation of GNSS platforms, mainly tailored for regional applications, in
favour of a global solution. Work programme • investigation of candidate concepts for the provision of the high integrity needed to protect the
detected position of the train, as required by the train control system application. Reference Infrastructure • GNSS (GPS and GALILEO) + SBAS (EGNOS and WAAS) + Local augmentation elements, • ARAIM techniques and other sensors on the train are the add-on specific assets for mitigating the
hazards due to the environmental effects which dominates the rail application. Ambition • Fast release of the potential benefits of the EGNSS in the fast growing train signalling market.
Railway High Integrity Navigation Overlay System
RadioLabs
Ansaldo STS
SOGEI
Nottingham University
DLR Deutsches Zentrum Fuer Luft - Und Raumfahrt EV
Univerzita Pardubice
Stanford University
ERTMS/ETCS Train Localization
GPS
RBC
BALISE
position report
ETCS
trainborne
• In ERTMS/ETCS Train location is determined by means of BALISES and
ODOMETRY
• The Balises are transponders deployed at georeferenced points
• The odometer provides the relative positioning w.r.t. the last balise
• When the Balise Reader energizes a balise, it receives a message with
the balise Id
• The on board computer (EVC) sends a POSITION REPORT to the Radio
Block Center
The Virtual Balise Concept
• The GNSS based VIRTUAL BALISE READER generates the same information produced by a Balise Reader detecting a physical Balise, through the same logical and physical interface, then emulating the Balise reader behavior with respect to the train equipment.
GPS BEIDOU
RBC
Virtual Balise
position report
ETCS
trainborne Interlocking
• In this way the On Board ERTMS/ETCS location determination
functions do not need to be changed.
Railway User Requirements
Requirement Aviation Railway
Integrity Risk IR=2x10-7 per 150 s THR= 4.8x10-6 per 1 hour
Tolerable Hazard Rate for VB detection: THRVB < 10-9/ hr (SOM)
MTBF CR=4x10-6 per 15 s MTBF=621 hours
MTBF=3x105 hours
Alert Limit/ Max. Confidence Interval
35 m (Vertical) • Along track: 3 m • Distance between parallel track
axes: min. 4.75 m in stations, typically 6 m
Availability 0.99 – 0.99999 0.9999854
5
• Translation of Railway RAMS (Reliability, Availability, Maintainability, and Safety) requirements into GNSS integrity requirements
• Comparison of aviation and railway user requirements:
VBR Accuracy Requirements
Supervised Location
Req.: The train shall not trespass the Supervised Location without specific Moving Authority
VBR Accuracy Requirements
Supervised Location
Breaking Distance
Brake Activation Location
VBR Accuracy Requirements
Supervised Location
Breaking Distance Command, Control & Signaling Latency (in [km])
Brake Activation Location
VB Detection
Limit
VBR Accuracy Requirements
Supervised Location
Breaking Distance
Brake Activation Location
Command, Control & Signaling Latency (in [km])
SIL-4 Train Location Confidence Interval
VB Detection
Limit
VB Location
Req.: To support INTEROPERABILITY Infrastructure Managers require that the same engineering rules are employed to deploy physical and virtual balises, In this way heterogeneous traffic consisting of trains equipped with physical BTM and trains equipped with Virtual BR can be handle by a a Radio Block Center, without modifications.
Railway Integrity Fault Tree ETCS core THR allocation
to LDS safety functions
10
GNSS-based services for Train Control
Functionality Current EU Technology (ERTMS)
SIS Integrity Monitoring
Augmetation Accuracy
Train Location Determination • Single track
Based on Balise X X Medium
Train Location Determination • Multiple tracks
Based on Balise, Track Circuit
X X Medium, High
Train Integrity Track Circuit + On Board Circuitry
X N High
• GNSS based train location determination can be considered a
disruptive technology.
• It will succeed in replacing the current technologies based on balises
and track circuits if and only if it will be COST-EFFECTIVE.
CHALLENGE: SIL-4 COMPLIANCE
Overlay Architecture
• Selection of candidate solutions concerning both augmentation and integrity monitoring infrastructures, and On Board Units starts from the mitigation actions related to the hazards identified during the Hazard Analysis
Hazards Mitigations
Clock runoffs SBAS & LADGNSS
Ephemeris Faults SBAS & LADGNSS
Ionospheric storms LADGNSS, Multifrequency
Signal Distortions SBAS & LADGNSS
Constellation Rotations SBAS & LADGNSS
Multipath Train Autonomous Integrity Monitoring
Jamming, Spoofing DBF + High Resilience DSP Train Autonomous Integrity Monitoring
Hazard Analysis
Mitigation Actions
Candidate Solutions
RHINOS MAIN ACHIEVEMENTS
1. Railway Environment MODELLING
2. Definition of candidate solutions
• Integrity Monitoring and Augmentation System Reference Architecture
• On Board System Reference Architecture
• Local hazards detection and effects mitigation,
Train
ARAIM
Confidence
Interval
Global Hazards
Monitoring
Local Hazards
Monitoring
GNSS SISs
RHINOS MAIN ACHIEVEMENTS
3. Cost and Benefits trade-off and selection of the reference architecture among the candidate solutions.
• GNSS based LDS Integrity Model (appropriate analytical methods and tools)
• cost and benefits,
• SWOT (Strength, Weakness, Opportunities, and Threats)
4. Realization of a Proof of Concept
5. Verification of the reference architecture performance
6. Dissemination and consensus sharing.
• RHINOS team members are part of the
• EU-U.S. Cooperation on Satellite Navigation Working Group-C on Next Generation GNSS
Track Discrimination
ENABLING TECHNOLOGIES
Scenario Satellite Visibility
Multipath Interference Risk
Train Dynamics
Railway station
Reduced High High Null
Along the line
Good Low Low High (HS)
• MULTICONSTELLATION GPS+GALILEO+ ГЛОНАСС+BEIDOU
• SBAS EGNOS, WAAS, SDCM
• Integrity Monitoring and Augmentation Services
• High Accuracy RTK, PPP (?)
• Multifrequency Rx E1 E5a E5b, L1 L2C L1C
• ARAIM
RTK: state of the art
RTK Availability and Reliability may be critical, due to their fast decay with
respect to baseline length
Source: Feng et al. “GPS RTK Performance Characteristics and Analysis”
100%
95%
100% Reliability
60%
Availability
35 km
35 km
35 km
TTFF (seconds) vs.
baseline length
Dual Frequency
When the railway consists of multiple tracks, the single track PVT
estimate is combined with the track detection.
We assume that the train can be located along one of M tracks and we
denote with Hk the hypothesis corresponding to the k-th track.
The Bayesian (optimal) track detection rule selects the hypothesis
corresponding the largest generalized likelihood ratio
Track Discrimination
• For each track we estimate the train mileage under the hypothesis that Hk is true;
• We then use these mileage estimates in a likelihood ratio test, as if they were correct
Measured
double
difference
RTK with Track Constraint
q
p
RS
Baseline expansion in Taylor’s w.r.t. s
Then we have
( 1)
( 1 ( )
ˆ
)ˆm
m
s s
m ss
bb
b;
( 1) ( )ˆ m ms s s
( 1)ˆ mb
b
Initial phase
Ambiguity
( 1) ( 1)ˆ m m tr o
ii
op i n
ic c Hb DD β τ τL
( 1) ( )ˆ , 1,2ii
mm s i N nGH
( )
( )
ˆ m
Trainm
s ss
XG
( )j T
MSH S e
1 1( )0
0Sat Sat
j jj
N j N j
I 1S
1 I
( 1)
( )
ˆ m
m
s s
ss
b
( 1)ˆ ms ( 1)ˆ ms
s
RTK with Track constraint
( 1)
1 1
( 1)
2 2
( 1)1
( 1)
( 1)
( 1)
( 1
1
(
1
1)2 2
)
2
(
)
ˆ
ˆ
0 0
0 0
0
0
ˆ
ˆ
m
m
m
m P
m P
m
mm
ms
HbP DD β
P Hb DD β
L DD β
HG
HG
HG I
HG
Hb
L DDN
βHb
N
I
1
2
1 1
2 2
P
P
n
n
n
n
Generalized Likelihood Ratio
Dual frequency linearized system
/ ( / )( )
( )
kH k
k
p H
w
DD RR
R
1 2
1 2 1 2
,/ ,
( / , ) ( / )
(
, ,
)
kk kH
p H P H
w
N N
DD NR N N N
R
N
R
Since we have undesired unknown parameters (the phase ambiguities) we have
RTK with Track constraint
( 1)
( 1)
( 1)
( )
1
1 1
( 1)
2 2
( 1)1 1
( 1)2 2
( 1)
) 2( 1
ˆ
ˆ
0 0
0 0
0
0ˆ
ˆ
ˆ
m
m P
m P
m
m
m
m
m
m
s
HbP DD β
P Hb DD β
L DD β
HG
HG
HG I
HG
Hb
L DDN
βHb
N
I
1
2
1 1
2 2
P
P
n
n
n
n
To reduce the search space we may employ the Wide Lane (WL) and
Narrow Lane (NL) combinations
Dual frequency linearized system
Combination Weakness
Wide lane Too noisy
Narrow lane
Slow Ambiguity fixing
WLL
WLP
NLP
NLL
Track Constrained RTK
Measurement equations
( 1)
( 1
( 1) ( 1)
( 1) (
(
1
)
))
ˆˆ
ˆ
m
WL
m
NL
mW
m P m P
NL NL
m m
WL WL L WL WL
s
N
DD β HG 0 n
DD β HG I n
P Hb
L Hb
One advantage of the use of this pair is that an initial estimate of the phase ambiguities can be obtained by the Melbourne-Wübbena combination
MW WL NL B L P
Here the Melbourne-Wübbena combination is used to reduce the number of candidate ambiguities to be considered in the Generalized Likelihood Ratio computation.
A Posteriori Track Probability
For each candidate phase ambiguity vector the estimated train mileage is
( )
, /
ˆˆ
ˆˆ k k k
k hWL
kk k
WL
PH H H NLNLm
H HWL WL H WLH H
s
N
H b DD βP 0K
L I DD βHN
b
1
( ) ( 1) 1 1 ( 1) ( 1) 1 1
k NL WL k k NL WL
T Tm m T m m T
H P L H H P L
K G H R R H G G H R R
, 1
, 1
2
ˆ ,
2
ˆ ,
1ˆexp ( )
2Prob .
1ˆexp ( )
2
WLHk WL
WL WL
WLHk WL
WL WL
WLs
k
WLsm
P
H
P
N
ν
N
ν
NN R
NN R
ν N
ν N
The A posteriori Probability of each tack is
,
ˆ , , ,ˆ ˆ ˆ( ) ( )
k kWL WLHkW
WLL
P
NLs H H L
WL
Ws s
N
N NN
0Iν R Hb DD β
IIN
Test bed description
Parallel tracks: 2
interaxis: 3 m
Train on Track #1
Tracking
Channels
120 channels
GPS: L2, L2P, L2C, L5
GLONASS: L1 C/A, L2P, L2C
Galileo: E1, E5a, E5b, E5a+b
SBAS: WAAS, EGNOS,
GAGAN, MSAS
Measurements
Quality
Very low noise GNSS carrier
phase measurements (RMS<
0.2 mm)
Fixed Amb. RTK
positioning
accuracy
10 mm + 1 ppm (horizontal)/10
mm + 1ppm (vertical)
Antenna Standard Dorne Margoline with
Choke Ring Antenna
Measurements
update rate
50 Hz
A posteriori track probability (Results)
Memoryless Single Epoch Detection
Mileage
Track Posterior Probability
A posteriori track probability (Results)
Memoryless Single Epoch Detection
Geometry Free
Residual MSQE
Track Error Probability
• Single Epoch, M equispaced Tracks
• N0 epochs, M equispaced Tracks (Slow motion)
• DGNSS - N0 epochs, M equispaced , Rank Order Statistics Detector
( , ) 11
2 2
O hN I
e OP erfc N bM
Γ e
0
( , )1 1O O
O
NhON II N h
e e e
h k
NP P P
h
ˆb e
Track Discrimination Performance
( )
h h h k
j
h H H H H νΓ C I H G K S E
11
2 2
h
eP erfc dM
Γ e
Conclusions
• As confirmed by the experimental activity, the coarse estimate of the train location provided by the Wide Lane combination is good enough to reliably discriminate the track.
• The main advantage is the TRADE-OFF between ACCURACY and TIME
needed to discriminate the track. In fact, in our case, we do not have to wait for ambiguity fixing.
• To achieve track error probabilities compatible with SIL 4 operational requirements even in strong multipath environment, TEMPORAL INTEGRATION and MULTIPLE CONSTELLATIONS can be applied.
• Nevertheless, effectiveness of temporal integration can be impaired by MULTIPATH errors highly correlated in time.
• To reduce this effect, proceeding at the maximum authorized speed when in Start of Mission mode is recommended.
THANKS FOR YOUR ATTENTION
SIS Fault
• The Hazard essentially reduces to the ephemeris errors and Tropospheric errors
• Effect modeled as a satellite position error b on the i-th satellite
• Let
Single difference error
SIS Fault
• Let
cos sinp q
b
B b B B
Bp
Bq b
12 2 2
, ,q, ,
( ) 2i iRIM RIMp q
i i
m p mi
SD i i
Train Train
b
B r r
e eB Bb b
b
Project Objectives
• Objective 1: To DEFINE THE ARCHITECTURE of a train Location Detection System (LDS) and of the supporting infrastructure, with the following properties
• joint use of GPS and GALILEO and wide area integration monitoring and
augmentation networks (WAAS, EGNOS)
• standard interface for providing Safety of Life services for railways through SBASs, regional augmentations or hybrid SBAS/GBAS systems;
• compliance with European and US railway requirements and regulations;
• sharing as much as possible of the supporting infrastructure and on board processing, including new developments such as ARAIM, with the avionics (and automotive) field,
• provisioning of a set of functionalities tailored to the specific needs of the rail sector.
Project Objectives
• Objective 2: To assess the performance of the defined architecture by means of:
• a PROOF-OF-CONCEPT integrating, in a virtualized testbed,
• real railway environment data sets,
• rare GPS SIS faults
• simulated faults for the new constellations;
• ANALYTICAL METHODS for verification and safety evidence of defined architecture according to railway safety standards (e.g. CENELEC EN 50129, etc.)
3.975 3.98 3.985 3.99 3.995 4 4.005
x 105
0
5
10
15
20
25
30
35
40
45
50Mean values on double differences for all the satellites in view
GPS time [s]
Mean o
n d
ouble
diffe
rences [
m]
Project Objectives
• Objective 3: To contribute to the missing standard in the railway sector about the way of integration of GNSS-based LDS, into current Train Control System standards (e.g. ERTMS)
• by publishing a comprehensive GUIDE on how to employ, in a cost-
effective manner, GNSS, SBAS and other local infrastructures in safety related rail applications worldwide,
• by defining a STRATEGIC ROADMAP for the adoption of an international standard based on the same guide.