Rome Navigation Innovations 5/14/07 1 GIS and GPS for Railroad Environment Presented by: Jim Rome May 15, 2007 Rome Navigation Innovations Inc 27 Old County.

Rome Navigation Innovations 5/14/07

1

GIS and GPS for Railroad Environment

Presented by:

Jim Rome

May 15, 2007

Rome Navigation Innovations Inc27 Old County Rd.

Gloucester, MA [email protected]

H. James Rome, President978-281-5623

mailto:[email protected]





2

Why is Track Navigation different from most any other Navigation

• It is mostly ONE DIMENSIONAL, Linear referencing!

• Curves are mild , linear accelerations are mild.

• A Map (if we have one) tells us everything except where along the track we are. We simply have to pin down that ONE quantity.


3

Its Sooo.. Easy!

• Compare the Issue an aircraft pilot in flying and landing virtually blind.

• Track Navigation is like shooting fish in a Barrel

• Don’t Shoot yourself in the Foot!


4

What Follows

• Preliminaries

• Principles of GPS ( very very short course)

• Issues of Rail Navigation

• Data Integration and Handling– For Navigation– Navigation with Measured Data

• The Railroad Receiver


5

Definitions

• Purpose of GIS and Navigation: Locate something so you can find it again

• Accuracy Required : Good enough to complete your “Mission”

• Methods of Navigation on the Rails :– Dark Ages.. Locate by visual clues– Until Recently Locate by ft from Milepost

(Requirements are often driven by capability)– NOW GPS…Adds a whole new Dimension of the One

Dimensional Problem


6

Comparing Apples to Prunes

• Never Try to Land a plane where a train navigates• A Plane for the most part can go any direction at

any time and hardly ever stops. A Train for the most part goes forward and backward and stops all the time.

• GPS on a plane is generally available. On a train, only if we are lucky.

• On a train we Independently can find distance along the ground traveled to great precision. Not so on a plane.


7

Accuracy Integrity Reliability

• Accuracy: RMS error 95 % limit of error, etc.

• Integrity: know when to trust! Know when readings are false! ( Can cost lots of money when you don’t)

• Reliability what fraction of the time ( space) is it working?


8

Why Know the Principles of Operations of GPS

• So you know where it won’t work and where it might Shoot you In the foot!

• Know what complements GPS.

• Know what you might do with GPS.. And What you can’t . That is, understand issues.

• Next is a very brief summary.


9

sat. time

Observed Time

sat. time

Observed Time

Observed Time Observed

Time

sat. time

sat. time

Antenna /Receiver Compute Position

Sat Positions

Other Parameters

Vehicle Position

Principles of GPS Operation

DifferentialCorrections R1

R2

R3 R4


10

NAVIGATION SOLUTION IF WE HAD 4 TIME MEASUREMENTS T1, T2,T3,T4

AND FOUR UNKNOWNS: X,Y,Z, Can Find position, X,Y,Z RANGE TO VEHICLE

T1= [(Xs1-X)2+(Ys1-Y)2+(ZS1-Z)2] 1/2 /c + n1

T2= [(Xs2-X)2+(Ys2-Y)2+(ZS2-Z)2] 1/2 /c + n2

T3= [(Xs3-X)2+(Ys3-Y)2+(ZS3-Z)2] 1/2 /c + n3

T4= [(Xs4-X)2+(Ys4-Y)2+(ZS4-Z)2] 1/2 /c + n4

Speed of Light

Synch Error

Noise

X orbit Pos. sat #2

Transmission Delay Time


11

ADDITIONAL FEATURES IN NAVIGATION SOLUTION

THERE ARE USUALLY MORE MEASUREMENTS: T5= [(Xs5-X)2+(Ys5-Y)2+(ZS5-Z)2] 1/2 /c + n5

T6= [(Xs6-X)2+(Ys6-Y)2+(ZS6-Z)2] 1/2 /c + n6

USED FOR SELF CHECK AND IMPROVING ACCURACY• Least Squares Estimation•Integrity and Exclusion

1,2,3,4

1,2,3,5

1,2,4,5

1,3,4,5

2,3,4,5

1,3,4,6

Least Squares Estimate: HDOP*RMS Range=Nav. Error

Integrity Region

Outlier, Sat6 Bad?

One SolutionUsing 4 Sats.


12

ADDITIONAL FEATURES IN NAVIGATION SOLUTION

T3= [(Xs3-Csx3 -X)2+(Ys3 -Csy3 -Y)2+(ZS3 -Csz3 -Z)2] 1/2 /c – EP3+ n3

Differential Corrections:

Orbit Corrections Propagation Corrections

Types of Differential Corrections:VBS satellite (OmniStar) ~3 ft (they say 1ft) (800/yr)NDGPS (Free.. If available) 3-6 ftWASS (Satellite, Free) 10 ft.(Omnistar) 1500 HP Satellite 5 INCHES! ( 1500/yr)

But They only work when there is GPS!


13

Types of Receivers• Hand Held ( WASS enabled) accuracy ~ 30 ft (10 ft)

<$300• Computer Cards .. Right in your laptop Accuracy 30 ft-~

2 ft .. With cycle matching and Good differential• Receivers with cycle matching… 2 ft with good

differential… better for “local corrections”• Dual Frequency ( gets rid of Ionospheric errors) Receivers

have cycle matching, with base station positioning to cm or mm. ( Overkill?)

• Inertial Aiding… probably won’t benefit railroad users, but fast recovery from outage will.


14

Track Beds Are Not Designed for GPS Visibility -Issues

HeadingOdometer

Blockage of Sats./DifferentialMulti-pathTotal WipeoutError in Differential

Dead Reckoning is Important

Issues of GPS and Train Navigaton


15

So What does One Dimension Buy US?

• Linear Referenced track Map .. It will fit on a memory stick

• If only we had a perfect Odometer, we would only need GPS occasionally

• If only we didn’t have to use Figure skates to play Hockey….?– Example Next that could probably reduce the

unavailability of GPS by a factor of 3


16Surface of Constant Time Difference

Line of Track

Solution: Two Satellites and a Track Map

This is a great Concept,Easily Implemented.But Don’t expect a GPSManufacturer to offer oneSoon


17

Why is this so special?

• If any satellites in view, they will be fore and aft, and above .. Where they would be most effective!

• No Receiver Does this because nobody with clout asked!( That is it is a feature that could be burned into a receiver chip)


18

Data IntegrationKalman Filter Software

• Kalman Filter software usually uses GPS and aiding data . Example:odometer , heading reference, and/ or inertial system

– Provides smooth outputs between fresh GPS samples

– Can Provide Position Solutions ( albeit with growing errors) during GPS outages.

– Can slightly improve accuracy over just Raw GPS

– Can Provide good heading information

– Can Reject bad data (A Key Feature )


19

Kaman Filter Should Also Provide:

• Distance Along track• Heading• Quality of Solution ( Combination of Quality of

Raw data, Expected error, Number of rejections) – Used to Tie Down Distance , assess precision of

solution

• If not, they are selling you a Figure skate for Hockey!


20

Comments on Integration Components in Kalman Filtering

• Aircraft Grade GPS Inertial System would provide >100 ft along track accuracy extrapolating 1 mile at 60 Mph with no GPS.

• GPS/Odometer/ heading reference (Could be curvature from Geometry) filter would provide <12 ft along track accuracy extrapolating 1 mile at 60 Mph with no GPS.– Cheap and simple implemented in post time

• Guess which is a lot cheaper?


21

But then You have the GPS/INERTIAL /ODOMETER• Should be as good as the Odometer and

HR…If it is integrated properly… at 5X to 10X the cost.

• Your Geometry measurement System may need Inertial Data for Stabilization of measurements.. So get it , but not Just for position improvement!


22

How to get the Most out of Your GPS

• Make Sure Your Receiver is Navigating !!!!!!!• Know where the GPS antenna is with respect to

sensors (then you can compensate)!• Have a Track Map at least as accurate as your GPS (Do

you have to? No more than you have to have air conditioning in you car in Texas)

• A Chain is only as strong as its weakest link

– When you plan think of Location and Re-Location!


23

About Synchronization • SYNCHRONIZE Your Sensors to Odometer

– Odometer is your interpolator

• SYNCHRONIZE Your GPS to Odometer– Odometer Distance between GPS outputs should be

consistent with GPS distance Difference

• SYNCHRONIZE GPS to Sensors ( Redundant? Maybe)– If you travel at 50 ft/sec, and there is a synch error of .2

sec between GPS output and Sensor,You are off by 10 ft! – Timing or Synch Jitter is the worst.. Because you can’t

compensate!• GPS Latency: GPS solution is before time of output. Good

receivers will extrapolate to to output time.


24

Using Odometer as Strobe

• Most Geometry Car data is strobed to the (ft) count of the odometer. Its is logical

• GPS Data should be ( and is in many systems) Strobed in the same way. – Fresh GPS is detected at change.

– Sampling error can be no more that a ft.

– When it is done right, everything is time synched( even though we might not exactly what time it is)

– And yes, you can have a Kalman filter “without time”


25

GPS Sampling Rates

• When there is an odometer providing data every ft or less, there is no need for GPS outputs faster that 1/sec… Software can do any necessary interpolation.

• Faster GPS sampling means less sophisticated Interpolation software.

• BUT it won’t significantly improve accuracies.


26

What should be on a Track Map?• Dependent Variables :LAT/Long (Elevation), AZIMUTH

every ~15 –50 ft

• Independent Variable: Distance Along track Si: Preferably Derived from a GPS input based Algorithm

• LAT/Long derived from survey,fly maps, or GPS derived maps.

• Beware of Datum shifts (Does Positioning During Mapping use same earth model as Navigating GPS?)!

L1,

L2, L3,

S1 S2 S3

Format?


27

Can Make A Track Map With Geometry Car GPS , Ancillary data and Optimal Smoothing


28

Some Uses Of Track Map• Convert Lat/Long to distance along Track and

vise/versa. Lets High tech meet low tech

• Provide heading for Lever arm correction, and Estimation.

• Provide common Base for Communication, data collection

XDefect file/map

CrackedTie

312 ft X


29

What data to Record on the Big Guys

• Latitude , Longitude, {Elevation} Time • Distance along track ( odometer) • Heading change( if Curvature is not on Geometry) • Integrity Related:

– Differential Quality , – Hdop ( measure of accuracy),– number of satellites in solution (good for in finding

integrity)…– Better: Sat numbers in solution ( can be used In post

processing for correction)


30

If Kalman Filter Output is Available on Geometry Car

• Kalman filter output of LAT/Long/Heading {Distance along Track or correction to Odometer}

• {Filter Integrity Measure}• Measure of Raw heading change• + All data on previous slide

– To Recover from “Computer and software crashes”

– To Apply Optimal smoothing for more accurate positioning , and making semblance of map


31

Data To Collect on “Low Tech Little Brothers” Having GPS

• At fault, lat/long , milepost and ft. If available

• Distance from fault to next good GPS reading. Could be “0”– Lat/long at this point– With track map this is now a redundant

measurement


32

Odometer Error Sources and Suggested Accuracies

• Scale factor stability(.002 in say 200000 ft)• Random ( show example) • Low or no (Reasonable) speed sensitivity• Event related:

– Wheel skips ( best to have odometer on “free wheel”)

– Low speed effects( with magnetic odometers…test it at very low speeds)


33

6000004000002000000E0

160

140

120

100

80

60

40

20

0

-20

-40

-60

Distance along track, ft

Odo

met

er E

rror

, ft.

Scale Factor, slope =.00018

Random Error

This odometer Error Would be O.K.


34

Odometer Sampling Rates

• For Geometry car collection

• One pulse/ft MINIMUM

• 10 pulses or better/ft better


35

Resultant Capabilities of Railroad Specific Receiver

• Fast or seamless acquisition or reacquisition of signal after short length outages.

• Differential Corrections wherever there is a GPS signal ( alternate modes of receiving differential corrections)

• Capability 1-D navigation when as few as two satellites are in view and a track map available. (Extension of 2-D navigation when only three satellites in view) (Don’t Hold Your Breath)

• Output only when there is High integrity or warning when Integrity cannot be validated


36

Rome’s Version of A Railroad Specific Receiver

• For signal acquisition,reacquisition and tracking in weak signal environments

– Maximum Bandwidths of Receiver tracking loops consistent with the Railroad Operation accelerations ( Should be Adjustable at least by Manufacturer )

– Tracking loop aiding consistent with Odometer /heading reference inputs ( should be better and cheaper than MEMS aiding, but is not currently available)

• Antenna with rate Gyro (Certainly not Standard)


37

Rome’s Version of A Railroad Specific Receiver (Cont.)

• RAIM (Internal Integrity Monitoring) tuned to accuracies of Differential Systems

• Software consistent with (1-D) navigation … factoring in track map into solutions (may be a long time coming)

– Built in integrity based on track maps

• Receptors of Differential Corrections coming from “Broadband ( like internet, cell phone, even perhaps TV, track based communications… It will come!)

Rome Navigation Innovations 5/14/07 1 GIS and GPS for Railroad Environment Presented by: Jim Rome May 15, 2007 Rome Navigation Innovations Inc 27 Old County.

Documents