1

GNSSGNSS--Based AutoBased Auto--Guidance Guidance Test Program DevelopmentTest Program Development

ECPA (ECPA (SkiathusSkiathus, Greece), Greece)June 6, 2007June 6, 2007

Viacheslav I. AdamchukViacheslav I. Adamchuk Roger M. HoyRoger M. HoyGeorge E. MeyerGeorge E. Meyer Michael F. Michael F. KocherKocher

Biological Systems EngineeringBiological Systems EngineeringUniversity of NebraskaUniversity of Nebraska--LincolnLincoln

BackgroundBackground• Auto-guidance (auto-steering) systems

provide numerous benefits• Performance of different systems

depends on internal and external factors

• Quality of auto-guidance performance must be quantified in repeatable manner

• ASABE has two active projects:– GPS Dynamic Test Standard (X578)– Auto-Guidance Test Standard (X605)

OutlineOutline

• Why do auto-guidance systems perform differently?

• What is “guidance error”?• 2005 auto-guidance field day demo

review• 2006 pilot test of different systems

using Nebraska Tractor Test Laboratory’s test track

• 2007 instrument development and future plans

Agricultural Vehicle GuidanceAgricultural Vehicle Guidance• Reference method

– Local triangulation– Crop-based methods

• Mechanical feelers• Laser row tracking• Machine vision

– GNSS-based guidance• Level of assistance

– Navigation aids• Lightbar parallel tracking

– Auto-guidance• Auto-steering

– Autonomous vehicles• Field robots

GNSSGNSS--Based AutoBased Auto--GuidanceGuidancePositioning

Sensor

DGPS (< 4 in)

RTK (< 1 in)

Terrain Compensation

Gyroscope

Multiple Units

Vehicle Control

Mechanical

Hydroelectric

Popular GPS SolutionsPopular GPS Solutions

• Single Frequency Receivers – WAAS, EGNOS, Beacon– OmniSTAR VBS, John Deere SF1– Subscription/Free (sub-meter accuracy)

• Dual Frequency Receivers – OmniSTAR HP/XP, John Deere SF2– Subscription (decimeter accuracy)

• RTK Receiver– Base Station (centimeter accuracy)

2

AutoAuto--Guidance ApplicabilityGuidance Applicability

• Sub-meter accuracy – 2-4 ft year-to-year– 1 ft pass-to-pass– Mechanical steering

• Decimeter accuracy– 8 in year-to-year– 4 in pass-to-pass– Electro-hydraulic controls

• Centimeter accuracy– 1 in year-to-year– 1 in pass-to-pass– Base station

Tillage/DiskingSpraying/SpreadingHarvestingSeedingMappingPlantingCultivatingBeddingStrip TillingDrip Tape PlacementLand LevelingTopographic Mapping

What does “What does “±±”” actually mean ?actually mean ?

• Nature of the test– Static vs. dynamic

• Duration of the test– Pass-to-pass vs. long-term

• Definition of the test– Precision versus accuracy

• Statistic used– 68% (1σ) vs. 95% (2σ) prediction

• Type of error– Positioning vs. guidance error

Cross-track error

AutoAuto--Guidance ErrorGuidance Error

Guidance Error

Implement Tracking

GPS Positioning

Error

Field Conditions

Vehicle Dynamics

2005 Auto2005 Auto--Guidance Field DayGuidance Field Day

August 18, 2005 – ARDC, Mead, Nebraska

2005 Field Demonstration2005 Field Demonstration

• Pull-type cart

• J-type course

• Coulter marker

• RTK-level GPS position logging

Demonstration CourseDemonstration Course

0

20

40

60

80

0 20 40 60 80 100 120 140 160

Easting, m

Nort

hing

, m

Starting point

Straight pass

Contour pass

Terrace crossing

3

Straight PassStraight Pass

5 0

5 1

5 2

5 3

5 4

5 5

5 6

5 7

5 8

5 9

6 0

4 0 6 0

E a sti n g , m

Nort

hing

, m

5 5

5 6

5 7

5 8

5 9

6 0

6 1

6 2

6 3

6 4

6 5

8 0 1 0 0

E a sti n g , m

Nort

hing

, m

Contour PassContour Pass

20

40

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Easting, m

Nor

thin

g, m

Error DefinitionError Definition

α

(X1,Y1)

(X2,Y2)

(XA,YA)

ε

L12

LA2

L1A

212

21

22

212

212

1221

212

21

22 2

coscos2LLLLL

LLLLLA

AAAAA

−+=⇒−+= αα

αα 2cos1sin −= 2

212

21

22

212

21

11 21sin ⎟⎟

⎠

⎞⎜⎜⎝

⎛ −+−=⇒=

LLLLLLL

A

AAAA εαε

-25

-20

-15

-10

-5

0

5

10

15

20

25

0.5 1 .5 2 .5 3 .5 4 .5 5 .5

Cros

s-Tr

ack

Err

or, c

m

Curved PathStraight and Level PathStraight and Sloped Path

Error DistributionsError Distributions

Exact backtracking

System 1 System 2 System 3 System 4 System 5

-8

-6

-4

-2

0

2

4

6

8

0 1 2 3 4 5 6 7 8 9 1 0 1 1

Cros

s-tr

ack

Err

or (c

m)

Signed ErrorUnsigned ErrorNormal ModelRayleigh ModelZero ErrorAverage ErrorMedian Absolute ErrorMean Absolute ErrorRMSE

Error DistributionsError Distributions

Signed error distribution Unsigned error distribution

Cumulative DistributionCumulative Distribution

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

-8 -6 -4 -2 0 2 4 6 8

Cross-Track Error, cm

Cum

mul

utiv

e Pr

obab

ility

Signed ErrorUnsigned ErrorNormal ModelRayleigh ModelZero Error95% Threshold

4

ProducerProducer--Viewed DifferencesViewed Differences

• Interface and ease of use• Modes of operation• Installation time and options• Setup and calibration procedure• Cost and possible upgrades• Versatility and secondary use• Technical support

2006 NTTL Track Testing2006 NTTL Track Testing

Trigger

Linear potentiometerReference GPS antennae

Test cart

Two straight parallel passes

RTK Base Station

Linear Potentiometer System (LPS)Linear Potentiometer System (LPS)

Trigger

V R1 Resistor array

washers Common array

washers

Vertical support

Nuts

1.95 cm

Rw Rw Rw Rw Rw Rw

Steel rod

PVC tubing

PVC spacer

2006 Pilot Test2006 Pilot Test

• System A – Centimeter-level accuracy• System B – Decimeter-level accuracy• LPS – based analysis• GPS – based analysis• Straight passes• Three 15-min runs• Two consecutive days• Urban environment• Concrete pavement

Test SequenceTest Sequence

0

25

50

0 25 50 75 100 125 150 175 200 225 250

Easting, m

North

ing,

m

Test Run 1 Test Run 2 Test Run 3 Test Run 4

North – 9 triggers

South – 11 triggers

Spacing – 15.2 m

Swath – 38.1 m

A-B line

5

LPS OutputLPS Output

02468

101214

-39.0

-35.1

-31.2

-27.3

-23.4

-19.5

-15.6

-11.7

-7.8

-3.9 0.0 3.9 7.8

11.715 .6

19.523 .4

27.331.2

35 .139.0

R e la tive p o si tio n , cm

Num

ber o

f occ

uran

ces

N o rth 1N o rth 2N o rth 3S o u th 1S o u th 2S o u th 3

System A

02468

10121416

-39.0

-35.1

-31.2

-27.3

-23.4

-19.5

-15.6

-11.7

-7.8

-3.9 0.0 3.9 7.8

11.715.6

19.523.4

27 .331.2

35.139 .0

R e la tive p o sitio n , cm

Num

ber o

f occ

uran

ces

N o rth 1N o rth 2N o rth 3S o u th 1S o u th 2S o u th 3

System B

Reference GPS OutputReference GPS Output

R2 = 0.95

R2 = 0.95

38.0

38.5

39.0

39.5

40.0

40.5

41.0

41.5

42.0

0 50 100 150 200 250

Easting, m

Nort

hing

(Nor

th p

ass)

, m

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

6.0

North

ing

(Sou

th p

ass)

, m

NorthSouthLinear (South)Linear (North)

Error DefinitionError Definition• Standard error

– Difference between each trigger hit position and an average (balanced in terms of direction of travel) of all relevant hit positions calculated for each trigger separately (within 15 min for short-term error and all test data for long-term error)

– Standard error of a simple linear regression for the GPS data

• Cross-track error– Expected difference between two tracks randomly

selected from either a short-term or long-term pools of data

– Theoretically, cross-track error equals to standard error multiplied by square root of 2

System A System A –– centimeter accuracycentimeter accuracySystem A - GPS reference (short-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

System A - GPS reference (long-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

System A - LPS reference (short-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

System A - LPS reference (long-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

System B System B –– decimeter accuracydecimeter accuracySystem B - GPS reference (short-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

System B - GPS reference (long-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

System B - LPS reference (short-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

System B - LPS reference (long-term error)

0

1

2

3

4

5

6

7

8

-40 -30 -20 -10 0 10 20 30 40

Guidance error, cm

All dataSouth CW

South CCWSouth all data

North CWNorth CCW

North all data

6

ShortShort--term Error (Passterm Error (Pass--toto--Pass)Pass)Cumulative short-term error distribution

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Error, cm

System A (LPS)System A (GPS)System B (LPS)System B (GPS)

95% probability

LongLong--term Error (Dayterm Error (Day--toto--Day)Day)Cumulative long-term error distribution

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Error, cm

System A (LPS)System A (GPS)System B (LPS)System B (GPS)S t A GPS (95%)

95% probability

Standard Related Discussion ItemsStandard Related Discussion Items

• Test location– Surface conditions– Clarity of sky– Test course segments

• Method of measurement– Reference receiver– Total station– Mechanical (contact)

sensor– Optical (non-contact)

sensor

• Test sequence– PDOP requirements– Pass-to-pass test– Day-to-day test

• Error terms– Dealing with bias– Parametric and non-

parametric estimates• Machinery selection

– Tractors– Sprayers

Sensor calibration

Optical Sensor DevelopmentOptical Sensor DevelopmentVideo camera with internal processor

Direct hard mounting to the tractor’s chassis

Locating the center of a painted line to within 2 mm

2007 Test Plans2007 Test Plans• Measurement close to the drawbar pivoting

point• Three 15-min test run sequences repeated

three times• Actual cross-track error calculation• Non-parametric statistics only• Three test segments

– Linear flat– Curved flat– Curved sloped

• Line acquisition test trial http://bse.unl.edu/adamchukE:mail: vadamchuk2@unl.edu