Interfacing Sensors with (VR)Application Equipment

Scott Drummond Ken Sudduth IT Specialist Agricultural Engineer

Objectives

• Understand the “big picture” of developing a sensor based system for VRA of N.

• Recognize design details that often get ignored or at least “underappreciated”.

• See how these details affect the design and development of one such research system.

Sense Decide Control

Crop Sensing

• Remote sensing– Satellite based– Aerial based

• Real time sensing– Passive Sensors – Active Sensors

Active Sensors

• By using an internal light source, these sensors eliminate problems with sun angle and cloud variations– GreenSeeker by NTech – Crop Circle by Holland Scientific– CropSpec by Topcon

Effect of soil on active sensors?

V7 corn with 0 N

N applied at Planting (kg ha-1)

Days after planting / Growth stage

0 45 246

23 DAP V4

NDVI: 0.36 ISR: 0.47 SPAD: 39.8

NDVI: 0.36 ISR: 0.47 SPAD: 41.1

NDVI: 0.36 ISR: 0.47 SPAD: 43.7

41 DAP

V7

NDVI: 0.53 ISR: 0.31 SPAD: 48.6

NDVI: 0.53 ISR: 0.31 SPAD: 52.8

NDVI: 0.57 ISR: 0.28 SPAD: 58.8

47 DAP

V10

NDVI: 0.64 ISR: 0.22 SPAD: 49.9

NDVI: 0.66 ISR: 0.21 SPAD: 52.8

NDVI: 0.70 ISR: 0.18 SPAD: 57.6

56 DAP

V13

NDVI: 0.66 ISR: 0.21 SPAD: 45.1

NDVI: 0.68 ISR: 0.19 SPAD: 52.4

NDVI: 0.73 ISR: 0.16 SPAD: 59.8

Is soil an important part of the signal?

Do we need to consider a way to remove the effect of soil?

What happens when the soil “color” varies across time or across the landscape?

Stability of sensor readings?

0.6

0.65

0.7

0.75

0.8

0.85

0.9

6:2

7

6:5

5

7:2

3

7:5

1

8:1

9

8:4

7

9:1

5

9:4

3

10

:11

10

:39

11

:07

11

:35

12

:03

12

:31

12

:59

13

:27

13

:55

14

:23

14

:51

15

:19

15

:47

16

:15

16

:43

17

:11

17

:39

18

:07

18

:35

19

:03

19

:31

19

:59

20

:41

Time on 10 July 2006

ND

VI

40 inch

10 inch

20 inch

RainDew

Courtesy: Dr. Peter Scharf

Is variable crop height an issue?

If more H2O affects readings – how about less?

Variable Rate Controllers

• Things you must consider when selecting the controller for your VRA system…

– VRA control type– Range of rates– Response time– Precision and accuracy– Communication method(s)

Variable Rate Controllers

• Many systems claim VRA control but…

– Real time control

• Message based • Controller includes decision module

– Map based control• Useful for image based methods – much less

attractive for active sensor applications

Variable Rate Controllers

• Range of rates for:– Dry fertilizers

• Range generally not an issue

– Liquid fertilizers• Standard pressure regulated • Capstan spray system (PWM) • SprayTarget variable flow nozzles

Variable Rate ControllersCorn Yield

(bu/ac)

124 to 180 110 to 124 101 to 110 95 to 101 89 to 95 83 to 89 76 to 83 68 to 76 52 to 68 30 to 52

Delay Time = 9 sDelay Time = 14 s

What COULD happen IF our response time was too slow?

Variable Rate Controllers

• Communication issues

– Serial (RS-232/RS-422/RS-485)– CAN Bus– As applied maps – stored where/how?– Message formats can be open or proprietary

Sense Decide Control

Decision Module

• Things to consider when selecting the decision module for your VRA system…

– Communication– Algorithm(s)– Flexibility

Decision Module

• Questions to ask yourself…

– How many algorithms are available?– Is my algorithm “stable”?– Can I adjust (timing/layout/parameters)?– What happens when a new piece of

information (sensor/map) appears?

Designing a VRA System

• Now that we have an idea of some of the questions to ask… let’s look into the design of a system based upon a set of requirements.

• This system was designed for research applications, and may have more stringent requirements than some.

Requirements

• Use existing Spra-Coupe • Plot sizes down to 5x10 m in size• Range = 0-210 lb/a N• Precision = 30 lb/a• Accuracy < 5% of full scale.• Map based and sensor based VRA needed• GS & CC sensor data collected and/or used• Algorithm – complete flexibility needed

Application System

• Used existing AGCO Fieldstar controller in the SpraCoupe to change system operating pressure to compensate for changes in ground speed.

• To get fast response, we chose a “bypass” or 3-way valve system. – When a particular valve (1x, 2x, or 4x) was

not sending N to the ground, that same volume of flow was returned to the sprayer tank through a matched orifice.

– The pump was always putting out the same volume at the same pressure, and the pressure control system did not have to respond (at least theoretically).

Application System

• We chose a 6-row system for reasonable plot widths– Near maximum capacity of the

SpraCoupe pump at normal operating speeds

• Drop nozzles with 1x, 2x, and 4x orifice plates were installed in row middles

• Nominal application rates:– 1x = 30 lb N/acre– 2x = 60 3x = 90– 4x = 120 5x = 150– 6x = 180 7x = 210

Data Flow

Collect Reference Strip Data

Interpolate/ extrapolate whole-field

reference map

Get Current GPS data

Prior to Application

Get Reference Value at

Current Point

N Recommendation

Algorithm

Smoothing, Deadband, Hysteresis

Solenoid Valve

Control

0, 1x, 2x, 3x, 4x, 5x, 6x, or 7x

Green GreenSeeker

1

Green GreenSeeker

2

Crop Circle 3

Crop Circle 4

Select and/or Combine Sensor Outputs

Spatial or time-base

filtering

Decision Module

Finding the target sensor data…

Drop NozzlesN Sensors

a bGiven that:Sensor data buffered at 10 Hzv = GPS velocity (m/s)a+b = dist from sensors to drops (m)

L = system latency (s)

The target sensor data was takenthis many readings ago…

t = 10*(((a+b)/v)+L)

In practice, we have averaged 1s of data (10 values per sensor) centered around this target point.

Positioning details…

Drop Nozzles

N Sensors

GPS AntennaGives Easting(x), Northing(y),

h(eading) and v(elocity)

a b

Application Boom Centereboom = cos(90-h)*b+egps

nboom = sin(90-h)*b+ngps

Individual Sensor Locationseright = cos(90-(h+atn(c/a))*sqrt(c2+a2)+egps

nright = sin(90-(h+atn(c/a))*sqrt(c2+a2)+ngps

eleft = cos(90-(h-atn(d/a))*sqrt(d2+a2)+egps

nleft = sin(90-(h-atn(d/a))*sqrt(d2+a2)+ngps

c

d

Sensor Boom Locationesens = cos(90-h)*a+egps

nsens = sin(90-h)*a+ngps

Software Control Loop…Collect store and buffer

data: sensors, GPS, psi,

status, etc.

Find Target DataFind N-Ref Data

Calc Raw N-Rate

Time>1s?

N-Rate< MIN?

N-Rate > MAX?

Map to 0X-7X

BeyondDeadband?

N-Rate = MIN

N-Rate = MAX

Send New Rate

To Controller

No No

No

No

Yes Yes

Yes

Yes

How Well Did it Work ?

• Accuracy and consistency of response