Precision Agriculture (2019) 20:840–856

https://doi.org/10.1007/s11119-018-9616-z

Use of real‑time extend GNSS for planting and inverting

peanuts

Adão Felipe dos Santos1 · Rouverson Pereira da Silva1 · Cristiano Zerbato1 ·

Patricia Candida de Menezes1 · Elizabeth Haruna Kazama1 ·
Carla Segato Strini Paixão1 · Murilo Aparecido Voltarelli2

Published online: 19 October 2018

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract
Among the main techniques employed in precision agriculture, yield mapping and auto-
matic guidance of agricultural machines are the best-known to farmers. The objective of
this study was to evaluate, using statistical process control tools, the quality of automatic
guidance using satellite signals, to reduce positioning errors and losses in peanut digging.
The treatments consisted of the use of manual (operator guidance) and automatic (autopi-
lot) guidance with RTX satellite signals in sowing and digging operations. The quality of
the operation was evaluated after collection of 30 points spaced at 100 m for each qual-
ity indicator, which are the losses and the errors of alignment of the mechanised sets in
sowing and digging operations. From the perspective of statistical control, manual guid-
ance was shown to be compromised for the quality indicators of digging losses. Despite
the instability in the sowing and digging operations, the use of automatic guidance proved
to be accurate. The use of automatic guidance increases the precision and reduces overlaps
(< 38 mm, as stipulated by the supplier) for sowing and digging. The manual sowing mean
error between overlaps was stable; however, it did not remain constant over time.

Keywords Global navigation satellite system (GNSS) · Precision point positioning ·

Mechanized harvest · Peanut digging

Electronic supplementary material The online version of this article (doi:https​://doi.org/10.1007/s1111​

9-018-9616-z) contains supplementary material, which is available to authorized users.

* Adão Felipe dos Santos

[email protected]
1
Department of Agricultural Engineering, São Paulo State University, Via Access Prof. Paulo
Donato Castellane s/n, Jaboticabal, SP ZIP: 14884‑900, Brazil
2
Federal University of São Carlos, Lagoa do Sino Campus, Highway SP‑189, Lauri Simões de
Barros, km 12, Buri, SP ZIP: 18290000, Brazil

1 Vol:.(1234567890)
3
Precision Agriculture (2019) 20:840–856 841

Introduction

Peanut farming is carried out in several parts of the world, such as China, India, Nigeria
and the United States producing 18.10, 6.00, 3.20 and 2.78 Mt, respectively (USDA 2018).
In Brazil, production of the 2016/17 crop was 411.30 kt, which represented an increase of
2.9% in relation to the previous harvest (CONAB 2017). This increase may seem relatively
small; however, it is significant for the most part where peanut is sown, with São Paulo
being the main producer of the grain, accounting for 90% of all production during the sug-
arcane rotation.
With the increase in production and the lack of space to sow peanuts, producers can
use automatic guidance of machines as an aid in production, which increases the quality
and precision of operations in the field by enhancing the efficient use of resources, reduc-
ing input costs and minimising environmental degradation (Tey and Brindal 2012). Addi-
tionally, automatic guidance increases hours worked, allowing possible working day and
night. (Oliveira and Molin 2011) and reduction of overlap between the passes (Kroulík
et al. 2011). These benefits can be observed in various mechanised operations, from soil
preparation to harvesting (Roberson and Jordan 2014).
To ensure these benefits, the positioning method should be accurate. In agriculture, real
time kinematic (RTK) positioning allows the navigation of agricultural machines using sig-
nals with radio correction emitted from a fixed geodetic station (base station) located near
the machine and guarantees errors in the order of 25 mm (Baio and Moratelli 2011). How-
ever, when the distance between base and rover of machine is greater than 10 km, the radio
frequency begins to lose efficiency and the signal begins to lose accuracy in the positioning
(Perez-Ruiz and Upadhyaya 2012), which indicates the need for the installation of a repeti-
tive signal base to assure quality in positioning. This is not useful for Brazilian peanut
producers, who are mostly tenants of land areas used for sugarcane within a crop rotation;
that is, they do not have fixed areas for cultivation and the areas are distant from each other.
An alternative for correction signal via radio may be the use of satellite signal with high
accuracy, such as RTX (Real Time eXtend), this type of signal uses the precision point
positioning (PPP) method. The RTX signal consists of several reference bases distributed
across the globe, enabling machine position corrections in the field, with a guaranteed
accuracy of 38 mm for 95% of the time (Trimble RTX 2017). The reference bases emit a
signal to a central base where the correction calculations on the positioning are made (via
algorithms). The corrected signal is transmitted to geostationary satellites, which send the
signal to the rover of the machines independent of the distance and location. The method
reduces the problems of signal degradation, which arise from long distances, and the loss
of signal usually found in radio correction. However, this type of signal can still be delayed
before it reaches the rover of the machine.
It is worth mentioning that the good performance of guidance through the Global Navi-
gational Satellite Systems (GNSS) is mostly associated with errors that are transversal to
the trajectory of the machines and they can be generated during execution of the operation
in the field. Easterly et al. (2010) stated that positioning errors can be attributed to fac-
tors such as vehicle dynamics, type of implement traction and especially the environmental
conditions in which the operation is performed. In addition, they are sensitive to large and
small climatic variations such as increased cloudiness and ionospheric scintillation.
Brazil, having a large part of its territory located in the geomagnetic equator, suffers
intense effects of ionospheric scintillation. In critical scenarios characterized by poor sat-
ellite coverage and scintillation incidence, loss of positioning service can occur, which

13
842 Precision Agriculture (2019) 20:840–856

impairs activities that depend on this type of service, such as air navigation and precision
agriculture (Rocha et al. 2017).
In the United States, many peanut farmers have adopted automatic guidance because it
can theoretically result in high productivity gains when used in sowing and digging (Lei-
dner 2012; Vellidis et al. 2014). Ortiz et al. (2013) indicated that for every 20 mm of line
deviation, an average of 186 kg ha−1 losses can be expected. The study also showed that
the farmer who uses an automatic guidance system and the RTK signal with an accuracy
within 25 mm can expect net returns between 94 US $ ­ha−1 and 404 US $ h­ a−1 compared
to deviations of 90 mm based on gross margin. In Brazil, Santos et al. (2016a), verified that
the use of the RTK signal provides a better quality of the peanut digging operation. How-
ever, this is an automatic guidance system with a high initial investment cost, which makes
it not feasible for tenant farmers in Brazil.
The studies cited above evaluated the radio signal (RTK) associated with the use of a
tractor equipped with a hydraulic autopilot, however, there are few studies that use electric
pilot in the peanut crop. The main difference between these autopilot mechanisms is the
response time. On the electric pilot, the corrected signal acts first on the steering wheel
then comes to the wheels to correct the machine’s position (slower response time), while
the hydraulic acts directly on the tractor’s wheel (faster response time).
Thus, it is necessary to evaluate the quality of automatic guidance by an electric pilot
using satellite signals of RTX to examine its effectiveness in the guidance of the machines
in the field and to reduce peanut harvesting losses.
The tool often used by researchers to evaluate the quality of agricultural operations is
statistical process control (SPC) using control charts (Zerbato et al. 2017a; Paixão et al.
2017; Voltarelli et al. 2017; Tavares et al. 2017). The control charts give an idea of the var-
iability and stability of the operation over time or space by means of control limits (Mont-
gomery 2009). Efficiency of monitoring of operational quality of machines using a GNSS
signal was proven by Santos et al. (2016b). They tested different standard deviation estima-
tors and concluded that twice the standard deviation should be used in the evaluation of
the error between the passes of the mechanised set using GNSS signals to establish control
limits, which guarantees greater rigidity in the analysis and interpretation of the process.
From the SPC perspective, errors can be caused by common (or random) factors and
special (or marked) causes of variation. According to Samohyl (2009), errors are caused
by the action of the 6 M factors (Machine, Manpower, Material, Environment, Mother
Nature and Measurement). Therefore, monitoring the quality of the process is essential so
that these causes can be identified and later corrected to ensure that the process returns to
normal.
Thus, the objective of the study was to evaluate, using SPC tools, the quality of auto-
matic guidance using the electric pilot and satellite signals and to investigate the reduction
of errors between the mechanised set and losses in the mechanised digging of peanuts, as
well as to examine the feasibility of using this positioning method in agriculture.

Materials and methods

Experimental sites and treatments

The experiment was conducted in farm Santa Candida, a municipality of Luzitânia, São
Paulo, Brazil, located near the geographic co-ordinates 21°06′39″S and 48°14′09″W using

13
Precision Agriculture (2019) 20:840–856 843

the Geodetic System WGS84 reference. The average altitude and slope was 540 m and
2.5%, respectively, with climate Aw according to the Koppen (1948) classification. The
cultivar used was Runner, sown with 0.90 m between rows, in soil of clay texture (Embrapa
2013). The lines were designed using project carried out with the A ­ groCAD® software
(TECGRAF, Campinas, São Paulo, Brazil), so that the tractor would always move in a rec-
tilinear course in both treatments.
The treatments used were the Autopilot (A/G) and Manual guidance (M/G) in sowing
and digging peanuts.
In the A/G treatment, the tractor was guided by a satellite signal of high precision,
which was the Real Time eXtend (RTX) from ­Trimble® (Sunnyvale, CA, USA). This sig-
nal consists of the PPP method, kinematic and real-time. Its principle of operation is based
on several reference bases spread all over the globe that receive and send the signal to a
control centre, where calculations and processing of information are performed. The cor-
rected information is then sent to a geostationary satellite that relays the information to
all receivers that are within reach, ensuring an accuracy of up to 38 mm, according to the
signal provider.
To receive the signal, the tractor was equipped with the EZ-steer electric pilot (Sunny-
vale, CA, USA) located in the tractor steering wheel, with the antenna model AG25 (fea-
turing multiband receivers, L1 and L2, and works on the GPS, GLONASS, SBAS, RTK,
and OMNISTAR signals, as well as the RTX signal), and the T ­ rimble® CFX750 monitor/
processor.
In the M/G treatment, only the line mechanical marker was used in the tractor-sowing
direction (the line marker demarcates on the ground where the tractor wheel should return
next pass); for digging, the operator’s own vision and experience were used for directions
on the lines of the crop during the operation.
The working width of the sower as well as the spacing of the crop were considered in
carrying out the project, with the lines for orientation autopilot spaced 3.60 m apart. From
the sowing project, the peanut crop project was developed.
The sowing and digging operations for both treatments were performed by the same
mechanised sets. In sowing, a John Deere tractor, model 6110 J (80.9 kW), with a speed of
displacement of 7 km h−1 and a seeder-fertiliser PHT4 Suprema, were regulated to obtain
a seeding density of 21 seeds ­m−1. The digging operation was undertaken on 2016-03-03,
120 days after sowing. An MF 7390 Dyna-6 (139.7 kW) tractor with an average displace-
ment speed of 4.5 km h−1 was used, with an engine rotation of 1500 rpm, following the
recommendation of the digger–shaker–inverter manufacturer. The digger–shaker–inverter
used was the KBM brand agricultural equipment (Dumont, São Paulo, Brazil), 2-row
model.

Characteristics evaluated

Alignment errors in sowing and digging operations

To evaluate the error between lines/windrows of peanut, about 100 m of 4 lines/windrow

of peanut were mapped using the Trimble R6 GNSS receiver (semi-kinematic relative
positioning method); the receiver uses the RTK signal, which guarantees an accuracy of
25 mm. This receiver has a dual frequency (bands L1 and L2) and consists of a fixed base
(known latitude and longitude), a mobile antenna that is coupled to a pole, as well as a

13
844 Precision Agriculture (2019) 20:840–856

Trimble TSC3 data collector. The accuracy of this equipment is measured according to the
occupancy time at each point; to ensure good accuracy, 15 s per point was used.
Approximately 200 points were collected (50 points per line/windrow of peanut). In
both sowing and digging operations, the error between the passes was obtained by consid-
ering the useful width of the equipment (3.60 m and 1.80 m for seeder-fertiliser and dig-
ger–shaker–inverter, respectively).
After field mapping, the data were exported to ­AutoCAD® software (Autodesk, San
Rafael, California, USA) and the respective lines/windrows of peanut were drawn. Then,
lines perpendicular to the lines/windrows of peanut were generated to verify the distance
between the passes of the mechanised sets in each operation.
At each sampling point, the error between the passes was calculated by the difference
between the predicted value (3.6 m and 1.8 m) and the actual value found in the field in
each operation (Fig. 1) without considering the GPS error. That is, the value found between
the passes is the error of the mechanized set plus the error of both signals.

Harvest losses (digging)

The digging losses were classified as follows: pods above soil (PVA), below soil (PIA), and
the sum of the previous two (PTA).
To evaluate the losses (Fig. 2), a metal frame with dimensions of 1.8 m (length) and
1.11 m (width) was used, giving a sample area of approximately 2 m2. The length of the
frame corresponds to the working width of the digger–shaker–inverter, where it must be
placed transversely to the windrows of peanut. A total of 30 samples were collected in the
4 evaluated lines, spaced each 50 m.
For PVA, plants were carefully removed from within the frame by manually collecting
all pods that were on the ground within the frame (Fig. 2a). Without removing the frame,
the PIA (Fig. 2b) was collected at the same point, digging up to a depth of 150 mm and
mixing all the soil inside the frame, and then the pods that had adhered to the soil were

Fig. 1  Sample diagram of errors between the mechanized set at sowing (a) and digging (b)

13
Precision Agriculture (2019) 20:840–856 845

Fig. 2  Sampling scheme for evaluation of peanut crop losses. Losses on soil (a), and losses below soil (b),
both collected at the same sampling point

collected. After collecting the losses at each point, the pods were packed in paper bags,
weighed and identified.
The samples were sent to the Unesp/Jaboticabal Agricultural Machines and Mechani-
zation Laboratory (LAMMA) in São Paulo, Brazil, where they were placed in an oven at
105 °C for 24 h until they reached a constant mass. After drying, the pods were again
examined by separating and discounting the mineral impurities adhering to the pods and
obtaining the values of the losses that were extrapolated to kg ha−1, with a subsequent cor-
rection to account for 8% of water content, a value used for the storage of peanuts.
The value of losses were calculated in kg ha−1 in relation to the average gross productiv-
ity of each treatment, which was determined by collecting all the pods that were attached to
the plants and those below and above the soil within 2 m2.

Other variables affecting losses

The maturation was evaluated according to the proposed methodology of Williams and
Drexler (1981), which consists of scraping the pods with a jet of water to remove the exo-
carp and expose the mesocarp. The darker the colour of the pod, the greater the degree of
maturation.
The water content of the soil was determined using the HydroSense II H2S equipment
coupled to the CS658 sensor with a 200 mm rod (Campbell Scientific, Logan, USA). The
equipment used a time domain reflectometer (TDR); the TDR rod in contact with the
ground determines the dielectric constant of the soil (Silva and Gervásio 1999).
The working depth of the digging was used to check whether oscillation at a depth
could influence the increase of losses as well as the destabilisation of the automatic
steering. At each sampling point (same points used to evaluate losses), five measure-
ments were taken with the aid of a 500 mm scale graduated iron rod inserted in the
soil until finding the area moved by the digger–shaker–inverter. Each measurement was

13
846 Precision Agriculture (2019) 20:840–856

measured within the sample area (2 m2), spaced at 350 mm from one another. The aver-
ages of the five observations were calculated, which were used to evaluate the depth of
cut of the digger–shaker–inverter at each sampling point.

Experimental design and statistics

The design followed the standard adopted by SPC, in which samples are collected over
time, where 30 points were collected for each variable in each treatment, except error
between passes. Sampling was performed to randomly collect data at a given frequency.
As the mean velocity of displacement was set at 4.5 km h−1, the samples were collected
each 80 s.
The analysis of the variability of process was carried out using SPC, with the help of
the ­Minitab® 17 program (www.minit​ab.com). The tools used were the control charts
of variables. The control card model selected was the Mobile Amplitude Individual,
which contains two graphs: the upper one, corresponding to the individual values sam-
pled at each point; and the lower one, obtained by the calculated amplitude between
two successive observations. The control limits were established while considering the
variation of data owing to uncontrolled causes in the process (special causes) and were
calculated based on the standard deviation of the variables, as shown in Eqs. (1) and (2)
with α = 0.01.
UCL = x̄ + 3𝜎 (1)
LCL = x̄ − 3𝜎 (2)
where UCL is upper limit of control; x̄ is general average of the variable; 𝜎 is standard
deviation; and LCL is lower limit of control.
For error charts between passes, the specific limit of control (CLS), represented by blue
lines, was established based on the information of the signal provider, which guarantees an
accuracy of 38 mm in 95% of the observations (Trimble 2017).
The calculation of control charts for error between the passes of the mechanised sets
followed the recommendation of Santos et al. (2016b), according to Eqs. (3) and (4), which
guarantee that 95% of plotted points are within the acceptance region (α = 0.05):
UCL = x̄ + 2𝜎 (3)

LCL = x̄ − 2𝜎 (4)
Control charts can be key components in the assessment of the accuracy and precision
of a GNSS, as precision is the variation value repeatedly measured under the same experi-
mental conditions around the observed mean value. Accuracy refers to how close the meas-
ured value is to the real value (Cappellli et al. 2006).
Thus, control charts where the values do not exceed specific limits, or the upper and
lower control ones indicate that the process has precision and accuracy. When values
exceed specified limits, but are still within the control limits, the process has only accu-
racy; when the process is out of control (points above or below the control limits), it has
neither precision nor accuracy and needs to be changed.
In addition, descriptive analyses (with absolute values of the data) to verify the behav-
iour of the data and analysis of normality by the Anderson–Darling test (p > 0.05) were
performed, where the normality of the data is a fundamental assumption for the develop-
ment of individual graphics and mobile range.

13
Precision Agriculture (2019) 20:840–856 847

Results and discussion

Descriptive statistics

From the results of the descriptive analysis (Table 1), the variables studied in the mecha-
nised peanut digging, with and without the use of automatic guidance, can be described as
normal except for the error between the digging passes with and without manual guidance,
according to the Anderson–Darling test (p value > 0.05).
According to Vaccaro et al. (2011), it is fundamental for the analysis of control charts
that the data present normal or near-normal probability distribution to avoid possible false
alarms during the process. However, in this study, the normality assumption criterion pro-
posed by Montgomery (2009) (which considers the data set as normal when the distri-
bution is close to a 5% probability) was used so that it does not interfere with the data
interpretation.
Regarding the maximum and minimum values presented in Table 1, it is noted that the
use of automatic guidance maintains the alignment between passes of the mechanised sets
for sowing and digging operations. Although the maximum loss values found in the dig-
ging using automatic guidance were lower, when the average of the two treatments were
evaluated, the losses were 13% more than those found in the manual guidance.

Statistical process control (SPC)

The use of the automatic guiding system allowed the reduction of errors between passes
of the mechanised set (tractor-seeder) as shown in Fig. 3, resulting in an improvement in
the quality of the sowing operation. When the improvement of the operation is reflected in

Table 1  Descriptive statistics for the quality indicators of the mechanized peanut digger process with and
without automatic guidance
Variables Mean σ CV Median Max Min AD

Automatic guidance
Error between sowing passes (mm) 29.4 ± 2.5 17.0 58.62 27.0 74.0 0.00 0.674N
Error between the past in the digging 41.1 ± 2.5 44.0 897.9 38.0 164.0 0.00 0.951A
(mm)
PVA (kg ha−1) 102.79 33.15 32.25 97.83 163.04 43.48 0.361N
PIA (kg ha−1) 93.51 29.10 31.12 95.11 146.74 31.53 0.182N
PTA (kg ha−1) 196.30 51.39 26.18 192.93 282.61 102.17 0.264N
Manual guidance
Error between sowing passes (mm) 54.6 ± 2.5 37.0 649.1 51.0 174.0 0.00 0.404N
Error between the past in the digging 176.0 ± 2.5 161.0 914.4 144.0 995.7 0.00 5.91A
(mm)
PVA (kg ha−1) 80.43 32.34 40.21 81.52 157.61 21.74 0.222N
PIA (kg ha−1) 89.86 51.06 56.82 76.09 315.61 38.04 2.042N
PTA (kg ha−1) 170.3 65.5 38.48 152.2 396.70 87.0 0.983N

PVA pods above soil, PIA pods below soil, PTA sum of the PVA and PIA, ± 2.5 error of GPS, σ standard
deviation, CV (%) coefficient of variation, Median median, Max maximum, Min minimum, AD Anderson–
Darling normality test, α = 0.05 (N normal distribution, A non-normal distribution)

13
848 Precision Agriculture (2019) 20:840–856

error tractor-sowing set passes (m)

0.2 A/G M/G
1

1
UCL
0.1
1

_
CSL
X
0.0 (I)
CSL
1
1 LCL
-0.1 1 1
11 1

1 13 25 37 49 61 73 85 97 109 121

0.12 UCL
mobile range (m)

1
0.09
1

0.06 (II)
C SL
__
0.03 MR

0.00 LCL=0
1 13 25 37 49 61 73 85 97 109 121
Observations

Fig. 3  Individual control charts (I) and mobile range (II) for error between tractor-sowing set passes
(α = 0.05). A/G automatic guidance, M/G manual guidance, UCL upper control limit, LCL lower limit of
control, CLS specific limit of control, x̄ average, MR mean of the mobile range. Black dots: observations
that do not go beyond the limits of control; Red dots: errors that can affect the quality of the operation (6M
factors) (Color figure online)

terms of smaller errors between the passes of the mechanised set, the probability of having
planted areas is greater, as the overlap rate between passes is reduced, ensuring that each
line has a spacing of 0.9 m.
In contrast, sowing with manual guidance and only using the experience of the opera-
tor for sowing operations, caused alignment errors greater than 0.90 m 19.67% of the time,
which is evidence that the variability of the process can be caused by interference from the
action of special causes, such as a failure in the machine direction, the action of the opera-
tor (manpower) in conjunction with the deregulation of line markers (machine), and the
lack of automatic guidance, which reduces the quality of the operation.
However, despite the quality of the operation being compromised from the perspective
of the SPC, the average error in sowing with manual guidance was satisfactory (54.6 mm).
The variability in the errors as the machine assembly moves reduces the quality of the
operation with manual guidance, in which there are maximum errors of 170 mm. The fre-
quency of occurrence of errors greater than the spacing of the crop reduces the area sown
due to the increase in distance between the lines. These results are interesting because
Holpp et al. (2013) did not find specific comparisons between manual and automatic guid-
ance during sowing operations.
The average error found in the present study when using an RTX signal in peanut
sowing was 29.4 mm. Thus, if the peanut sowing operation requires a precision up
to 900 mm (row spacing) the RTX can be used without risk of overlapping. Oksanen
(2015) found that autonomous vehicles using the RTK signal can perform cereal sowing

13
Precision Agriculture (2019) 20:840–856 849

with lateral error typically less than ± 100 mm. Thus, it is suggested that future peanut
sowing operations may also be done by autonomous vehicles using high-precision sat-
ellite signals, since Rizos et al. (2012) reported that the high-precision satellite signal
(RTX) can provide a maximum real-time error of 40 mm up to 95% of observations in
autonomous vehicles.
It is worth remembering that one of the advantages of using the RTX satellite signal is
that it dispenses with any initial investment in a base station to make the differential correc-
tion, which consequently reduces the cost of adoption of the technology. However, Lipin-
ski et al. (2016) emphasised that the choice of signal to be used depends on the precision
required by the operation.
When the quality of the sowing operation is evaluated (Fig. 3), the mechanised set
passes are accurately and precisely sown with automatic guidance (error < 38 mm and
standard deviation 17 mm), while the manual guidance was sowed with poor accuracy and
standard deviation 37 mm. This deduction stems from the greater variability observed in
both the individual values and the mobile range values for sowing with manual guidance,
which also presented 10% of the points outside the control limits with errors that exceed
the recommended spacing for the peanut crop. It is important to note that accuracy is only
affected by random errors in the measurement process, while precision is affected as well
by unknown errors or systematic errors (Cappellli et al. 2006).
The error between the passes of the mechanised set in the peanut arrangement was
unstable, with points above or below the UCL and LCL in the two treatments stud-
ied (Fig. 4a, b). In the M/G treatment, 9.15% (Fig. 4b) of the points were out of control,
whereas for the A/G digging (Fig. 4a), this percentage was 65.86%; however, mean errors
in this treatment were lower than those found in the M/G treatment.
It is worth noting that there is a considerable difference in the variability presented in
each process. When the automatic guidance was used, there was a reduction in variability
due to the lower mean (41.1 mm) in association with the standard deviation (44 mm), a fact
which reduces the limits of control and, consequently, reduces variability while increasing
the quality of the process.
It is known that errors during the displacement of the mechanised assembly can happen
to the left side or to the right side. In this sense, it can be observed that the mechanised set
remained within the specific limits of control (CLS = ± 38 mm) 70% of the time when the
automatic guidance was used. However, it is observed that in some moments (Observa-
tions 78–88 and 92–101), there is a consecutive increase in the alignment error between
the passes, exceeding even the UCL. However, among these sequences of observations, the
maximum error point (observation 85) represented 6.13% more than that established by the
signal provider (± 38 mm), which did not cause detrimental reductions in the quality of the
operation.
However, when automatic guidance was not used in the peanut digging operation, maxi-
mum errors of 1 m were recorded, as shown in the control chart at point 117 (Fig. 4b). The
occurrence of errors of this magnitude in the peanut digging operation is significant, as
the centre point of the digger–shaker–inverter will pass outside the area where the peanut
pods are found, leading to higher loss rates. However, if the operator is following the actual
alignment of plant rows then losses are likely to reduce.
An alternative to reduce very large errors in the peanut digging without using the auto-
matic guidance would be to include visible marks on the ground for the alignment of the
mechanised set; however, the peanut plant spreads and prevents observation of the soil.
Keller (2005) pointed out that many factors can affect visibility and therefore automatic
guidance systems considerably increase accuracy in alignment.

13
850 Precision Agriculture (2019) 20:840–856

error digger-shaker-inverter (m)

0.2
11

0.1 1111 1 1
1 1 1 1
1111
1
1 1 1 111 1 1
UCL
CSL
_
0.0 X (I)
CSL
1 111 1 11 11 1 LCL
1111 1
1 1
1
1 1
-0.1 1
1 1

11
1 16 31 46 61 76 91 106 121 136 151

0.16
1
Mobile range (m)

1
0.12
1 1

1 1
(II)
0.08 1
1 1
1
1
UCL
0.04 CSL
__
MR
0.00 LCL=0
1 16 31 46 61 76 91 106 121 136 151
Observation
(A)
error digger-shaker-inverter (m)

1.0
1
1
1 1 1
0.5 1
1
UCL
CSL
_
0.0
X
(III)
CSL
LCL
-0.5 1 1
1 1
1
1
-1.0
1
1 16 31 46 61 76 91 106 121 136 151

1.00 1
1
11
mobile range (m)

0.75
1 1
1 1
(IV)
0.50 1
UCL

0.25
__
MR
CSL
0.00 LCL=0
1 16 31 46 61 76 91 106 121 136 151
Observation
(B)

Fig. 4  Control charts individual (I) and mobile range (II) for error between digger–shaker–inverter passes
using automatic guidance (a) and individual control charts (III) and mobile range (IV) using manual guid-
ance (b) (α = 0.05). Black dots: observations that do not go beyond the limits of control; Red dots: errors
that can afect the quality of the operation (6M factors) (Color figure online)

13
Precision Agriculture (2019) 20:840–856 851

When evaluating the quality of the digging operation from the perspective of statisti-
cal control (Fig. 4), it is observed that the error between the passes has low precision
and accuracy, both with and without automatic guidance. As with the observations dur-
ing sowing, it was verified that there was greater variability in the latter of individual
values and the mobile range for the M/G digging, which presented a lower percentage
of points outside the control limits to the highest values of mean (32 mm) and standard
deviation (107.3 mm), making the operation lose quality.
The low precision and accuracy in the peanut digging operation is due to the inter-
action of the type of pilot (electric actuator) and the signal used (signal via satellite),
which consequently causes greater errors depending on the operation. This is mainly
because of the shorter response time of the electric actuator in the alignment correction
since it is located on the steering wheel and is therefore far from the guide wheel (front
axle).
At the moment of the digging operation (2 h p.m.), the ionosphere layer had minimum
scintillation values, Total Electron Content (TEC) of 38.92 and ionospheric index equal to
4 (index S4). The errors found in the field may have been affected by the amount of TEC in
the ionosphere, since the values of scintillation and ionospheric index are considered low,
according to Tiwari et al. (2011).
The TEC consists of the number of electrons present along the path traversed by the sig-
nal and is directly related to the magnitude of the errors on the GNSS (Rocha et al. 2017).
In addition, it is noted that the TEC is sensitive to variations as a function of the solar ioni-
zation flux, magnetic activity, sunspot cycle, season of the year, user location and direction
of the vector radius between the satellite and the receiver.
In general, Seeber (2003) points out that the use of GNSS signals using PPP as a posi-
tioning method can be affected by several factors, such as satellite errors (DOP satellites),
signal propagation (ionosphere delay and troposphere), multipath and loss of cycles) and
errors related to the receiver/antenna (clock errors and channel error). In addition, accord-
ing to Almeida and Dal Poz (2016), the ionospheric effect is the main factor that affects the
accuracy of the positioning of GNSS using PPP.
Regarding losses of the pods above soil, PVA, (Fig. 5), the operation remained within
the control limits regardless of the use of automatic steering, i.e. there is only natural vari-
ability as a result of the digging operation. However, there is a high variability in the pro-
cess of automatic orientation, which is confirmed in the mobile amplitude chart of this
variable and may be related to the high maturation of pods at the time of digging (> 80% of
maturation).
When the pods present a high degree of maturation, the inversion of the branches, car-
ried out by the vibratory belt of the excavator-agitator-inverter can increase the number
of pods that are detached from the plants, mainly due to the weakening of the gynophore
and the low volume of vegetable material on the vibratory belt. Therefore, it is essential
to maintain the correct regulation of the vibrating belt rotation to avoid the detachment of
pods during the inversion process. It is also necessary to develop methods to evaluate and
establish resistance values of the gynophores to the rupture, a main indicator of maturation.
Figure 6 shows the individual and mobile amplitude charts for PIA, which show the sta-
bility of the process in operations with the use of automatic guidance and instability with-
out automatic guidance. The occurrence of below-ground losses is linked to several factors,
such as soil type (Zerbato et al. et al. 2017b), gynophorous resistance at breakage (Ince and
Guzel 2003), digger–shaker–inverter settings (Bragachini and Peiretti 2008, and depth of
cut (Roberson 2009). In this study, it was noticed that the interference mainly stems from
the depth of work (Table 1) in which the high variability may have caused higher below

13
852 Precision Agriculture (2019) 20:840–856

A/G M/G

200
visible losses (kg ha-¹)

UCL
150

100 _
X (I)
50

0
LCL=0
1 7 13 19 25 31 37 43 49 55

160
mobile range (kg ha-¹)

120
UCL

80
(II)
40 __
MR

0 LCL=0
1 7 13 19 25 31 37 43 49 55
Observation

Fig. 5  Individual control charts (I) and mobile range (II) for visible losses. A/G automatic guidance, M/G
manual guidance, UCL upper control limit, LCL lower limit of control, x̄ average, MR mean of the mobile
range (α = 0.01). Black dots: observations that do not go beyond the limits of control; Red dots: errors that
can affect the quality of the operation (6M factors) (Color figure online)

A/G M/G
1
invisible losses (kg ha-¹)

300

200 UCL
(I)
100
_
X

0
LCL=0
1 7 13 19 25 31 37 43 49 55

1 1
240
mobile range (kg ha-¹)

180
(II)
UCL
120

60 __
MR
0 LCL=0
1 7 13 19 25 31 37 43 49 55
Observation

Fig. 6  Individual control charts (I) and mobile range (II) for invisible losses (pods below soil). A/G auto-
matic guidance, M/G manual guidance, UCL upper control limit, LCL lower limit of control; x̄ average, MR
mean of the mobile range (α = 0.01). Black dots: observations that do not go beyond the limits of control;
Red dots: errors that can affect the quality of the operation (6M factors) (Color figure online)

13
Precision Agriculture (2019) 20:840–856 853

ground losses in the M/G treatment. However, it is more appropriate to attribute this insta-
bility to machine and manpower acting together.
This assumption is because the digger–shaker–inverter used did not have a pantographic
system to copy the soil profile, since this requires a greater ability of the operator to main-
tain the depth desired for the digger–shaker–inverter and, consequently, to minimise the
variability of the process by reducing the PIA. Vellidis et al. (2014) stated that the operator
has difficulty with aligning the longitudinal axis of the tractor with the row of the crop,
which can increase levels of losses in digging. This is explained by the structure of the pea-
nut plant fully covering the soil surface, making visualisation of the peanut lines difficult.
Even though it was unstable, the M/G process obtained variability and mean PIA simi-
lar to the A/G process in relation to the gross productivity, with indices of 1.7% and 1.6%
of PIA, respectively. These values were considered low by studies that indicate losses rang-
ing from 3.1 to 47.1% in several experiments found in the literature (Santos et al. 2013;
Zerbato et al. 2017b; Cavichioli et al. 2014). According to Santos et al. (2016a), the use of
automatic guidance is one of the main factors peanut farmers must consider in increasing
the quality of the digging operation, leading to a reduction in below-ground losses.
The use of automatic guidance in peanut digging has a central objective to promote the
reduction of total losses. However, it was observed in this experiment that there was no
significant reduction when compared to manual guidance based on control charts (Fig. 7).
However, Ortiz et al. (2013) pointed out that the producer can obtain a higher financial
return from lower loss rates close to 11% when digging is performed with low soil water
contents even when using automatic guidance, since this factor can increase losses as men-
tioned previously. In the present study, at the time of harvesting, the average water content
in the soil was 25%, which caused the greatest variability of the total losses when using

A/G M/G
1
400
total losses (kg ha-¹)

300 UCL

200 _ (I)
X

100

LCL
0

1 7 13 19 25 31 37 43 49 55

1
300
mobile range (kg ha-¹)

1
200
UCL (II)

100
__
MR

0 LCL=0
1 7 13 19 25 31 37 43 49 55
Observation

Fig. 7  Individual control charts (I) and mobile range (II) for total losses of pods. A/G automatic guidance,
M/G manual guidance, UCL upper control limit, LCL lower limit of control, x̄ average, MR mean of the
mobile range (α = 0.01). Black dots: observations that do not go beyond the limits of control; Red dots:
errors that can affect the quality of the operation (6M factors) (Color figure online)

13
854 Precision Agriculture (2019) 20:840–856

automatic guidance. In contrast, when using manual guidance, the mean range of water
content in the soil was close to 20%. It is emphasized that harvesting was done at the same
time and same soil type.

Conclusion

The use of automatic RTX signal guidance was more efficient in maintaining the alignment
between tractor-sowing passes and digging operations when compared to manual guidance.
Alignment of operations (sowing and digging) using RTX with autopilots can increase
the harvest efficiency (higher productivity) but not reduce harvest losses when compared to
manual guidance.
The use of RTX signal and the electric pilot can be used in agricultural operations on a
peanut crop that require a precision of up to 38 mm.

Acknowledgements This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de
Nível Superior - Brasil (CAPES) - Finance Code 001.

References
Almeida, M. S., & Dal Poz, W. R. (2016). Precise point positioning and relative positioning with GNSS:
What most accurate currently? Bulletin of Geodetic Science Curitiba, 22(1), 175–195.
Baio, F. R. R., & Moratelli, R. F. (2011). Auto guidance accuracy evaluation and contrast of the operational
field capacity on the mechanized plantation system of sugar cane. Journal of the Brazilian Association
of Agricultural Engineering, 31(2), 367–375.
Bragachini, M. E., & Peiretti, J. M. (2008). En la eficiencia de cosecha de maní (On the efficiency of the
peanut harvest). Gacetilla de Prensa: 09/2008 março 2008. Retrieved September 3, 2018, from http://
www.cosec​haypo​stcos​echa.org/data/gacet​illas​/2008/20080​319_mani.asp (in Spanish).
Cappellli, N. L., Umezu, C. K., Silveira, A. C., & Garcia, A. P. (2006). Comparative performance among
GPS receivers. Journal of the Brazilian Society of Agroinformatics, 8(1), 63–77.
Cavichioli, F. A., Zerbato, C., Bertonha, R. S., Silva, R. P., & Silva, V. F. A. (2014). Pods losses during the
mechanical harvesting of peanuts. Journal of Agrarian Science, 42(3), 211–215.
CONAB. (2017). Monitoring of the Brazilian grain harvest, v.4 Harvest 2016/17-12° data collection, Bra-
sília, pp. 1–158 September 2017. Retrieved September 3, 2018, from http://www.conab​.gov.br/Olala​
CMS/uploa​ds/arqui​vos/17_09_12_10_14_36_bolet​im_graos​_setem​bro_2017.pdf (in Portuguese).
Easterly, D. R., Adamchuk, V. I., Kocher, M. F., & Hoy, R. M. (2010). Using a vision sensor system for per-
formance testing of satellite-based tractor auto-guidance. Computers and Electronics in Agriculture,
72(2), 107–118.
EMBRAPA - Brazilian Agricultural Research Corporation. (2013). Brazilian system of soil classification
(3rd ed., p. 353). Brasília: Embrapa.
Holpp, M., Kroulik, M., Kviz, Z., Anken, T., Sauter, M., & Hensel, O. (2013). Large-scale field evaluation
of driving performance and ergonomic effects of satellite-based guidance systems. Biosystems Engi-
neering, 116(2), 190–197. https​://doi.org/10.1016/j.biosy​stems​eng.2013.07.018.
Ince, A. & Guzel, E. (2003). Effects of gynophore breaking resistance on losses in mechanized peanut har-
vesting. In G. Quick (Ed.) International conference on crop harvesting and processing, ASAE Publica-
tion Number 701P1103e. St Joseph, MI, USA: ASAE.
Keller, J. (2005). Auto-Guidance-System-Effiziente Flächenbearbeitung, Diesel verbrauchsoptimierung,
Steigerung der Wirtschaftlichkeit (Efficient surface treatment, diesel consumption optimization,
increase in economic efficiency). In: Max-Eyth-Gesellschaft Fur Agrartechnik (Ed.) VDI-Berichte
1868-Landtechnik für Profis (pp. 75–80). Düsseldorf, Germany: VDI Verlag GmbH.
Koppen, W. (1948). Climatologia: Con um estudo de los climas de la Tierra. (Climatology: With a study of
the climates of the Earth) México: Fondo de Cultura Economico.
Kroulík, M., Kvíz, Z., Kumhála, F., Hula, J., & Loch, T. (2011). Procedures of soil farming allowing reduc-
tion of compaction. Precision Agriculture, 12(3), 317–333.

13
Precision Agriculture (2019) 20:840–856 855

Leidner, J. (2012). Precision farming payoff in peanuts. Southeastern Peanut Farmer, 50(5), 10–12.
Lipinski, A. J., Markowski, P., Lipinski, S., & Pyra, Paweł. (2016). Precision of tractor operations with
soil cultivation implements using manual and automatic steering modes. Biosystems Engineering, 14,
22–28.
Montgomery, D. C. (2009). Design and analysis of experiments (6th ed., pp. 179–268). Hoboken, NJ, USA:
Wiley.
Oksanen, T. (2015). Accuracy and performance experiences of four wheel steered autonomous agricultural
tractor in sowing operation. In L. Mejias, P. Corke, & J. Roberts (Eds.), Field and service robotics.
Springer tracts in advanced robotics (Vol. 105). Switzerland: Springer. https​://doi.org/10.1007/978-3-
319-07488​-7.
Oliveira, T. C., & Molin, J. P. (2011). Use of autopilots on citrus orchards establishment. Journal of the Bra-
zilian Association of Agricultural Engineering, 31(2), 334–342.
Ortiz, B. V., Balkcom, K. B., Duzy, L., Van Santen, E., & Hartzog, D. L. (2013). Evaluation of agronomic
and economic benefits of using RTK-GPS-based auto-steer guidance systems for peanut digging opera-
tions. Precision Agriculture, 14, 357–375.
Paixão, C. S. S., Santos, A. F., Voltarelli, M. A., Silva, R. P., & Carneiro, F. M. (2017). Times of efficiency
and quality of soybean crop mechanical operation in geometry functions of plots. Journal of the Bra-
zilian Association of Agricultural Engineering, 37(1), 106–115. https​://doi.org/10.1590/1809-4430-
eng.agric​.v37n1​p106-115/2017.
Perez-Ruiz, M., & Upadhyaya, S. K. (2012). GNSS in precision agricultural operations. In F. B. Elbahhar
& A. Rivenq (Eds.), New approach of indoor and outdoor localization systems (pp. 1–24). InTech:
London, United Kingdom.
Rizos, C., Janssen, V., Robert, C., & Grinter, T. (2012). PPP versus DGNSS. Geomatic. World, 20, 18–20.
Roberson, G. T. (2009). Planting, harvesting, and curing peanuts. In: D. L. Jordan, R. L. Brandenburg, A. B.
Brown, S. G. Bullen, G. T. Roberson, B. Shew & J. F. Spears (Eds.) Peanut information (pp. 131–148).
North Carolina Coop. Ext. Series AG-331. Raleigh, NC, USA.
Roberson, G. T., & Jordan, D. L. (2014). RTK-GPS and automatic steering for peanut digging. Applied
Engineering in Agriculture, 30(3), 405–409.
Rocha, R. S. M., Jerez, G. O., Brassarote, G. O. N., & Monico, J. F. G. (2017). Assessment of the iono-
spheric scintillation and different data time intervals effect on the precise point positioning in its on-
line versions. Revista Brasileira de Geomática, 5(2), 251–276.
Samohyl, R. W. (2009) Statistical quality control (1st ed.). São Paulo, Brazil: Elsevier: Elsevier Campus. v.
1. ISBN 978-85-352-3220-2.
Santos, A. F., Kazama, E. H., Ormond, A. T. S., Tavares, T. O., & Silva, R. P. (2016a). Quality of mecha-
nized peanut digging in function of the auto guidance. African Journal of Agricultural Research,
11(48), 4894–4901.
Santos, A. F., Silva, R. P., Corrêa, L. N., Borba, M. A. P., & Girio, L. A. S. (2016b) Accuracy and precision
monitoring through statistical process control. In: Brazilian Congress of Precision Agriculture-Con-
BAP, 2016, Goiânia. Anais Goiânia: SBEA, CD-Rom.
Santos, E. P., Silva, R. P., Bertonha, R. S., Noronha, R. H. F., & Zerbato, C. (2013). Productivity and losses
in the peanut on five different harvesting dates. Revista Ciência Agronômica, 44(4), 695–702.
Seeber, G. (2003). Satellite geodesy: Foundations, methods and applications (p. 586). Berlin, New York:
Walter de Gruyter.
Silva, E. L., & Gervásio, E. S. (1999). The use of TDR for moisture determination in different layers of a
dystrophic dusky red latossol. Revista Brasileira Engenharia Agrícola Ambiental, 3(3), 417–420.
Tavares, T. O., Damasceno, A. F., Voltarelli, M. A., Silva, R. P., & Furlani, C. E. A. (2017). Effective power
and hourly fuel consumption demanded by set tractor-coffee harvester in function of adequacy tractor
ballasting. Journal of the Brazilian Association of Agricultural Engineering, 37(4), 699–708. https​://
doi.org/10.1590/1809-4430-eng.agric​.v37n4​p699-708/2017.
Tey, Y. S., & Brindal, M. (2012). Factors influencing the adoption of precision agricultural technologies:
A review for policy implications. Precision Agriculture, 13, 713–730. https​://doi.org/10.1007/s1111​
9-012-9273-6.
Tiwari, R., Skone, S., Tiwari, S., & Strangeways H. J. (2011). 3WBMod assisted PLL GPS software receiver
for mitigating scintillation affect in high latitude region. IEEE. Retrieved September 26, 2018 from,
http://www.ursi.org/proce​eding​s/procG​A11/ursi/FG-4.pdf.
TRIMBLE RTX. (2017). Retrieved September 03, 2018 from http://www.trimb​le.com/Posit​ionin​g-Servi​ces/
Trimb​le-RTX.aspx.
USDA. (2018). United States Department of Agriculture. Peanut Area, Yield, and Production. Retrieved
June 10, 2018 from http://apps.fas.usda.gov/psdon​line/psdre​port.aspx?hidRe​portR​etrie​valNa​
me=BVS&hidRe​portR​etrie​valID​=918&hidRe​portR​etrie​valTe​mplat​eID=1.

13
856 Precision Agriculture (2019) 20:840–856

Vaccaro, G. L. R., Martins, J. C., & Menezes, T. M. (2011). Statistical analysis of the quality of tension lev-
els in electrical energy distribution systems. Production, 21, 539–552. https​://doi.org/10.1590/S0103​
-65132​01100​50000​47.
Vellidis, G., Ortiz, B., Beasley, J., Hill, R., Henry, H., & Brannen, H. (2014). Reducing digging losses by
using automated steering to plant and invert peanuts. Agronomy, 4, 337–348. https​://doi.org/10.3390/
agron​omy40​30337​.
Voltarelli, M. A., Silva, R. P., Cassia, M. T., Daloia, J. G. M., & Paixão, C. S. S. (2017). Quality of base
cutting in sugarcane using knives of different angles and coatings. Revista Ciência Agronômica, 48(3),
438–447.
Williams, E. J., & Drexler, J. S. (1981). A non-destructive method for determining peanut pod maturity.
Peanut Science, 8(2), 134–141.
Zerbato, C., Furlani, C. E. A., Ormond, A. T. S., Gírio, L. A. S., Carneiro, F. M., & Silva, R. P. (2017a). Sta-
tistical process control applied to mechanized peanut sowing as a function of soil texture. PLoS ONE,
12(7), e0180399. https​://doi.org/10.1371/journ​al.pone.01803​99.
Zerbato, C., Furlani, C. E. A., Silva, R. P., Voltarelli, M. A., & Santos, A. F. (2017b). Statistical control of
processes applied for peanut mechanical digging in soil textural classes. Journal of the Brazilian Asso-
ciation of Agricultural Engineering, 37(2), 315–322.

13
Precision Agriculture is a copyright of Springer, 2019. All Rights Reserved.