11
Manual GPS guidance system for agricultural vehicles C. Amiama-Ares*, J. Bueno-Lema, C. J. Alvarez-Lopez and J. A. Riveiro-Valiño Department of Agroforestry Engineering. University of Santiago de Compostela. Escuela Politécnica Superior. Campus Universitario. 27002 Lugo. Spain Abstract In this paper, the performance of a manual GPS guidance system to assist farming operations is evaluated. The distribution of granular fertilizer was simulated in order to discretize areas with excessive application of fertilizers and areas with fertilizer application rates below the intended rate. The path of travel followed by a tractor with the manual GPS guidance system was analysed and compared with a commercial parallel tracking system and without guidance assistance. In addition, the analysis evaluated how the use of manual GPS guidance systems improves the performance of field operations that require large distances between passes. Under the experimental conditions used, the best results were obtained using a commercial parallel tracking system but, for our purposes, small differences were observed between the results obtained with the commercial system and the results obtained with the developed manual GPS guidance system, getting pass-to-pass average error values of 0.26 and 0.73 m, respectively. The results obtained with both systems were significantly better than the results obtained when no guidance assistance was used. In our trials, area with appropriate fertilizer rate was clearly increased when guidance assistance was used. Values of area with correct fertilizer rate applied ranged between 87% with commercial parallel tracking and 59% without guidance assistance. The use of the manual GPS guidance system presented in this paper has proved sufficient to obtain good results for mechanical fertilizer spreading. Additional key words: parallel tracking; precision agriculture; spreader simulation. Resumen Sistema GPS de guiado manual para vehículos agrícolas En este trabajo se ha evaluado un sistema de asistencia al guiado manual para la realización de labores agríco- las. Se simuló la distribución de fertilizante granulado con el objetivo de discretizar áreas con excesiva cantidad de fertilizante y áreas con cantidades inferiores a las previstas. Se comparó la trayectoria seguida por un tractor uti- lizando el sistema GPS de asistencia al guiado manual con un sistema comercial de guiado paralelo, y sin asisten- cia al guiado. Nuestro análisis ha permitido evaluar las mejoras que estos sistemas suponen para la realización de labores que requieran elevadas distancias entre pasadas. En nuestras condiciones, los mejores resultados se obtu- vieron con un sistema comercial de guiado paralelo, si bien, considerando nuestro propósito, las diferencias fue- ron reducidas respecto a las obtenidas con el sistema de asistencia al guiado manual desarrollado, con valores me- dios de error pasada a pasada de 0,26 y 0,73 m, respectivamente. Los resultados obtenidos con ambos sistemas fueron significativamente mejores que los obtenidos cuando no se utiliza ningún sistema de asistencia. En nues- tros ensayos, el área con dosis adecuadas de fertilizante se incrementó de forma clara con la utilización del siste- ma de asistencia al guiado manual. Los valores de superficie con dosis correctas de fertilizante aplicado oscilaron entre el 87% con el sistema comercial de guiado paralelo y el 59% sin asistencia al guiado. Los resultados obteni- dos evidencian que el sistema de asistencia al guiado manual desarrollado es válido para la aplicación mecánica de fertilizantes. Palabras clave adicionales: agricultura de precisión; guiado paralelo; simulación de fertilización mecánica. * Corresponding author: [email protected] Received: 21-09-10. Accepted: 28-06-11. Abbreviations used: DGPS (differential global positioning system); GIS (geographical information system); GPS (global positio- ning system); PDOP (position dilution of precision); PT (parallel tracking); RMSE (root mean square error); RTK (real time kine- matic); STD (standard deviation of error); TCM (terrain compensation module). Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA) Spanish Journal of Agricultural Research 2011 9(3), 702-712 Available online at www.inia.es/sjar ISSN: 1695-971-X eISSN: 2171-9292

Manual GPS guidance system for agricultural vehicles

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Manual GPS guidance system for agricultural vehicles

Manual GPS guidance system for agricultural vehicles

C. Amiama-Ares*, J. Bueno-Lema, C. J. Alvarez-Lopez and J. A. Riveiro-ValiñoDepartment of Agroforestry Engineering. University of Santiago de Compostela.

Escuela Politécnica Superior. Campus Universitario. 27002 Lugo. Spain

Abstract

In this paper, the performance of a manual GPS guidance system to assist farming operations is evaluated. Thedistribution of granular fertilizer was simulated in order to discretize areas with excessive application of fertilizersand areas with fertilizer application rates below the intended rate. The path of travel followed by a tractor with themanual GPS guidance system was analysed and compared with a commercial parallel tracking system and withoutguidance assistance. In addition, the analysis evaluated how the use of manual GPS guidance systems improves theperformance of field operations that require large distances between passes. Under the experimental conditions used,the best results were obtained using a commercial parallel tracking system but, for our purposes, small differenceswere observed between the results obtained with the commercial system and the results obtained with the developedmanual GPS guidance system, getting pass-to-pass average error values of 0.26 and 0.73 m, respectively. The resultsobtained with both systems were significantly better than the results obtained when no guidance assistance was used.In our trials, area with appropriate fertilizer rate was clearly increased when guidance assistance was used. Values ofarea with correct fertilizer rate applied ranged between 87% with commercial parallel tracking and 59% withoutguidance assistance. The use of the manual GPS guidance system presented in this paper has proved sufficient to obtaingood results for mechanical fertilizer spreading.

Additional key words: parallel tracking; precision agriculture; spreader simulation.

Resumen

Sistema GPS de guiado manual para vehículos agrícolas

En este trabajo se ha evaluado un sistema de asistencia al guiado manual para la realización de labores agríco-las. Se simuló la distribución de fertilizante granulado con el objetivo de discretizar áreas con excesiva cantidadde fertilizante y áreas con cantidades inferiores a las previstas. Se comparó la trayectoria seguida por un tractor uti-lizando el sistema GPS de asistencia al guiado manual con un sistema comercial de guiado paralelo, y sin asisten-cia al guiado. Nuestro análisis ha permitido evaluar las mejoras que estos sistemas suponen para la realización delabores que requieran elevadas distancias entre pasadas. En nuestras condiciones, los mejores resultados se obtu-vieron con un sistema comercial de guiado paralelo, si bien, considerando nuestro propósito, las diferencias fue-ron reducidas respecto a las obtenidas con el sistema de asistencia al guiado manual desarrollado, con valores me-dios de error pasada a pasada de 0,26 y 0,73 m, respectivamente. Los resultados obtenidos con ambos sistemasfueron signif icativamente mejores que los obtenidos cuando no se utiliza ningún sistema de asistencia. En nues-tros ensayos, el área con dosis adecuadas de fertilizante se incrementó de forma clara con la utilización del siste-ma de asistencia al guiado manual. Los valores de superficie con dosis correctas de fertilizante aplicado oscilaronentre el 87% con el sistema comercial de guiado paralelo y el 59% sin asistencia al guiado. Los resultados obteni-dos evidencian que el sistema de asistencia al guiado manual desarrollado es válido para la aplicación mecánica defertilizantes.

Palabras clave adicionales: agricultura de precisión; guiado paralelo; simulación de fertilización mecánica.

* Corresponding author: [email protected]: 21-09-10. Accepted: 28-06-11.

Abbreviations used: DGPS (differential global positioning system); GIS (geographical information system); GPS (global positio-ning system); PDOP (position dilution of precision); PT (parallel tracking); RMSE (root mean square error); RTK (real time kine-matic); STD (standard deviation of error); TCM (terrain compensation module).

Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA) Spanish Journal of Agricultural Research 2011 9(3), 702-712Available online at www.inia.es/sjar ISSN: 1695-971-X

eISSN: 2171-9292

Page 2: Manual GPS guidance system for agricultural vehicles

Introduction

The reform of the Common Agricultural Policy (CAP)adopted by the European Union in 2003 has introduceda system of single farm payment and has decoupledproduction support. Council Regulation (EC) No1782/2003 of 29 September 2003 has introduced theconcept of cross compliance. Based on this concept,farmers will receive direct payments on condition thatthey adopt good agricultural practices and meet therequirements established by community legislation inthe areas of public, animal and plant health, and envi-ronment and animal welfare (cross compliance). Inaddition, growing food safely leads to the need to con-veniently document production processes (traceabi-lity). With regard to documentation, considerable pro-gress is being made to monitor equipment performanceand, consequently, agricultural production processes.

Increased technological sophistication of equipmentand increased field speeds require extra attention ofthe driver, which contributes to a decrease in guidanceaccuracy (Kaminaka et al., 1981). Moreover, with thegradual increase in equipment widths, it becomes in-creasingly difficult for the operator to determine visuallyfrom the cab whether path spacing is appropriate(Wilson, 2000). Such an effect is more pronouncedwhen f ield operations are performed by very widemachines and there are no visual indicators. Undersuch conditions, keeping the correct distance betweenconsecutive passes without guidance assistance beco-mes difficult (Stoll and Kutzbach, 1999). Improvingfertilizer distribution is beneficial to the environment,insofar as environmental problems caused by excessivefertilizer application rates are avoided and a moresustainable agriculture is achieved.

Widespread use of global positioning systems (GPS),such as GPS/GLONASS, has allowed for the develop-ment of many applications in a variety of areas relatedto agriculture and forestry, such as operating efficiency(Taylor et al., 2001), fleet management (Veal et al.,2001) or yield mapping (Shannon et al., 2002).

In the field of fertilizer application, GPS systemshave significantly contributed to the development ofguidance systems. Farmers are striving to adapt ferti-lizer distribution to the requirements established byboth legislation and the market. Currently, some commer-cial systems allow for data storage on a computermounted on the fertilizer spreader. Such data can befurther loaded into the farm management system com-puter program (Marquering and Scheufler, 2006).

Manual GPS guidance systems typically use a visualindicator to show where the vehicle is in relation to theintended path. The operator have to steer the equipment,but the lightbar or other indicator tells the driver whenand how much to correct. The efforts of several authorshave contributed to the development of manual guidancesystems (Gómez and Santana, 2007) and automaticguidance systems (Bell, 2000; Harbuck et al., 2006).Automatic guidance systems require driving an initialpass, but the system then steers the tractor through eachsubsequent pass. The operator still has to steer at turn-rows and around any obstacles, but passes are madewithout operator input. Automatic guidance reducesdriver fatigue and improves machinery performanceduring f ield operations insofar as the operator canfocus attention on the tasks performed (Kocher et al.,2000).

Among the benefits of combining fertilizer applica-tion systems with guidance systems are reduced fieldinput costs due to more even fertilizer application andreduced fertilizer application overlaps, skips or ratesbelow the desired rate (Thuilot et al., 2002; Batte andEhsani, 2006), which affect fertilizer use efficiency.Many authors have focused on testing the performanceof a number of devices, generally by evaluating guidanceerror, i.e. the deviation of the vehicle from the desiredpath (Taylor et al., 2004, Dunn et al., 2006). For preci-sion agriculture, Han et al. (2002) have suggested thatthe cross-tracking error determines overlaps and skips.For parallel tracking (PT) applications, the most impor-tant measure is the error measured between subsequentpasses. Ehsani et al. (2002) related mean error withoverlaps and skips, using root mean square error(RMSE) as a measure of safety and standard deviationof error (STD) as a measure of accuracy. They genera-ted regression lines to determine the path followed bythe machine. Wu et al. (2005) concluded that least squareregression was a suitable approach to f ind the bestfitting line through GPS data.

In addition to GPS signal accuracy, the ergonomicdesign of the system becomes particularly relevantwhen using guidance systems. Karimi et al. (2006)observed that the most accurate lightbar guidancesystems involved increased operator workload becausethe driver needed to make continuous steering adjust-ments. Ima and Mann (2004) suggested that the qualityof the operations performed using guidance displaysdepended on a number of ergonomic factors, such ascolour perceptibility, flash rate, display size, viewingdistance and height of placement of the display in the

Manual GPS guidance system for agricultural vehicles 703

Page 3: Manual GPS guidance system for agricultural vehicles

cab. Hence, ergonomic factors must be considered inthe design of the guidance display and in the placementof the display in the tractor cab.

Wilson (2000) reported two main limitations of GPSfor vehicle guidance: first, the difficulty to obtain theaccuracies required under adverse conditions, such asthe presence of buildings or trees. Second, inherenttime delay required for signal processing to determinethe location, which would become more relevant athigher field speeds. Nevertheless, the author suggestedthat such limitations would be overcome with thetechnological development of new equipment. Anotherlimitation of guidance systems is that the dynamicoperation of receivers is extremely variable in time,such that variable results are obtained depending ondate and time of day (Han et al., 2004).

In this paper, the performance of a low-cost manualGPS guidance system, developed by the AgriculturalMechanization Research Group at the University ofSantiago de Compostela and integrated into a dataacquisition system previously built by our researchgroup (Amiama et al., 2008a) and adapted to anagricultural tractor was evaluated. To accomplish thistask, the distribution of granular fertilizer was simula-ted, and a comparison was made between the resultsobtained for the uniformity of application using low-cost manual GPS guidance system and the results ob-tained with a commercial manual GPS guidance system.Besides, both systems were compared with fertilizerapplication without guidance assistance, the traditionalfertilizer application method in the study area. Essen-

tially, the paths of travel of a vehicle equipped withboth systems were compared by analyzing the expecteddistribution of the fertilizer rate applied in the fieldsand by discretizing areas with excessive application offertilizers and areas with fertilizer application ratesbelow the intended rate.

Material and methods

Characteristics of the equipment used

The tractor used for this study was a John DeereTM

6910S (John Deere Co., Moline, III). A Bogballe ET3200 (Bogballe Co., Uldum, DK) trailed twin-discfertilizer spreader was attached to the tractor duringthe tests (see Fig. 1). In the tests, fertilizer was notactually applied, but a theoretical or expected distributionpattern was used. By using a theoretical distribution,we removed the potential effects of a wrong spreadpattern of the fertilizer spreader on our results. Thisapproach allowed us to compare the error in guidanceunder the three systems of interest.

The tractor was equipped with a data acquisitionsystem developed for harvesters by the AgriculturalMechanization Research Group at the University ofSantiago de Compostela, (Amiama et al., 2008a),previously adapted to an agricultural tractor. The dataacquisition system was adapted by removing thesensors for capture of the parameters of harvesteroperation while maintaining the rest of the modules,

704 C. Amiama-Ares et al. / Span J Agric Res (2011) 9(3), 702-712

Figure 1. Tractor and trailed twin-disc fertilizer spreader used in trials (a) and location of GPS antennas at the top of the cab (b).

a) b)

Page 4: Manual GPS guidance system for agricultural vehicles

and by adapting the software for data collection andmanagement to the operations specific to the tractor.

An open source software application was developedin Visual Basic (Microsoft) using the MapObjects™2.3 ActiveX controls (Environmental Systems ResearchInstitute, Inc.), which allowed for the visualization ofthe path of travel through the field on the touch-screenmonitor of the computer installed in the tractor cab.Our manual GPS guidance system aims to maximizethe versatility of the equipment installed in the cab,such that many applications, among which guidance,remote location, route management or record ofmachine activity status, can be implemented using thesame hardware. In addition, equipment replication inseveral tractors does not involve the purchase of costlysoftware licenses for every implemented application.In fact, the only extra cost required if compare with thesystem developed for harvesters, is the cost of replacingthe GPS used in harvesters (C/A code GPS receiver)with a GPS that provides higher accuracy (± 300 mmaccuracy, 95% of the time, is considered sufficient forthe field operations tested). The GPS signal used wasthe free SF1 signal, which was captured with theStarFire™ iTC position receiver. The differential correc-tion signal used was the free SF1 signal owned byDeere & Company, which offers ± 300 mm accuracy.

Position data were recorded using the StarFire™ realtime kinematic (RTK) system. The system was compo-sed of a RTK Base Station located near the fields wherethe tests were performed, and a vehicle StarFire™ iTCreceiver. The purpose of the base station was to act asa local reference station. It transmitted correction signalsto the vehicle StarFire™ iTC receiver through the RTKradio. The vehicle StarFire™ receiver interpreted thecorrection signal coming from the base station. TheStarFire™ system provided a horizontal accuracy ofabout 2 cm. The position data were stored in the vehiclepanel computer installed in the tractor cab. This systemallowed the user to capture the actual path of the tractorfor each of the system tested.

The guidance application shows in the display ashaded strip on each side of the tractor path, whichreflects the treated area. The width of the strip of landshaded was defined by the user and was equivalent tothe distance between passes. Thus, the driver could seeon the monitor which areas had been treated and whichareas had been left untreated. Additionally, a virtualpath parallel to the actual path of travel was markedon the screen at a distance equivalent to the widthinitially defined by the user. This virtual line was used

as an additional reference to the shaded path and madetractor guidance easier for the driver.

The guidance application allows for PT with anytype of path (straight and curved) and is devised toconform to small, irregular shaped f ields where thedriver can freely choose the path of travel of the tractor.In small, irregular shaped fields, which are characte-rized by smaller field capacities (Amiama et al., 2008b),straight paths might cause a sharper decrease in theefficiencies of the field operations performed. Develo-ped system has implemented functions as automaticscreen sliding, switching of orientation viewpoint,different levels of zoom, nº available satellites and po-sition dilution of precision (PDOP).

The tractor was also equipped with a GreenStar™parallel tracking manual steering system that utilizedthree components, GreenStar display, mobile processorand StarFire™ iTC position receiver. The differentialcorrection signal used was the SF1. Filter configura-tions were set based on the manufacturer’s recommen-dations. StarFire iTC position receiver had integrateda terrain compensation module (TCM) for improvedguidance performance. TCM corrects for vehicle dy-namics such as roll on sideslopes, rough terrain orvarying soil conditions. Before performing the testscorrect information about differential global positio-ning system (DGPS) mounted direction, fore/aft valueand height must be introduced. TCM calibrations weredone with the vehicle on a hard, level surface and in acomplete stop (cab not rocking), TCM was calibratedfor 2D placing f irst the tractor in one direction andthen, after the first calibration, turning the vehicle 180°to face opposite direction and repeating the calibrationin this position. Figure 2 shows the displays viewed bythe driver with each guidance system.

In tests performed without guidance assistance, theonly visual reference used by the driver was the wheeltracks of the previous pass.

The antenna of the guidance systems was placed ontop of the tractor cab, at the centre of the machine. Theantenna of the RTK GPS rover unit was mounted pa-rallel to the antenna of the guidance systems at a dis-tance of at least 75 cm in order to reduce possible sig-nal interference, as shown in Figure 1b.

Experimental design

Field tests were conducted on ten fields on Gayoso-Castro farm (43° 06’ N, 7°27’W), in Castro de Ribeiras

Manual GPS guidance system for agricultural vehicles 705

Page 5: Manual GPS guidance system for agricultural vehicles

de Lea, 20 km north of Lugo, Spain. Site elevation was420 m, with a level slope class (dominant slope < 4%).The experimental design involved performing parallelpasses in the direction of the widest side of ten fieldsfor the three systems considered. For all tests, swathwidth was 24 m and operating speed was set at 8 kmh–1 because this is the usual speed for a fertilizer appli-cation rate of 400 kg ha–1. The selected fields were flat,without overhead obstructions in the immediate testarea and with a width that allowed for at least five passes.The area of the fields ranged 4.7 to 11.6 ha, and thetotal area of the f ields was 87 ha. The tests wereperformed in February 2008. To reduce the effect oferror from different sources, all the tests performed ona field were performed on the same date. In addition,the same driver was used for all the tests. The driverwas experienced in mechanical fertilizer spreading anddrove his own tractor.

Data were collected at 1 Hz frequency. For all thesystems tested, the driver defined an initial straightpath that was used as a reference for the following passes.The initial reference straight path was defined placingtwo marks in the field for driver orientation in this fistpass. The linear regression equation obtained from theRTK GPS points read at this first pass was used as areference to compare the other swaths in each field.This will be the reference to calculate next desiredpaths (function of the swath width). The adjusted pathwill be defined by the parallel path, parallel to the re-ference pass, who better adjust to the RTK DGPSmeasurements, as shown in Figure 3. First, the driver

operated the tractor without using any guidancesystem, such that the path from the previous test couldnot be used and consequently bias the test.

Data analysis

Swath uniformity

There are two key aspects in the assessment of theperformance of fertilizer spreaders: accuracy of ferti-lizer rates and uniformity in application. Accuracy isdefined as the capability of applying a fertilizer rateat a given location, and uniformity is defined as theevenly distributed application of a fertilizer rate. To

706 C. Amiama-Ares et al. / Span J Agric Res (2011) 9(3), 702-712

Figure 2. Display visualization for GreenStar™ Parallel Tracking system (a) and for the software application developed (b).

b)a)

Figure 3. Layout of the test and definition of the mean cross-track error (X

_T_

E__

i).

Reference pass

RTK DGPS measurements Adjusted path Desired path

X_T_E_

i–1 X_T_E_

i

Page 6: Manual GPS guidance system for agricultural vehicles

achieve the greatest distribution uniformity, distancebetween passes must be constant.

For each pass, cross-track error (XTE) for a point j,is defined as the distance, measured perpendicular tothe direction to the travel, between his true position,calculated from the data provided by RTK GPS, andthe desired path.

Pass accuracy can be calculated with the use of twoexpressions: mean cross-track error (X

_T_

E__

i) and rootmean square error (RMSi).

[1]

[2]

where Ni = total number of measurement points for theith pass, XTEij = the XTE at jth point for the ith pass,X_T_

E__

ij = the mean of the XTE for the ith pass, and RMSi

= the RMS error for the ith pass.But these expressions translate the error from a pass

to the next passes. To avoid this problem pass-to-passaccuracy must be definite. For this purpose, pass-to-pass average error (Epi) and pass-to-pass root meansquare error (spi

2) were used.

[3]

[4]

where Epi = the pass-to-pass average error for the ithpass, and spi

2 = the pass-to-pass root mean square errorfor the ith pass.

Following Ehsani et al. (2002), the curved segmentsand the points at the end and at the beginning of eachswath were removed from the original raw data to eva-luate the straight-line accuracy.

Statistical processing was performed by using thevariable Epi, which was considered most representativeof the performance of each guidance system tested.

In addition to the comparison of the three systemstested, the experimental design allowed us to determinewhether the characteristics of the different fields affec-ted the results. In principle, the shape or size of fieldsshould not affect results when some guidance systemwas used. However, the shape, size and relief of a fieldcould influence the results obtained when no guidanceassistance system was used.

In order to assess the effects of field characteristicson the results, each field was considered as a block by

applying a randomized block design. Such a designwas used to assess whether considering fields as blockfactors was useful in reducing the size of experimentalerror. Otherwise, using a blocked design would becounterproductive insofar as it would generate weakerhypothesis tests and wider confidence intervals than afully randomized design.

Given that the number of readings within each blockcould be not equivalent for the different treatments be-cause of the different performance of the guidancesystems tested and the different number of readingsobtained for different blocks due to differing field shapesand areas, we worked with the mean values of readings(xij) by block (j) for each treatment (i) as response va-riable. Accordingly, the response to treatment i forblock j was computed from Eq. [5].

[5]

The assumptions of homoscedasticity, normality andindependence of results are required to obtain valid re-sults from the analysis of variance. With such a trans-formation, the assumption of homoscedasticity cannotbe met, but a balanced design is achieved. As shownby Scheffé (1999), the assumption of homoscedasticityof the response variable can be relaxed (results areapproximately valid) in balanced designs if Eq. [6] ismet, as in our test.

[6]

If using a blocked design was not useful, a one-wayexperimental design with three treatments was consi-dered and all the values of the test were used.

The Epi values obtained were compared accordingto ANOVA F-test. A Tukey’s test was performed in thecases in which significant values were obtained. TheSPSS computer package was used to solve the statis-tical tests.

Fertilizer rates applied

To analyze the uniformity of the spatial distributionof fertilizer, fertilizer application was simulated byusing a geographic information system (GIS), namelythe ArcMap application (ESRI, 2005), and followinga procedure similar to that used by Lawrence and Yule(2007). The distribution of the fertilizer rate appliedto the fields was obtained by taking as a reference the

max[var(Yij)]

min[var(Yij)]< 3

Yij =∑Xijnij

spi2 = Ni

1 ∑Ni

j =1

(XTEij – XTEi)2

Epi = (XTEi – XTEi–1)

RMSi = Ni

1 ∑Ni

j =1

(XTEij)2

XTEi = ∑Ni

1Ni

j =1

XTEij

Manual GPS guidance system for agricultural vehicles 707

Page 7: Manual GPS guidance system for agricultural vehicles

path of travel of the tractor through test f ields andintroducing the transverse distribution diagram of thefertilizer spreader for granular fertilizer (15-15-15). Awide range of colours was used to visually identifyareas with the intended fertilizer rate, areas with ex-cessive fertilizer rates and areas with fertilizer ratesbelow the intended rate. The deviation range was deter-mined for fertilizer rates above or below the intendedrate. ArcMap allowed us to accurately obtain the theo-retical fertilizer rates applied at every point of the field.The fertilizer rate chosen was 400 kg ha–1. Deviationsfrom that rate were considered correct for values of upto ± 10% of the intended fertilizer rate.

Results and discussion

Swath uniformity

Simple visual inspection of the paths of travel gene-rated by the two systems allowed us to predict betterperformance of the fertilizer spreader in the fields wherea guidance system was used than in the fields whereno guidance system was used (Fig. 4). The preliminaryvisual analysis of results suggested that the performan-ce obtained with the John Deere Parallel Trackingsystem was better than the performance obtained withthe guidance system developed by our research group.

708 C. Amiama-Ares et al. / Span J Agric Res (2011) 9(3), 702-712

Figure 4. Comparison of paths (a) and fertilizer rates (b) for the three systems tested, PT (Parallel Tracking), SA (Software Ap-plication) and WA (Without Assistance).

N

S

W E

N

S

W E

N

S

W E

WA PT SA

PT

SA

< 200 281-360 441-520

201-280 361-440 521-600

> 601

WA

PT

WA

SA

a)

b)

Page 8: Manual GPS guidance system for agricultural vehicles

As explained in previous section “Swath unifor-mity”, we first assessed whether considering the fieldas a block factor was useful in reducing experimentalerror. A two-way ANOVA F-test was performed consi-dering a block factor (f ield) and a treatment factor(Epi). The test revealed that the block factor was notsignificant (p = 0.05). Such a result could be justifiedby the large size, regular shape and flat relief of fields.Under different experimental conditions (narrow fieldswith rough relief, etc.), the field factor could largelyinfluence the results obtained when no guidance assis-tance was used.

Table 1 shows the results obtained for a singletreatment factor (guidance system). To determine X

_T_

E__

and Epi, the absolute values of the deviations wereused.

The numerical values obtained confirmed the resultsshown in Figure 4. By analyzing mean cross-track error(X

_T_

E__

) and pass-to-pass average error (Epi), the PTtreatment yielded lower deviations between the targetpath and the actual path of travel obtained with othersystems. In addition, the fit of the path to the regressionline defined by the data points for the virtual path (pa-rallel to the initial path) was better for the PT system,i.e. the paths were straighter when using PT than whenusing the other systems. The ANOVA F-test performedrevealed differences among systems. Tukey’s test didnot show significant differences at 0.05 level between

the two systems that used a guidance system. Yet, theresults obtained for each treatment would be differentat 0.01 significance level. These results are in agreementwith the results reported by other authors (Stoll andKutzbach, 1999; Wilson, 2000) and suggest that it isdifficult to keep the required distance between passeswith large implement widths without guidance assistance.

For the differences observed between the two gui-dance systems, results appear to suggest that the PTsystem performs better than the guidance system deve-loped by our research group. Such differences can beattributed to the following causes:

— The screen refresh was too high. Owing to sucha high refresh rate, slight steering wheel movementsmade by the driver were quickly transferred to thescreen, thus demanding great guidance effort from thedriver, who tended to make steering adjustments conti-nuously. These results are in agreement with the resultsreported by Ehsani et al. (2002) and Karimi et al. (2006)for tests performed with lightbar guidance systems.

— The manner in which guidance was displayed onthe screen by shading a strip of land on both sides ofthe path of travel produced a tendency to overlap inexcess because the driver focused on not leaving anyarea unshaded between passes rather than on followingthe centre of the path indicated by the line.

Figure 5 graphically represents the values of Epi asabsolute values and shows a better performance of thePT system.

For a better interpretation of the results shown inFigure 5, Table 1 shows the values of Epi at differentpercentiles. These values reveal a better performanceof the PT system, with an Epi value below 0.30 m for75% of the passes. By using our guidance system, the

Manual GPS guidance system for agricultural vehicles 709

Table 1. Mean values (m) obtained, cumulative frequenciesof Epi (m) and mean values of Epi (m) considering overlapsand skips, for each of the systems tested

Without Parallel Softwareassistance tracking application

Mean values (m)

X_T_

E__

4.89 0.54 1.85RMS 5.36 0.71 2.23Epi 2.62a 0.26b 0.73b

spi2 4.42 0.28 1.26

Cumulative frequencies of Epi (m)

25% 0.80 0.07 0.2550% 2.17 0.17 0.4875% 3.78 0.30 0.84100% 8.51 1.49 3.15

Mean values of Epi (m)

Epi for overlaps –2.60 –0.27 –0.78Epi for skips 2.63 0.25 0.42

Different letters following figures in the same row suggest sig-nificantly different data (α = 0.05).

0

20

40

60

80

100

120

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0Epi (m)

(%)

WA PT SA

Figure 5. Comparison of pass-to-pass average errors for thethree systems tested, WA (Without assistance), PT (ParallelTracking) and SA (Software Application).

Page 9: Manual GPS guidance system for agricultural vehicles

value of Epi at the 75th percentile amounted to 0.84.This value would be at the limit of suitability but wouldsubstantially improve fertilizer distribution withoutguidance assistance; the most frequent method in ourlocality, insofar as such a high level of accuracy wasobtained only in 25% of the passes performed withoutguidance assistance.

Fertilizer distribution and fertilizer ratesapplied

Figure 6 plots the percentage of passes for whichthe correct distance between passes was kept againstthe percentage of passes with overlaps or skips. De-viations from the intended path resulted in fertilizerexcess or deficit as compared to the intended fertilizer

rate. Taking into consideration that the accuracy of theGPS used for guidance was ± 30 cm, the passes definedas correct were the passes performed at a distance ofless than 30 cm from the desired path. For the PT gui-dance system, the number of correct passes prevailedover passes with overlaps or skips, which were verysimilar. The percentage of correct passes withoutguidance assistance was much lower. By using theguidance system developed by our research group, thenumber of correct passes increased compared to opera-ting the applicator without any guidance. However, avery high percent passes with overlaps was observed.Passes with overlaps result in excessive fertilizer appli-cation.

Table 1 summarizes the mean values of Epi discrimi-nated according to overlaps and skips. These valuessuggest that not using any guidance system results inlarger mean errors and, consequently, in an increase inthe areas with fertilizer application rates above or belowthe intended rate, which is in agreement with theresults reported by many authors (Stoll and Kutzbach,1999; Thuilot et al., 2002; Batte and Ehsani, 2006).

Figure 4b shows the results of the simulation perfor-med. Reddish shades suggest fertilizer rates higherthan the intended rate whereas greenish shades suggestfertilizer rates below the intended rate (400 kg ha–1

± 10%). The best results for uniformity of fertilizerdistribution were achieved using the PT system.

Figure 7 shows the distribution of the fertilizer ratesapplied in the set of f ields tested for each guidancesystem. The proportion of fertilizer rates below 350 kgha–1 is justified due to the presence of field bounda-

710 C. Amiama-Ares et al. / Span J Agric Res (2011) 9(3), 702-712

Figure 6. Comparison of paths for the three systems tested, WA(Without assistance), PT (Parallel Tracking) and SA (SoftwareApplication).

80

70

60

50

40

30

20

10

0

Correct

WA PT SAGuidance system

(%)

Skips Overlaps

WA

6%7%

13%

59%

9%5% 1%

PT

3%4%

5%

87%

0%0%

1%

SA

5%6%

5%

77%

5%2% 0%

< 200 201-280 281-360 361-440 441-520 521-600 > 600

Figure 7. Percent area with correct fertilizer rates, excessive rates or rates below the intended rate for the three systems tested, WA(Without assistance), PT (Parallel Tracking) and SA (Software Application) (ranges in kg ha–1).

Page 10: Manual GPS guidance system for agricultural vehicles

ries, where the fertilizer rate applied is usually below300 kg ha–1.

When the PT system was used, the correct fertilizerrate was applied to 87% of the area and only 1% of thearea received excessive fertilizer rates. The free SF1differential correction signal, which is the least accuratesignal offered by John Deere, proved sufficient to per-form fertilizer application operations. When f ieldoperations were performed without guidance assistance,59% of the area received the correct fertilizer rate,whereas 15% of the area received excessive rates.Intermediate results were obtained with the guidancesystem developed by our research group, with 7% ofover fertilized area. These results demonstrate thatusing guidance systems considerably reduces pollutionproblems caused by poor fertilizer distribution becauseareas with excessive fertilizer rates are minimized byusing guidance assistance.

Our results are in agreement with the results repor-ted by Thuilot et al. (2002). Tests performed usingguidance systems showed a more even fertilizer distri-bution and a decrease in areas with fertilizer rateshigher than the intended rate, skips or fertilizer ratesbelow the intended rate. These results match the resultsshown in Figure 6. In Figure 6, the guidance systemdeveloped by our research group showed the highestpercentage of path with overlaps. However, the devia-tion observed in these areas was lower than the devia-tion observed when working without guidance assis-tance (see Table 1). Despite the lower percent of pathwith overlaps, the percent area affected by over-laps without guidance assistance was higher than thepercent area affected with the guidance system de-veloped because of the higher pass-to-pass deviationsobserved. Figure 8 provides a graphical summary ofthe results obtained in this study. The improvementsachieved in fertilizer distribution by using guidancesystems are expected to prompt a better crop response,increased economic benefits and reduced environmen-tal impact.

The small differences found between the two guid-ance systems tested, with better results for the commer-cial guidance system, do not justify the greater invest-ment required to purchase a commercial system, ofabout € 3,400, as compared to the € 1,600 investmentrequired to replace the GPS used in harvesters (lowaccuracy) by a system that provides higher accuracyfor tractor guidance. In addition, our solution preventsthe duplication of monitors in the tractor cab, andavoids space and visibility reduction.

Conclusions

Based on this study in which we have tested acommercial GPS guidance system, a low-cost GPSguidance system developed in this study, and fertilizerapplicator operation without guidance in ten differentfields, we reached the following conclusions:

— Using guidance systems for granular fertili-zer distribution improved uniformity of fertilizerapplication. Although the improvement is higher forthe PT commercial system, the differences obser-ved between commercial PT and the software appli-cation developed by our research group were not signi-ficant.

— The free SF1 differential correction signal,which is the least accurate signal offered by John Deere,proved suff icient to perform fertilizer applicationoperations.

— With regard to the fertilizer rates applied, thebest distributions were obtained with the commercialPT system. When no guidance system was used, theareas with fertilizer application rates higher than theintended rate and the areas with rates below theintended rate tended to be equivalent. In contrast, whenguidance systems were used, areas with fertilizer ratesbelow the intended rate tended to be larger than areaswith excessive fertilizer application. The use of GPSguidance systems reduces the area where excess ferti-lizer is applied, which contributes to a more sustainableagriculture that is more environmentally friendly. Thegreater investment required does not justify the betterresults obtained with the commercial GPS guidancesystem compared to the low cost system developed inthis study.

Manual GPS guidance system for agricultural vehicles 711

Figure 8. Percent area treated with the different ranges of fer-tilizer rates considered (in kg ha–1), for the three systems tes-ted, WA (Without assistance), PT (Parallel Tracking), and SA(Software Application).

100908070605040302010

0

(%)

< 200 201-280 281-360 361-440 441-520 521-600 > 600Fertilizer rates (kg ha–1)

WA PT SA

Page 11: Manual GPS guidance system for agricultural vehicles

Acknowledgements

Financial support for the research was provided bythe co-operative “Os Irmandiños S.C.G.” and the “Di-rección Xeral de Investigación e Desenvolvemento daXunta de Galicia” under project no. PGIDIT03RAG14E.

References

AMIAMA C., BUENO J., ÁLVAREZ C.J., PEREIRA J.M.,2008a. Design and f ield test of an automatic dataacquisition system in a self-propelled forage harvester.Comput Electron Agric 61(2), 192-200.

AMIAMA C., BUENO J., ÁLVAREZ C.J., 2008b. Influenceof the physical parameters of fields and of crop yield onthe effective f ield capacity of a self-propelled forageharvester. Biosyst Eng 100(0), 198-205.

BATTE M.T., EHSANI M.R., 2006. The economics ofprecision guidance with auto-boom control for farmer-owned agricultural sprayers. Comput Electron Agric53(1), 28-44.

BELL T., 2000. Automatic tractor guidance using carrier-phase differential GPS. Comput Electron Agric 25(1-2),53-66.

DUNN P.K., POWIERSKI A.P., HILL R., 2006. Statisticalevaluation of data from tractor guidance systems. PrecisAgric 7(0), 179-192.

EHSANI M.R., SULLIVAN M., WALKER J.T., 2002. A method of evaluating different guidance systems. ProcASAE Ann Int Meeting, Chicago, IL, USA. Paper No.021155.

ESRI, 2005. ArcMap 9.1. ESRI, Redlands, CA, USA.GOMEZ J., SANTANA J., 2007. Agroguía, dispositivo de

asistencia al guiado basados en GPS, sobre plataformahardware Pocket PC [Device for guidance assistancebased in GPS, over a Pocket PC hardware platform]. Agri-cultura-revista agropecuaria, enero, 62-70. [In Spanish].

HAN S., ZHANG Q., NOH H., 2002. Kalman filtering ofDGPS positions for a parallel tracking application. TASABE 45(3), 553-559.

HAN S., ZHANG Q., NOH H., SHIN B., 2004. A dynamicperformance evaluation method for DGPS receivers underlinear parallel-tracking applications. T ASABE 47(1),321-329.

HARBUCK T.L., FULTON J.P., McDONALD T.P., BRODBECKJ., 2006. Evaluation of GPS autoguidance systems overvarying time periods. Proc ASAE Ann Int Meeting, Port-land, OR, July 9-12, Paper Nº 061042.

IMA C.S., MANN D.D., 2004. Ergonomic concerns withlightbar guidance displays. J Agric Safe Health 10(2), 91-102.

KAMINAKA M.S., REHKUGLER G.E., GUNTZEL W.W.,1981. Visual monitoring in a simulated agricultural ma-chinery operation. Hum Factors 23(2), 165-173.

KARIMI D., MANN D.D., EHSANI R., 2006. A new me-thodology for evaluating guidance systems for agriculturalvehicles. Proc Canadian Society for Bioengineering/ Société Canadienne de Génie Agroalimentaire et de Bio-ingénierie, Annual Conference. Edmonton, Alberta, PaperNº 06-148.

KOCHER M.K., SMITH M.B., GRISSO R.D., BASHFORDL.L., 2000. Performance test of three-point moun-ted implement guidance systems. Appl Eng Agric 16(0),595-603

LAWRENCE H.G., YULE I.J., 2007. A GIS methodology tocalculate in-field dispersion of fertilizer from a spinning-disc spreader. T ASABE 50(2), 379-387.

MARQUERING J., SCHEUFLER B., 2006. Mineral ferti-lizer. In: Yearbook agricultural engineering 2006 (HarmsH., Meier F., eds). KTBL; VDMA Landtechnik; VDI-MEG. Münster: Landwirtschaftsverlag GmbH.

SCHEFFÉ H., 1999. The analysis of variance. Wiley-Inter-science. 447 pp.

SHANNON K., ELLIS C., HOETTE G., 2002. Performanceof “low-cost” GPS receivers for yield mapping. ASAEpaper Nº 021151. St Joseph, MI, USA.

STOLL A., KUTZBACH H.D., 1999. Concept of a guidancesystem for agricultural vehicles. Proc 2nd Europ Conf onPrecision Agriculture (Stafford J.V., ed). Sheffield Acade-mic Press, UK. pp. 825-835.

TAYLOR R.K., SCHROCK M.D., STAGGENBORG S.A.,2001. Using GPS technology to assist machinery mana-gement decisions. Proc Mid-Central Conference of theASAE, St Joseph, MO, USA, March 30-31, Paper NºMC01-204.

TAYLOR R.K., SCHROCK M.D., BLOOMFIELD J., GORAG., BROCKMEIER G., BURTON W., CARLSON B.,GATTIS J., GROENING R., KOPRIVA J., OLEEN N.,NEY J., SIMMELINK C., VONDRACEK J., 2004. Dynamic testing of GPS receivers. T ASABE 47(4), 1017-1025.

THUILOT B., CARIOU C., MARTINET P., BERDUCATM., 2002. Automatic guidance of a farm tractor relyingon a single CP-DGPS. Auton Robot 13(0), 53-71.

VEAL M.W., TAYLOR S.E., MCDONALD T.P., MCLEMORED.K., DUNN M.R., 2001. Accuracy of tracking forestmachines with GPS. T ASABE 44(6), 1903-1911.

WILSON J.N., 2000. Guidance of agricultural vehicles – ahistorical perspective. Comput Electron Agric 25(0), 3-9.

WU C., AYERS P.D., ANDERSON A.B., 2005. Influence oftravel directions on the GPS dynamic accuracy for vehicletracking. Proc ASAE Ann Int Meeting, Trampa, FL, USA,July 17-20, Paper Nº 051088.

712 C. Amiama-Ares et al. / Span J Agric Res (2011) 9(3), 702-712