Introduction
⌅It
is well proven that farm mechanization is essential for enhancing
farmer’s income as well as reducing human drudgery. Large and medium
farmers are capable of owning tractors and other big machines, and
hence, their farms are mechanized. However, small and marginal farmers
rely on custom hiring services cultivating more than 80% of the total
cultivable area in India. In such scenarios, custom hiring serves as the
only viable means to promote mechanization, particularly in India. The
practice of providing or availing machine services on a payment basis is
called “Custom hiring service”. Custom hiring services are provided by
state agro-industries corporations, government agencies, cooperative
societies, and private machine owners. In recent years, many researchers
(Thomson et al., 2014Thomson
SJ, Huang Y, Smith LA, 2014. Portable device to assess dynamic accuracy
of global positioning system (GPS) receivers used in agricultural
aircraft. Int J Agric Biol Eng 7(2): 68-74. https://doi.org/10.3965/j.ijabe.20140702.008
; Kaivosoja & Linkolehto, 2015Kaivosoja
J, Linkolehto R, 2015. GNSS error simulator for farm machinery
navigation development. Comput Electron Agric 119: 166-177. https://doi.org/10.1016/j.compag.2015.10.021
; Pranav et al., 2016Pranav PK, Phukan Y, Saha B, 2016. Computer program for cost estimation of agricultural machines. Agric Eng 1: 1-10.
) have explored ways to make custom hiring centers profitable and easily accessible to farmers. Pranav et al. (2016)Pranav PK, Phukan Y, Saha B, 2016. Computer program for cost estimation of agricultural machines. Agric Eng 1: 1-10.
developed a computer program for cost estimation of agricultural
machines as well as break-even analysis for the custom hiring center to
prevent losses in custom hiring centers. Jayawant & Joshi (2018)Jayawant
YA, Joshi N, 2018. Modeling operations of a custom hiring center using
agent based modeling and discrete event simulation. Proc III
Asia-Pacific Conf on Complex Systems Design & Management, CSD&M
Asia, pp: 13-24. http://dx.doi.org/10.1007/978-3-030-02886-2_2
also simulated the operations of custom hiring center (CHC) under different scenarios to estimate its profitability. Daum et al. (2020)Daum
T, Villalba R, Anidi O, Mayienga SM, Gupta S, Birner R, 2020. Uber for
tractors? Opportunities and challenges of digital tools for tractor hire
in India and Nigeria. World Dev 144:1-15. https://doi.org/10.1016/j.worlddev.2021.105480
developed a mobile application for hiring tractors
on demand, similar to taxi booking services through mobile applications.
Although these studies were conducted to promote custom hiring
services, none of them quantified the amount of work done by the machine
for which it was hired. There is no technology-based system to
calculate custom hiring charges immediately after completing the work,
similar to application-based taxi services. In agricultural services,
the hiring charges are calculated based on the area covered in the field
as well as time taken. Therefore, this project was undertaken to
develop a mobile based application that records the machine’s operation
using the Global Positioning System (GPS) while also segregating and
calculating the area of each plot.
Global Positioning System (GPS)
is a technology used to know the location of a tractor or any moving
vehicle at a given time. GPS is a U.S. government-owned Radio Navigation
Satellite System operated by the U.S. Space Force. It is one of the
Global Navigation Satellite System (GNSS) that provides time information
and geolocation to GPS receivers anywhere on Earth, where visibility to
at least four satellites remains unobstructed. Researchers have
successfully utilized GPS for various agricultural applications. Omrani et al. (2013)Omrani
A, Shiekhdavoodi MJ, Shomeili M, 2013. Determine Sugarcane harvester
field efficiency using global positioning system (GPS) data. Elixir
Agric 56: 13260-13263.
determined the field efficiency of sugarcane harvester by using GPS data. Stoll et al. (2003)Stoll A, Stafford J, Werner A, 2003. Automatic operation planning for GPS-guided machinery. Precis Agric 3: 657-664. https://doi.org/10.3920/9789086865147_101
developed an automatic operation planning system for GPS-guided machinery. Maurya et al. (2012)Maurya
K, Singh M, Jain N, 2016. Real time vehicle tracking system using GSM
and GPS technology-an anti-theft tracking system. Int J Electron Comput
Sci Eng 1(3): 1103-1107.
designed a real time, vehicle tracking system by using GPS and Global System for Mobile Communications (GSM). Xiang et al. (2016)Xiang
M, Wei S, Zhang M, Li M, 2016. Real-time monitoring system of
agricultural machinery operation information based on ARM11 and GNSS.
IFAC-Pap 49: 121-126. https://doi.org/10.1016/j.ifacol.2016.10.023
developed an ARM11 and GNSS based monitoring system
for measuring the information related to working condition of
agricultural machinery in real time. Inoue et al. (2009)Inoue
K, Nii K, Zhang Y, 2009. Tractor guidance system for field work using
GPS and GYRO: Tractor guidance system for field work using GPS and GYRO.
Proc Int Sci Conf Energy Eff Agric Eng, Oct 1. pp: 280-295.
developed a differential GPS tractor guidance system (DGPS) for
monitoring tractor driving to track target lines in the field for
sowing, ploughing, spraying, and harvesting. The primary aim of this
research was to develop a mobile-based application that utilizes GPS to
record the machine's operation. The application was designed to
segregate different fields where the machine was operated and calculate
the area of operation.
Materials and methods
⌅A mobile application named “Geolocation” was developed to record real-time geolocations at desired time intervals using the inbuilt GPS of the smartphone. The application uses different algorithms to segregate machine operations across different fields and calculate the area of each segregated field.
Development of mobile application
⌅The mobile application developed for this study had distinct front-end and back-end functionalities. The front-end/ user interface (UI) was designed using, React.js (https://react.dev/), a JavaScript library known for building responsive and intuitive user interfaces for web and native applications. The middle layer, responsible for handling Representational State Transfer Application Programming Interface (REST API) calls, was built with Express.js (https://expressjs.com/) a minimal and flexible Node.js mobile and web application framework that facilitates smooth communication between the front-end and back-end. The server-side logic, including all back-end operations, was implemented using Node.js, ensuring efficient processing and data management. The database used was MongoDB (https://www.mongodb.com/), a widely used Not only Structured Query Language (NoSQL) document-oriented database that securely stores the application data and enabled scalable data management. The mobile application itself was developed using React Native, enabling seamless data transfer and compatibility with smartphones. Specifically for Android devices, the application was built using MongoDB, Express, React, Node.js (MERN) stack. The application plotted data using Scalable Vector Graphics (SVG) to provide clear and detailed visualizations. This application was classified as a native application and could be installed through an Android Package Kit (APK) file. The programming language used for the application's development was JavaScript. The app securely stores data in the cloud, ensuring seamless access and data integrity. Additionally, the application was developed on a local platform with role-based access, providing different levels of access and functionalities to users based on their roles. This enhanced security and ensured that only authorized users could access specific features and data.
The methodology employed in developing the application follows the Scrum Agile framework, emphasizing a systematic approach to enhance time and area calculation for agricultural field operations. The application streamlines operations through an intuitive user interface developed in React, ensuring responsiveness and user-friendliness. Middleware integration with Express facilitates seamless API handling, optimizing operational efficiency. MongoDB serves as the backend database for scalable data management and cloud storage, enhancing accessibility and reliability. This approach supports easy deployment via APK files on Android devices, maintaining data integrity in cloud storage. Overall, the development process adheres to Scrum Agile principles, fostering iterative improvements tailored to meet agricultural field requirements.
Recording of Geolocation
⌅The developed application continuously collects geolocations using the smartphone’s inbuilt GPS. It allows users to start and stop the recording of geolocation as needed. Further, users can set the recording time interval between 1,000 and 10,000 milliseconds. For better tracking, a user must enter a unique device Id, can be a name, numeric digit or other identifier. The application also displays the coordinates on the main screen in the form of latitude and longitude when the user clicks the “GET CURRENT LOCATION” button. The snapshot of developed user interface of the developed application is shown in Figure 1.
Segregation of plots
⌅Segregation of plots refers to the process of identifying and distinguishing the specific plots of land that have been covered or worked on by a tractor during ploughing, cultivation or other activities. As the tractor operates across different plots, the smartphone application logs the traversed areas and differentiates them as individual plots.
Algorithm for plot segregation
⌅-
The process begins by checking whether there are three or more than three points or not because it is required to have at least three points to form a polygon; if no, it does not form any closed loop and if yes, then it starts to identify all closed loops. A closed loop is a path starting from an initial point, traverses some distance and gets return to or near the initial point and, there can be many such loops in the whole track.
-
Further, it identifies the range of boundary points. The range of boundary points refers to the break in continuity shown in Figure 2. For example, there is an input of geolocations from 1 to 600 data points in which, points 1-250 are in loop I (ABCD) and 400-600 in loop II (EFGH) so, from points 251-399 (A-E), represent a discontinuity i.e., these points are in straight path not in any loop. Flow chart for the segregation of plot as per above algorithm is shown in Figure 3.
Computation of plot area
⌅The developed application was capable of calculating the area of segregated plots using an algorithm that identified extreme points and formed a closed loop by joining them. Once the loop was formed, the area was calculated.
Algorithm for area computation
⌅The GPS records geolocations in a coordinate system where abscissa (x-coordinate) and ordinate (y-coordinate) represented longitude and latitude, respectively. The basic concept was to create a closed polygon by joining the extreme points. The following steps were taken to calculate the area of field from the recorded geolocations through GPS.
-
A geolocation having the lowest x-coordinate was chosen from the recorded geolocations which was the left most point of the field and named as Lm(XL,YL).
-
Similarly, the top most (Tm), right most (Rm) and bottom most (Bm) geolocations were chosen by searching the highest y-coordinate, highest x-coordinates and lowest y-coordinates, respectively. Coordinates of these points are assumed as Tm(XT,YT), Rm(XR,YR) and Bm(XB,YB). These points are shown in Figure 4 for better clarity of the developed algorithm.
-
Four corners, A, B, C and D were identified having coordinate A(XL,YT), B(XR, YT), C(XR, YB ) and D(XL, YB), respectively, and joined them to make a smallest rectangle enclosing all these eight points for traversing the path.
-
After identifying the four edge points and four corners, starting from the edge point Lm in a clockwise direction searching for a point through A to Tm in such a manner that it follows a path of extreme points i.e. point closer to A and Tm, then again searches for a point that is closer to Tm and finally reaches to Tm, that’s how the left most point gets connected to top most point.
-
After this, it starts to search for a point from Tm that is closer to B and Rm ignoring other points and reaches to Rm, resulting the top most point gets connected to the right most point. Similarly, right most point and bottom most point gets connected to bottom most and left most point respectively resulting in the formation of a polygon by joining P (x1, y1), Q(x2, y2), R(x3, y3), S(x4, y4), T(x5, y5), U(x6, y6), V(x7, y7), W(x8, y8), X(x9, y9) and so on.
-
The area of the polygon formed by the points P-Q-R-S-T-U-V-W-X-P was calculated using their coordinates. The area was calculated in square unit using Eq. 1
| Points | Coordinates |
|---|---|
| P | x1, y1 |
| Q | x2, y2 |
| R | x3, y3 |
| S | x4, y4 |
| T | x5, y5 |
| U | x6, y6 |
| V | x7, y7 |
| W | x8, y8 |
| X | x9, y9 |
Flow chart for the computation of area as per above algorithm is depicted in Figure 5.
Evaluation Procedure
⌅The evaluation was carried out in two steps. The first step involved assessing the accuracy of plot area calculation, while the second step focused on evaluating the segregation efficiency.
For accuracy assessment, a tractor equipped with a cultivator was used to till fields while recording the area using the developed application. The evaluation was conducted across 25 fields of different sizes. The areas were also measured manually with the help of measuring tape and calculated by using the formula, A= (L1+L2)/2 *(B1 + B2)/2 where L1 and L2 are the lengths of opposite sides of field while B1 and B2 are the width of opposite sides of field. The manually calculated areas were then compared with the areas recorded by the developed application. Three different plot sizes (small, medium, and large) were tested with three replications each. Geolocation samples for each plot size are shown in Figure 6.
In the second step, the efficiency of plot segregation was evaluated by conducting five different trials in different patterns as follows:
-
In the first trial, the tractor started from the origin and covered three different fields continuously without any stoppage (Figure 7).
-
In the second trial, the tractor started and covered two fields with three stoppages in between (Figure 8).
-
In the third trial, the tractor covered one field, then stopped for fifteen minutes before covering two more fields (Figure 9).
-
In the fourth trial, the tractor covered two fields, stopped for half an hour and then covered two more fields (Figure 10).
-
In the fifth trial, the tractor covered one field, stopped for 15 minutes, then covered two more fields continuously, took a 20-25-minute stoppage, and finally reached the destination (Figure 11).
The area, entry, and exit times for each field during the five days are presented in the Results and Discussion section.
Results and Discussion
⌅Validation of time recording
⌅For evaluation, the plot sizes were divided into three categories: small (less than 1,000 m2), medium (1,000 to 2,500 m2) and large (more than 2,500 m2). Tests were conducted to evaluate the accuracy of time recording which is presented in Table 1. The average time recorded for nine small plots, compared to the manually measured time, showed an average error of -1.01 %. The negative sign indicates that the recorded value was higher than the manually measured time. Similarly, for medium and large plots, the error was 0.82 and 0.27 %, respectively which is negligible. Further, it was analyzed that the difference in time between manually noted value and recorded value through developed application was within 10-15 seconds. Though this difference was small, it might have resulted from the selection of starting and ending geolocation points in the developed application, which formed the basis for time calculation. However, statistical analysis showed no significant difference between the manually recorded and application-recorded times, as the p-value was greater than 0.05.
| Category of plot | No. of plot | Manually measured time (minutes) | Time recorded by developed system (minutes) | Percentage time difference (minutes) | |||
|---|---|---|---|---|---|---|---|
| Average | SD[1] | Average | SD[1] | Average | SD[1] | ||
| Small | 9 | 9.3 | 3.33 | 9.43 | 3.52 | -1.01 | 2.53 |
| Medium | 9 | 14.46 | 6.47 | 14.31 | 6.31 | 0.82 | 1.49 |
| Large | 7 | 26.42 | 11.87 | 26.38 | 11.89 | 0.27 | 2.02 |
[1]Standard Deviation
Validation of area recording
⌅Similarly, the area calculated by the developed system was compared with the manually measured area, as presented in Table 2. It is revealed from Table that the maximum percentage error in area calculation, with respect to the manually measured values was found within ± 7%. Further, the average percentage error for the small field was higher (6.99%) followed by medium (5.83%) and large (3.08%) fields. Statistically, a significant difference was observed in the case of small fields as the p-value was less than 0.05 whereas no significant difference was found for medium and large fields as the p-value was greater than 0.05. The percentage error was calculated using Eq. 2 in all cases. Since the error was dependent on the plot size, a graphical presentation is given in Figure 12 which indicates that as the plot size increased, the absolute percentage error decreased. The possible reason for errors in the area calculation is the propagation of errors while recording geolocations using GPS. In large fields, error was lower compared to smaller fields due to fewer turns. This indicates that a greater number of turns leads to a higher error.
| Category of plot | No. of plot | Manually measured area (m2) | Area recorded by developed system (m2) | Percentage area difference (m2) | |||
|---|---|---|---|---|---|---|---|
| Average | SD[1] | Average | SD[1] | Average | SD[1] | ||
| Small | 9 | 822.56 | 115.55 | 760.71 | 115.18 | 6.99 | 11.32 |
| Medium | 9 | 1,778.99 | 356 | 1,668.99 | 313.13 | 5.83 | 4.69 |
| Large | 7 | 2,748.23 | 1,082.91 | 2,667.01 | 1,067.25 | 3.08 | 2.64 |
[1]Standard Deviation
The percentage error is presented in a histogram of the frequency distribution curve (Figure 13) which indicates that the maximum frequency of percentage error falls between 5% and 10%. This positive value indicates that, in most cases, the area obtained through the developed application was less than the manually measured area. This recommends that the developed system can be used with an expected error range of 5-10%. The error can be further reduced by using a more precise GPS.
Validation of area segregation
⌅The area of segregated fields recorded by the developed application over five days is tabulated along with the manually measured area in Table 3. The area segregation efficiency was 100% as total of 15 fields were covered on different days and accordingly, the developed system calculated the area for all 15 fields. . Further, the table shows that though maximum error was about 20%, the average error was 9.41 %. Additionally, as the field size increased, the error decreased. Since the data were recorded using the GPS of the smartphone, the possible cause of error can be the satellite position, GPS drift, signal obstruction, calculation inaccuracies, and rounding errors. However, the likelihood of error in area calculation is minimal, as the system employs standard mathematical formulas. The measured error can be further reduced by using a more precise GPS with higher accuracy.
| No. of Days | Field | Manually (m2) | Developed Application (m2) | Error (%) | |
|---|---|---|---|---|---|
| Day 1 | F1[1] | 1,890.22 | 1,576.625 | 16.59 | |
| F2[2] | 1,581.07 | 1,332.875 | 15.7 | ||
| F3[3] | 1,615.62 | 1,298.625 | 19.62 | ||
| Day 2 | F1[1] | 2,685.75 | 2,835.4 | -5.57 | |
| F2[2] | 3,200.5 | 2,973.59 | 7.09 | ||
| Day 3 | F1[1] | 2,769.37 | 2,511.75 | 9.3 | |
| F2[2] | 2,294.13 | 2,035.4 | 11.28 | ||
| F3[3] | 2,679.83 | 2,719 | -1.46 | ||
| Day 4 | F1[1] | 3,692.6 | 3,419.5 | 7.4 | |
| F2[2] | 2,788.69 | 2,495.5 | 10.51 | ||
| F3[3] | 4,285.22 | 4,009.875 | 6.43 | ||
| F4[4] | 4,200.62 | 3,987.5 | 5.07 | ||
| Day 5 | F1[1] | 1,353.37 | 1,083.75 | 19.92 | |
| F2[2] | 2,978.55 | 2,635.5 | 11.52 | ||
| F3[3] | 3,240.33 | 2,988.87 | 7.76 | ||
| Average | 9.41 | ||||
[1] [2] [3] [4] Denote different fields (Field 1, Field 2, Field 3, Field 4) in the study area.
Conclusion
⌅A mobile application was developed to record the geolocation of machine’s movement using the GPS of a smartphone. The developed application was capable of segregating the tractor’s operation in different fields and calculating the area of each field. It also recorded the time taken in each field. It was observed that the error in time measurement, compared to the manually noted values, was non-significant. However, in the case of area measurement, the average percentage error was found to be 6.91% for small fields (area less than 1,000 m²), 5.40% for medium fields (area between 1,000 and 2,500 m²), and 2.46% for large fields (area more than 2,500 m²). This indicates that as the plot size increases, the error decreases. Furthermore, no error was found in field segregation, confirming that the developed system is capable of identifying field operations and idle movement. Therefore, the developed application can be considered an effective tool for monitoring farm machines’ operation in terms of area coverage and time taken, which will ultimately be useful for charging the custom hiring services.
Acknowledgements
⌅Authors hereby express their profound gratitude, and cordial thanks to the Department of Farm Machinery and Power Engineering, College of Agricultural Engineering, Dr. Rajendra Prasad Central Agricultural University, Pusa, Bihar, India, and RetiUno Technologies Private Ltd. Bangalore, India for the completion of this research.
Statement about use of generative AI
⌅The translation of the title, abstract, and keywords from the original version in English to Spanish has been generated using OpenAI, ChatGPT GPT-4o mini (2024).
Competing interests
⌅The authors have declared that no competing interests exist.
Authors’ contributions
⌅Apoorva Sharma: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Visualization, Writing - original draft, Writing - review & editing. Prabhanjan Kumar Pranav: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Visualization, Writing - original draft, Writing - review & editing. Ritesh Prasad: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Visualization, Writing - original draft, Writing - review & editing.
| Funding agencies/institutions | Project / Grant |
|---|---|
| All India Coordinated Research Project on Farm Implements and Machinery (AICRP on FIM). | 2,50,000 |