Agricultural Fields 2D and 3D Models Dataset

Agricultural Fields 2D and 3D Models Dataset

Introduction

This dataset was created to address the lack of comprehensive datasets in the literature that provide necessary information to evaluate and validate path planning approaches on both 2D and 3D surfaces of agricultural fields. It comprises 30 manually-selected agricultural fields located in France, chosen to cover a diverse range of shapes and sizes (from 1.83 to 13.21 hectares). The dataset includes simple shapes that do not require field decomposition and more complex shapes that necessitate field decomposition, ensuring a broad representation of real-world scenarios. 

This dataset was initially produced to validate our Complete Coverage Path Planning approach, and we are pleased to make these data available for future research. In sharing this dataset, we kindly ask that users cite this dataset in any publications or presentations that make use of the data. This will help acknowledge our contribution and encourage further collaboration and research in this area.

Background

Agricultural field shapes result from a complex interplay of historical, geographic, and topographic factors, as well as cultural and economic practices. Fields in countries with a more recent history of land ownership and partitioning may have simpler shapes, while those with more complex histories may have irregular shapes. Geography and topography also influence field shapes, with fields in flat, open areas having simpler shapes than those in mountainous or hilly regions. This dataset focuses on French fields due to the variety of field shapes and the availability of high-precision elevation data from the French government.

Dataset Content

For each of the 30 agricultural fields, this dataset provides the following information in separate files:

Aerial image (PNG)

2D polygon (XML)

2D triangulated surface (PLY) with a grid resolution of 0.25 m

Elevation grid (PLY) with a grid resolution of 5 m

3D triangulated surface (PLY) with a grid resolution of 0.25 m

Set of 2D line segments representing access segments (XML)

Set of dividing lines for fields 20-30 to decompose them into sub-polygons in different ways

The obtained result by our "Advanced 3D Hybrid Path Planning with Multiple Objectives for complete coverage of agricultural field by wheeled robots", which includes:

A way-points in a CSV file

An illustration of the result projected on the field surface

Note: All coordinates are represented in Cartesian coordinates with centimeter precision.

The table below provides links to the field data in the Géoportail platform and coordinates (longitude and latitude) of a point inside each field for all 30 fields. These links and coordinates can be used to access the data and for visualization purposes.

Field
Link
Lon / Lat

1
bit.ly/3FYtuKu
7.435° / 48.7732°

2
bit.ly/3WGAyRI
7.474° / 48.7825°

3
bit.ly/3zX1vqJ
2.9205° / 49.8115°

4
bit.ly/3DJL0PI
1.6713° / 47.9864°

5
bit.ly/3htb8H3
3.3216° / 50.6623°

6
bit.ly/3WGTfER
7.4311° / 48.8245°

7
bit.ly/3DP8vqG
2.4845° / 50.3106°

8
bit.ly/3NLmQJf
7.5924° / 48.831°

9
bit.ly/3EeTvUo
7.4641° / 48.8146°

10
bit.ly/3UOyTrv
1.3491° / 48.012°

11
bit.ly/3zW7v30
3.4701° / 46.652°

12
bit.ly/3UMC6I3
7.5742° / 48.8071°

13
bit.ly/3TjkOkA
3.578° / 46.7016°

14
bit.ly/3UAmdo0
7.4269° / 48.8194°

15
bit.ly/3GpjdXZ
3.5611° / 46.6875°

16
bit.ly/3tcGhRN
2.5127° / 48.2645°

17
bit.ly/3zW26sE
2.6443° / 48.2546°

18
bit.ly/3Trsqlq
7.9196° / 48.9513°

19
bit.ly/3DWHgKJ
2.1269° / 46.8124°

20
bit.ly/3NN8pnT
1.5874° / 47.1346°

21
bit.ly/3DShkA3
0.6254° / 49.191°

22
bit.ly/3zZK1dg
2.7067° / 50.3336°

23
bit.ly/3TmwcMC
7.4416° / 48.7223°

24
bit.ly/3E3l8OK
3.1021° / 48.2449°

25
bit.ly/3E0Raeq
1.6183° / 49.9655°

26
bit.ly/3tvN0Xg
3.5476° / 50.1441°

27
bit.ly/3A0tZ2D
3.6644° / 48.0046°

28
bit.ly/3fTlQGl
1.7086° / 47.2054°

29
bit.ly/3hBeLL2
1.6893° / 47.1421°

30
bit.ly/3Edm2cN
3.1018° / 48.5853°

Hybrid_CCPP_Result Subdirectory

Hybrid_CCPP_Result subdirectory contains the results of our path planning algorithm for complete coverage of agricultural fields by wheeled robots. The provided files include way-points in CSV format and an illustration of the result projected on the field surface.

Approach Parameters

The results were obtained under the following considerations: The driving direction step size ($\ell_s$), the spacing of access segment discretization ($\ell_a$) and the spacing of working trajectory discretization for slope computation ($\ell_{slp}$). These parameters were respectively set to $3°$, $0.5m$, and $0.5m$. The values of other parameters are listed in the table below:

Parameter
Description
Value

$w$
working width
3m

$\gamma_{on}$
minimum turning radius - implement on
10m

$\gamma_{off}$
minimum turning radius - implement off
2.8m

$V_{on}$
average speed - implement on
4.5m/s

$V_{gap}$
average speed - implement transition
1.5m/s

$V_{off}$
average speed - implement off
3.5m/s

$\ell_t$
transition trajectory length
1.5m

$\ell_o$
robot-implement offset
1.0m

$\Delta_{mwd}$
minimum working distance threshold
3m

$p$
number of inner trajectories
2

$g$
number of gap-covering trajectories
1

$W_{cov}$
weight of $S_{cov}$
0.30

$W_{ovl}$
weight of $S_{ovl}$
0.15

$W_{nwd}$
weight of $S_{nwd}$
0.10

$W_{otm}$
weight of $S_{otm}$
0.10

$W_{slp}$
weight of $S_{slp}$
0.35

$W_{s0}$
weight of $\ell_{s0}$
0.00

$W_{s1}$
weight of $\ell_{s1}$
0.10

$W_{s2}$
weight of $\ell_{s2}$
0.15

$W_{s3}$
weight of $\ell_{s3}$
0.20

$W_{s4}$
weight of $\ell_{s4}$
0.25

$W_{s5}$
weight of $\ell_{s5}$
0.30

For an in-depth understanding of these parameters, we kindly invite you to consult our published article:

Pour Arab, D., Spisser, M. & Essert, C. (2024) 3D hybrid path planning for optimized coverage of agricultural fields: a novel approach for wheeled robots. Journal of Field Robotics, 1–19. https://doi.org/10.1002/rob.22422

Way-point Structure

A way-point is represented by the following format:

Point X, Point Y, Point Z, Heading, Type, Move

where Heading is in radians, and Type and Move are according to the following structures:

enum WayPointType {
WORKING = 1,
TURN_OFF = 2,
TURN_ON = 3,
TRANSITION_OFF_TO_ON = 4,
TRANSITION_ON_TO_OFF = 5
};

enum RobotMove {
FORWARD = 1,
REVERSE = -1
};

WayPointType

WORKING: The robot implement for driving at this point must be on.

TURN_OFF: The robot is performing a turn while its implement is off and elevated from the ground.

TURN_ON: The robot is performing a turn while its implement is on.

TRANSITION_OFF_TO_ON: The robot is traveling a straight transition trajectory for turning on its implement.

TRANSITION_ON_TO_OFF: The robot is traveling a straight transition trajectory for turning off its implement.

RobotMove

FORWARD: The robot is moving forward.

REVERSE: The robot is moving in reverse.

Files

CSV file contains the way-points generated by our proposed approach.

SVG file providing an illustration of the result projected on the field surface.

Usage

This dataset is intended for researchers and developers working on path planning algorithms for agricultural applications. Users can leverage the data to evaluate and validate their path planning approaches in various scenarios, from simple to complex field shapes, and on both 2D and 3D surfaces.

Please ensure that you cite this dataset appropriately in any publications or presentations that make use of the data.