Gardening Robot

Autonomous Systems Lab
Prof. Roland Siegwart
Master-Thesis
Gardening Robotics
Design of a Seed-Planting Robot for the
Creation of Large-Scale Growing Flower
Images
Spring Term 2011
Supervised by:
Dr. Cedric Pradalier, ASL
Dr. Paul Beardsley, DRZ
Dr. Wojciech Matusik, DRZ
Authors:
Stefan Riesen
Linus Rohrer
Contents
Abstract
iii
Abstract (German)
List of Figures
vii
List of Tables
ix
Preface
xi
1 Introduction
1.1 Project Goals and Requirements .
1.1.1 Organization . . . . . . . .
1
........
. . . . . . . . . . .
1
........
. . . . . . . . . . .
11.1.2 Seed properties . . . . . . . ........
.
. . . . . . . . . . 2
........
. . . . . . . . . . .
3
Horticulture
. . . . . . . . . . .
31.2.2 Seed Delivery Systems . . . ........
.
. . . . . . . . . . 31.2.3
........
. . . . . . . . . . .
4
1.2 Related Work . . . . . . . . . . . .

1.2.1 Robotics in Agriculture and
Creation of Flower Images .

2 Seed Delivery System
2.1 Basic Principles . . . . . . . . . . . . . . .
2.1.1 Selection Principle . . . . . . . . .
5
. . . . . . . . . . . . . . .
5
. . . . . . . . . . . . . . .
52.1.2 Delivery Principles . . . . .
. . . . . . . . . . . . . . .
62.1.3 Placement Principles . . . .
. . . . . . . . . . . . . . .
7
. . . . . . . . . . . . . . .
8
. . . . . . . . . . . . . . .
82.2.2 Ejection Mechanism . . . .
. . . . . . . . . . . . . . .
9
. . . . . . . . . . . . . . . 10
. . . . . . . . . . . . . . . 10
. . . . . . . . . . . . . . . 10
. . . . . . . . . . . . . . . 14
. . . . . . . . . . . . . . . 14
. . . . . . . . . . . . . . . 14
....
....
2.2 Concept . . . . . . . . . . . . . . . . . . .
2.2.1 Seed Selection Mechanism . . . . .
....
2.3 Engineering . . . . . . . . . . . . . . . . .
2.3.1 Selection Mechanism . . . . . . . .
2.3.2 Ejection . . . . . . . . . . . . . . .
2.4 Extendability and Possible Improvements
2.4.1 More Seed Types . . . . . . . . . .
2.4.2 Sensors on the Sliders . . . . . . .
3 XY-Table for Tiling
3.1 Concept . . . . . . . . . . . . .
3.1.1 Substrate . . . . . . . .
3.1.2 Glue . . . . . . . . . . .
3.2 Engineering . . . . . . . . . . .
3.2.1 Mechanical Design . . .
3.2.2 Electronic System . . .
3.3 Software . . . . . . . . . . . . .
3.3.1 Graphical User Interface
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
15
15
15
17
17
18
20
20
3.3.2
3.3.3
Arduino Microcontroller . . . . . . . . . . . . . . . . . . . . .
Planting Process . . . . . . . . . . . . . . . . . . . . . . . . .
4 Testing
4.1 Setup . . . . . . . . . . . . .
4.2 Number of Ejected Seeds . .
4.3 Seeding Precision . . . . . . .
4.4 Ejection Sensor Performance
21
21
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
25
25
25
26
26
5 Mobile Platform
5.1 Requirements . . . . . . . . . . . . . . . . .
5.1.1 Terrain Form Constraints . . . . . .
5.1.2 Obstacle Avoidance . . . . . . . . .
5.1.3 Precision . . . . . . . . . . . . . . .
5.1.4 Seeding Constraints . . . . . . . . .
5.2 Locomotion Concept . . . . . . . . . . . . .
5.2.1 Evaluation . . . . . . . . . . . . . .
5.2.2 Selection . . . . . . . . . . . . . . .
5.2.3 Wheel Configuration . . . . . . . . .
5.3 Localization . . . . . . . . . . . . . . . . . .
5.3.1 Evaluation . . . . . . . . . . . . . .
5.3.2 Selection . . . . . . . . . . . . . . .
5.4 Required Components and Size Estimation
5.4.1 Overview . . . . . . . . . . . . . . .
5.4.2 Manipulator . . . . . . . . . . . . . .
5.5 Control . . . . . . . . . . . . . . . . . . . .
5.5.1 Pose Estimation . . . . . . . . . . .
5.5.2 Planting . . . . . . . . . . . . . . . .
5.5.3 Path Planning . . . . . . . . . . . .
5.6 Available Platforms . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
31
31
31
33
33
33
33
34
34
35
35
37
37
37
40
40
40
40
42
44
6 Conclusion and Outlook

6.1 Conclusion . . . . . . . . . . . . . . . . . .
6.2 Outlook and Future Work . . . . . . . . . .
6.2.1 Engineering of the Mobile Platform
6.2.2 Grass seeding . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
45
45
45
45
46
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Datasheets
47
B Test Results
57
C Content of the DVD
67
Bibliography
69
ii
Abstract
This project deals with the development of a robot that is able to plant a small number of
seeds of diferent species at predefined positions for the creation of growing ower images.
For this purpose a seed delivery system that is able to select, deliver and place the seeds was
engineered. This mechanism has been integrated into a Cartesian robot which can plant a tile
of one square meter on a transportable substrate. It yields the possibility to prepare the ower
images independently of the time and location of the final deployment.
The seed delivery system consists of five dispenser modules that can select a specified amount
of seeds from a seedbox. Each module is adaptable to the seed form and size through an
exchangeable opening mechanism. More modules can be added easily to further increase the
performance. An air pressure based delivery shoots the seeds from the dispenser modules to
the ejection subsystem through a plastic tube. A sensor system was added at the outlet to
detect whether a seed has been placed correctly or not in order to close the control loop.
The Cartesian stage is assembled as an H-configuration with tooth-belted axes and electrical
motors operated at 24VDC. As the manipulator only needs to carry the ejection subsystem
and the delivery tube, the power demands on the motors are relatively low. Further a control
software and user interface was written for both Microsoft Windows and Linux. For the
Linux software a node in the robot operating system (ROS) framework has been
implemented.
The performance of the system has been tested extensively for reliability, precision and
repeatability with seeds of diferent sizes and forms. It has been observed that the number of
ejected seeds varies only within the specified tolerance. Concerning the precision, the
requirements were also fulfilled: 67% of all seeds were placed within one centimetre and
85% were detected in a radius of 1.5cm from the reference point. Further the performance of
the ejection sensor has also been evaluated. All seeds were reliably detected, especially for
large seed sizes or high numbers of ejected seeds.
Finally, a concept for a mobile robot which can plant ower images autonomously was
developed. A similar manipulator as in the static solution is intended to be used. Several
locomotion and localization concepts have been evaluated and it is proposed to use a
diferential drive platform combined with laser based triangulation. As observed in first
simulations, the precision of the pose estimate could be further increased by adding a
compass.
Keywords: Gardening Robotics, Flower Images, Seed Delivery System, Cartesian
Robot, Autonomous Mobile Robot
iii
iv
Abstract
Das Ziel dieser Arbeit ist die Entwicklung eines Bepanzungsroboters, der Blumen- samen
verschiedener Arten in einem vordefinierten Muster setzen kann. Dadurch
wird es erm ht ein beliebig grosses Blumenbild zu erstellen. Ein Mechanismus, oglic
der einzelne Samen aus einem Beh alter ausw ahlen und zu einem Auslass transportieren kann, wurde dazu entworfen. Um die Blumenbilder unabh angig von Zeit
und Ort der endg ultigen Bepanzung vorbereiten zu k onnen, wurde der Mechanismus in einen XY-Tisch integriert, wobei die Samen auf eine wasserl he Folie von oslic
je einem Quadratmeter Fl he platziert werden. Die Folie kann danach auf dem ac
Feld ausgelegt und bew assert werden.
Das System besteht aus f Dispenser-Modulen, die jeweils eine bestimmte Menge unf
an Samen aus einem Beh alter ausgeben. Die Module k onnen der Form und Gr osse
der zu panzenden Samen angepasst werden, indem der Ofnungsmechanismus aus-
gewechselt wird. Weitere Module k onnen leicht hinzugef werden, um die Funk- ugt
tionalit zu erweitern. Die Samen werden durch ein Luftdrucksystem von den at
Dispenser-Modulen zu einem Auslassmechanismus bef ordert, wo ein Sensorsystem
angebracht wurde, um die korrekte Ausgabe zu erpr
ub
ufen und dadurch den Regelkreis zu schliessen.
Der kartesische Roboter wurde als H-Portal, bestehend aus Zahnriemenachsen und 24V
Gleichstrommotoren, realisiert. Da die bewegte Masse nur aus dem Auslasssystem und dem Zuf uhrschlauch besteht, sind die Leistungsanforderungen an die Motoren relativ gering. Ferner wurde eine Steuerungs-Software und eine Benutzerober he sowohl f Microsoft Windows wie auch f Linux programmiert. Die
ac
ur
ur
Linux-Software wurde als Node im Robot Operating System (ROS) Framework
implementiert.
Die Zuverl assigkeit, Pr azision und Wiederholbarkeit des Systems wurde ausgiebig
mit Samen verschiedener Gr ossen und Formen untersucht. Dabei zeigte sich, dass
die Anzahl der platzierten Blumensamen nur innerhalb der vorgegebenen Toleranz
variiert. Im Bezug auf die Pr azision konnten die Anforderungen ebenfalls erf ullt
werden: 67% aller Samen wurden in einem Umkreis von einem Zentimeter und 85%
innerhalb von 1.5cm vom Referenzpunkt platziert. Der Auslasssensor erkannte alle
Samen zuverl assig, insbesondere grosse K orner oder eine hohe Anzahl davon.
Schliesslich wurde ein Konzept f einen mobilen Roboter entwickelt, der die Blu- ur
menbilder autonom bepanzen kann. Der Manipulator aus der statischen L osung
kann mit leichten Ver anderungen ernommen werden. Verschiedene Fortbewe- ub
gungs- und Lokalisierungskonzepte wurden evaluiert und es wird vorgeschlagen, eine
Plattform mit einem diferentiellen Antrieb zu verwenden. Zur Lokalisierung kann LaserTriangulation in Verbindung mit einem Kompass und der Odometrie eingesetzt werden. Wie erste Simulationen zeigten, kann dadurch eine hohe Pr azision
erreicht werden.
Stichworte: Bepanzungsroboter, Blumenbilder, XY-Tisch, Autonomer mobiler
Roboter
vi
List of Figures
2.1
Vacuum probe . . . . . . . . . . . . . . . . . . . . . . . . . .
62.2
...........................
82.3
limited opening volume . . . . . . . . . . . . . . .
92.4
limited opening volume and control pressure tube
92.5
delivery module for multiple seed types . . . . . . . . .
2.6 Relay driver . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7 Seed ejection mechanism . . . . . . . . . . . . . . . . . . . . .
2.8 Ejection sensor schematic . . . . . . . . . . . . . . . . . . . .
3.1
3.2
3.3
3.4
XY-Table (Overview) . . . .
XY-Table system schematics
Graphical user interface . . .
Planting process owchart . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4.1
4.2
4.3
4.4
4.5
4.6
4.7
Test setup . . . . . . . . . . . . . . . . . . . . .
Number of ejected seeds . . . . . . . . . . . . .
Position of ejected Viola seeds . . . . . . . . . .
Position of ejected Phlox seeds . . . . . . . . .
Position of ejected Impatiens seeds . . . . . . .
Position of ejected seeds from all measurements
Precision/Recall of the ejection sensor . . . . .
5.1
5.2
5.3
5.4
Terrain constraints . . . . . . . . .
Overview of the mobile platform .
ROS-Node structure for the control
Path planning . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
. . . .
Slider . . . .
. . . .
Slider with
. . . .
Slider with
. . . .
Seed
. . . . 11
. . . . 12
. . . . 13
. . . . 14
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
19
21
23
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
26
27
28
28
28
29
30
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
32
39
42
43
...........
...........
of the manipulator
...........
vii
viii
List of Tables
3.1
Substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
4.1
Ejection sensor performance . . . . . . . . . . . . . . . . . . . . . . .
29
5.1
5.2
5.3
Localization methods summary . . . . . . . . . . . . . . . . . . . . . Required

components . . . . . . . . . . . . . . . . . . . . . . . . . . Position control algorithm
parameter description . . . . . . . . . . .
38
39
41
B.1
B.2
B.3
B.4
B.5
Seeding
Seeding
Seeding
Seeding
Seeding
test
test
test
test
test
results
results
results
results
results
for
for
for
for
for
Impatiens, Slider
Violas, Slider S .
Violas, Slider M
Phlox, Slider M
Phlox, Slider L .
ix
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
57
59
61
63
65
Preface
This Master's thesis concludes our studies in Robotics, Systems and Control at the
Swiss Federal Institute of Technology in Z h. The project is a cooperation be- uric
tween the Autonomous Systems Lab (ASL) and Disney Research Z h (DRZ) and uric
deals with the robotic creation of a large-scale ower image. Growing images are widely
used as a decoration element of outdoor areas, such as city gardens, round- abouts, around
public buildings as well as in amusement parks. Such images are created manually and have
therefore high labour costs and limited visual efects. With an autonomous robot, this cost can
be minimized and the image complexity increased. This application has commercial potential
for outdoor advertisements and as an attraction for amusement park visitors.
Due to the diversity of challenges, this has been a very interesting project and we
wish to thank
Prof. Dr. Roland Siegwart for giving us the opportunity to conduct this project
at the Autonomous Systems Lab.
Our supervisors Dr. C edric Pradalier at ASL, Dr. Paul Beardsley and Dr. Wojciech Matusik at Disney Research, for their support and the encouraging discussions.
All other people directly afliated with the project for the inspiring discussions and the insight into their parts: Dr. Derek Bradley, Dr. Frank Liebisch, Dr.
Norbert Kirchgessner and Dr. Robert Wang.
The IT-supporters Stefan Bertschi and Thomas Baumgartner for their support
in administrative, IT-related and electronics-related questions.
The mechanic technicians Markus B uhler and Dario Fenner for the manufacturing of the mechanical parts and support in related questions.
All members of ASL and DRZ for their help in questions related to their
specific fields of research.
And last, but not least, our friends and families for their great support and
encouragement during both this project and the entire studies.
Z urich, August 2011
Stefan Riesen and Linus Rohrer
xi
xii
Chapter 1
Introduction
1.1
Project Goals and Requirements
The overall goal of the Disney Flower Images Project is to grow a bed of owers with the
appearance of a realistic image. Flower images are already found in city parks, but they are
created manually which is time-consuming and limits the visual efects. By automating the
process, it is aimed to achieve more sophisticated results. This thesis is focused on the
robotics component of the project, and the goal is to make a seed-planting robot. Two
variants of the robot are to be developed. The first is a stationary robot which can plant seeds
onto a tile of one square meter. The second variant is a fully autonomous mobile robot which
can plant directly in an outdoor setting with variable field sizes. Both robots shall be able to
handle a wide range of plant seeds, whereas diferent seeds can be used simultaneously. The
creation of an optimized ower pattern given from an input image was not in the focus of
this thesis and is considered to be available as an input to the system.
1.1.1
Organization
The project is organized in three engineering stages. These are not entirely dependent on the results of the previous stage and can therefore be developed concurrently.
Stage 1: Development of a Seed Delivery System
In the first stage a seed delivery system for the specified seed types (See section 1.1.2)
is to be developed. The mechanism shall be able to select a defined amount of seeds and
deliver them to an outlet where they are placed. The selection mechanism is required to be
designed in way that the seeds can be selected from multiple seed types, either diferent
colours of the same species or diferent species. A key point is the reliability: It is important
that the number of seeds selected and delivered does not have a high variance and that the
average number of seeds corresponds to a predefined number. This number may vary
between species. The desired ejection precision is that most seeds are in a range of1.5cm
from the reference point. The seed delivery mechanism has to be designed in a way to be
operated in one of the
following modes:
Multipass: The robot passes the field multiple times with another seed type
each time. Between the passes, it is required to exchange or refill the seedbox.
Singlepass: The robot passes the field only once seeding all seed types in this
pass. The mechanism selects the correct seed type from multiple seedboxes and places
it.
1
Chapter 1. Introduction
Hybrid: The robot passes the field multiple times with a set of diferent seed
types each time. Between the passes, the seedboxes are exchanged either manually or
automatically.
Stage 2: Implementation of the Seed Delivery System on a Stationary
2D-Robot
In a second step, the seed delivery mechanism is mounted on a static robot which
allows the planting of ower images onto a substrate. This can either be a box of soil, which
is then used for growing the plants in a greenhouse and then placed onto the field. An
alternative would be a bio-degradable substrate, which is laid out on the field where the
plants are directly grown. For simplicity, the tiles shall be as big as possible in order to
reduce the number, but on the other hand, for easy handling of the tiles, small tiles would be
preferred. Therefore the size of one tile has been defined to one square meter, as it is a good
compromise between handiness, transportability and a low number of tiles.
Stage 3: Implementation of the Seed Delivery System on an Autonomous
Platform
To allow the fully autonomous creation of large-scale growing ower images, an
autonomous mobile robot is to be developed in the third stage. The robot shall be equipped
with a manipulator based on the seed delivery system developed in the first stage. Further the
robot has to be able to localize itself with a suitable algorithm and to plan the path on the
field autonomously. If, for any reason, the achieved positioning precision of the robot is less
than the required precision for planting, the manipulator has to be designed in a way that it
can compensate any misplacements of the robot in order to increase the planting precision to
the desired level. This implies that the localization precision is higher than the required
planting precision.
1.1.2
Seed properties
From a list of species that are considered suitable for the creation of growing images,
the seeds have been analyzed. They were classified according to their form and size.
Dahlia hybrids: Flat, longish, about 1x4mm

Dorotheantus bellidiformis: Spherical, about 0.1mm in diameter
Gazania hybrids: Ellipsoid, about 0.8x2mm
Impatiens walleriana: Ellipsoid, about 0.5x0.8mm
Petunia hybrids: Spherical, about 0.2mm in diameter
Phlox drummondii : Ellipsoid, about 0.7x1mm
Portulaca grandiora: Spherical, about 0.3mm in diameter
Verbena hybrids: Ellipsoid, about 0.5x2.5mm
Viola wittrockiana: Nearly spherical, about 0.8mm in diameter. Note: Violas
are biannual and therefore only suitable for the long-term creation of ower images.
From the list a subset of species with similar form and size class was selected to
specify the requirements on the robot in this stage, but it has also been defined that the
selection mechanism shall be extendable to other species. The selected species are Violas,
Phlox and Impatiens. These seeds have their roundish form and their approximate size of about
0.5-1mm in all directions in common.
1.2. Related Work
1.2
1.2.1
Related Work
Robotics in Agriculture and Horticulture
As agriculture is extensively supported by technical means like seeding, mowing or

harvesting machines, it is widely considered to be a field with a high potential for robotic
application as it is a small step from these semi-automatically operated machines to fully
autonomous robots in both greenhouse and open field applications. Robots are available on all
development levels from experimental to market-ready in several agricultural applications
[4], but most of them are in research, where institutes have made progress to extend the
existing agricultural machines to robotic systems. Comba et al. [1] have reviewed diferent
publications in the context of automated crop management. Most of the robots considered in
this publication are developed for harvesting and weed control as these are the most
expensive, often manually conducted, operations in agriculture. Seeding is not yet as
important since there are already good tractor based seeding systems.
In horticulture there are significantly less robotic applications as in agriculture. The
big exception are small mowing robots for home use, but robots for precise planting of single
plants or autonomous hedge cutting are not yet available on the market, probably due to high
development cost and complexity in relation to the market size. It can be concluded that for
the creation of growing ower images, no existing platform can be used or further developed,
but a new one has to be designed from scratch.
1.2.2
Seed Delivery Systems
The automatic delivery of seeds out of a seedbox onto the ground has been developed quite
early in industrial history. These systems distribute the seeds in a predefined area using
drums, drilling or digging devices, but none of these systems is used for precise placement of
a single seed to a defined position. Also most of these systems deal with only one seed type.
Nevertheless, in the last 30 years some attempts to pick up a single seed have been made.
Maury [7] patented an apparatus where the seeds are placed on a rotating disk and then are
shot with air pressure into the ground when their position on the disk passes the output
valve. Similarly, Yaji et al. [14] first use a negative air pressure to pick a seed out of a
seedbox, move it to the outlet position and then apply a positive air pressure to shoot the seed
into the ground.
Keller et al. further developed this principle in [6] where the seeds are picked up
by a precise vacuum probe and let them fall by inverting the pressure. Their robot is not used
to plant seeds, but to transfer seeds from one seedbox to another. A diferent approach to
pick up a single seed is used by Tonus [9]. In his patent a needle moves from the ground of
a seedbox to its top. The tip is designed in a way that only one seed remains on it. After
reaching the top of the seedbox, the tip enters a small tube from where the seed on the tip is
transported to the outlet using air pressure.
For the application of this project, there is no ready-to-use solution available, but
several solutions that can be used as a base to design a mechanism. Especially, this applies to
the vacuum probe and to the mechanical selection combined with an air pressure based
delivery system. Those two approaches seem to be most promising as a starting point since
they allow precise selection and planting of single seeds. For the reuse of the vacuum system,
legal issues have to be taken into consideration.
1.2.3
Chapter 1. Introduction
Creation of Flower Images
The creation of ower images, as used for decoration of gardens or parks and/or for
advertising purposes, is still a manual operation. Trebbin et al. [10, 11] have started research
in this direction. They use the following two diferent approaches. The first is to plant the
seeds which form the image directly at the corresponding locations. They intended to use a
drilling machine that is moved by a tractor navigated by the global positioning system GPS.
As the GPS has a localization precision of about 2.5m, it is assumed that a GPS located and
manually driven tractor would not reach the desired planting precision, thus not leading to a
satisfying ower image. Additional sensors for an accurate pose estimation and an
automatically driven vehicle, as intended to be developed in the third project stage (See
section 1.1.1), are expected to give a significantly better result.
The second approach is to print the seeds on a substrate made of poly-vinyl-alcohol similar
to the patent by Wirz [13]. The substrate is then laid on the ground and covered with soil.
By watering the substrate it dissolves and the seeds remain in their location. This seems to
be a considerable alternative to planting on soil tiles for the second development stage (See
section 1.1.1). A substrate is lighter than a box of soil and is therefore easier to handle.
Further the tiles can be prepared independent of time and location of the final deployment
and can be stored until needed.
As Trebbin et al. [10, 11] have withdrawn their patent application for the first approach, it is
assumed that they focus on the second approach. Neither is published how they create an
optimal owering pattern out of the image to get the best solution.
Chapter 2
Seed Delivery System

Ideas how to handle a small number of seeds are discussed in this chapter whereby the task is divided into
the three subproblems Selection, Delivery and Placement. Based on this discussion, a solution that includes a
seedbox with a controllable downward opening and an air pressure based delivery is presented. Finally,
some theoretical thoughts of possible improvements are listed.
2.1
Basic Principles
This section describes the possible physical efects that are considered to be useful for the
seed delivery system. It is distinguished between the tasks Selection, Delivery and Placement.
Principles were developed for each task separately and then merged into the concept
described in section 2.2 below.
2.1.1
Selection Principle
In the selection task, the robot or parts of it must be able to select a specified number of a
defined seed species out of a box and hand the seeds over to the subsystem which performs
the delivery task. This can be achieved by the principles described below.
Vacuum
As illustrated in fig 2.1 a seed is picked up by a probe which can create a negative air
pressure at its tip. Vacuum is widely used for pick-and-place operations of objects in
various size classes. To select diferent seeds from multiple boxes the probe movement needs
only to be programmed according to the positions of the boxes. To hand the picked up seed
over to a delivery mechanism, a positive air pressure is created at the probe to blow the seed
away onto the delivery mechanism or directly onto the placement position. This method is
patented by Keller et al. [6] for the handling of seeds. Advantages are that any size and form
class can be handled, as well as the number of seeds is adjustable by a change of the tip size.
A major drawback is the need for an additional axis as the tip is required to move vertically
for reliable pick up when the number of seeds in the box changes. Further - if combined with
direct delivery - the planting time is significantly increased since the probe has to cover the
distance from the pick-up to the planting position for each seeding point.
5
Chapter 2. Seed Delivery System
Vacuum probe
picked-up seed
Seedbox
Figure 2.1: Vacuum probe
Gravity
When gravity shall be used as a selection principle, the seedbox needs to have a controllable
downward opening. This can be achieved with an electromagnet, a motor or a pneumatic
system that manipulates the opening mechanism. The simplicity of this concept is a big
advantage which leads to lower design efort and production costs. Disadvantages are the fact
that any moving parts can be blocked by seeds and that the number of selected seeds is only
defined by a statistical process. Multiple seed types can be easily combined by adding a
module for each species. Each selection mechanism can be tailored to the specific properties
of the corresponding seed.
Electrostatic Valve
It has been observed that the seeds stick to electrostatically charged surfaces. This could be
used in combination with a probe that has a controllable electrostatically chargeable surface
at its tip. Electrostatics have already been used by Balachandran et al. [3] in order to control
the ow of agricultural seeds in bulk transportation. This could be adapted to single seeds,
but would require fundamental research on the electrostatic properties of diferent seeds.
Mechanical Pick-Up
Yet another method would be to pick up seeds with a mechanical device as a gripper. Similar
to the vacuum principle, this also requires a third axis for pick-up and placement. Although
it is very exible for diferent seed types and sizes, a gripper requires a very precise
mechanical design in order to handle the full variation of diferent seeds without destroying
them. Further a complex control structure is needed which is closely tied to the mechanical
properties and that can also survey the selected amount of seeds.
2.1.2
Delivery Principles
In the delivery task, the corresponding parts have to deliver the selected seeds to the
placement mechanism. The following principles show possible solutions that can accomplish
this procedure.
2.1. Basic Principles
Direct Delivery
The selection mechanism is located directly on the manipulator of the placement
mechanism. In combination with a vacuum probe or a mechanical gripper, this means that
the probe is able to move to any position where a seed should be placed. In the case of an
electrostatic valve or a gravity-based selection mechanism, the seedboxes would also need to
be placed on the manipulator. This limits the number of seedboxes that can be carried and
the moving mass would be increased.
Air Pressure
The idea is that the selection subsystem picks up the seed and places it into a
tube. Then air pressure is applied and the seeds are shot to the placement device. This has the
advantage that the seedboxes do not have to be placed on the moving manipulator thus
reducing the moving mass and the energy consumption. Another advantage is that additional
seedboxes, eventually based on a diferent selection principle, can be added to the system
easily by extending the tube.
2.1.3
Placement Principles
This subsystem has to place the seeds at the specified position. The usage of the
principle depends also on the seed type which can either be required to lie on the ground, to
be slightly covered with soil or to be under ground in order to germinate. Further it is
reasonable to add a sensor which detects whether a seed has been placed correctly or not.
Gravity
Gravity would be the simplest proposition as it does not require a complex mechanism, but it is only suitable for seeds that are to be lying on ground. For seeds that need to
be covered with soil, the coverage could be applied manually or with another manipulator in
a second step.
Drilling
An obvious way to bring a seed under ground is to drill a hole with a borer or
to strike a hole with a rod. In both solutions the seed would then be dropped separately into
this gap. It is required to ensure that the hole does not fall together before the seed is placed.
Possibly the bore has to be closed again after the planting of the seed. Hence the mechanical
structure of the manipulator has to be extended.
Air Pressure
The air pressure - e.g. from an air pressure based delivery - can be used to create a
little hole in which the seed is shot into. This requires a strong air pressure source and might
fail if the ground is too hard (e.g. due to stones). Another risk would be that seeds already
placed nearby could be displaced by the pressure exhaust or by material being blown away.
2.2
Concept
As a basic concept, a combination of the following principles was chosen:
Select a seed using gravity through a controllable downward opening in the

seedbox
Deliver the seed to the ejection port by applying air pressure

Eject the seed directly onto the ground and detect falling seeds
The main reason for choosing these principles is their simplicity and cost-efectiveness. A further advantage of the gravity based selection, the extendability of the system, was
another reason for this selection. The simplicity of the air pressure based delivery lies in the
fact that it eliminates the need for a third axis and lowers the moving mass since only a
single tube has to be carried.
2.2.1
Seed Selection Mechanism
This basic concept for the selection mechanism (i.e. the controllable opening) was
developed and refined in three engineering iterations, starting with a very simple mechanism
in the first iteration and correcting drawbacks in the subsequent steps. Further the ejection
was adapted to reduce the drawbacks given by the air pressure delivery system.
Simple Slider
The first approach is a simple slider opening mechanism as shown in fig. 2.2. When
the slider is opened a small, but not exactly defined, number of seeds fall out of the seedbox
into the guide. Then the slider is moved forwards thus pushing the seeds into the delivery
tube and closing the opening. A first hand-made functional prototype gave good results, but
then the precisely manufactured prototype showed the critical point that the number of seeds
is not really limited. This gave a high variance in the number of ejected seeds, which was
not desired.
Slider with Limited Opening
Based on the experimental results of the first iteration, the opening volume was
limited in the second draft as shown in fig. 2.3. The limitation of the opening volume
resulted in a significantly lower variance in the number of seeds that are selected. The
corresponding prototype was not manufactured as its parts would have been too complicated
requiring the milling of very small components which would be feasible, but complicated
and costly.
Seedbox
Air Pressure Tube
Slider
p = p0
p = pmax
Figure 2.2: Slider
2.2. Concept
Seedbox
Air Pressure Tube
Slider
p = p0
p = pmax
Figure 2.3: Slider with limited opening volume

Seedbox
Control Pressure
p = p0
p = pmax
Slider
Delivery Tube
p = p0
p = pmax
Figure 2.4: Slider with limited opening volume and control pressure tube
Extended Slider
The extended slider mechanism as shown in fig 2.4 is the final design. In contrast to the
solution above, it has a vertical bore which allows easy manufacturing. The bore is a critical
point as the seeds may get stuck in it. Opposed to the second solution the opening cannot be
cleaned by the delivery pressure as the bore axis is across the direction of pressure. Therefore
a separate control pressure tube is added, which allows the slider to be cleaned before the
delivery pressure is applied. Further it is easily adaptable to diferent seed sizes, as the hole
size of the exchangeable slider can be manufactured with diferent diameters.
2.2.2
Ejection Mechanism
The seeds are ejected through a vertically fixed tube. In this tube, small holes to reduce the
air pressure are integrated in order to slow the seeds down before being placed on the
ground. After first tests, it has been observed that this is not sufcient. Therefore a blocking
mechanism is added, which stops the seeds completely and ejects them after the air pressure
has been reduced. Further a light barrier system is mounted at the ejection point. This sensor
can detect whether a seed has been placed or not to close the control loop.
10
2.3
2.3.1
Engineering
Selection Mechanism
Mechanical Design
After the concept has been developed, the definitive design of the selection mechanism has been engineered as illustrated in fig. 2.5. Since it is expected that a growing
image can be created with four to five types of diferently coloured owers, there are five
identical modules integrated into the system. The main element is the ground plate consisting
of an upper and a lower part containing both the delivery tube and the control tube, as well as
the guidances for the sliders of all five modules. The sliders were manufactured with three
diferent hole sizes with diameters of 2mm, 3mm and 4mm to be easily exchanged according
to the seed type used in the corresponding seedbox and the desired number of seeds to be
ejected. All parts have been fabricated using a 3D-printer.
To close the seedboxes, a cover and a lower closure have been designed. The cover
is required as there is pressure going from the control tube into the seedbox. This mixes the
seeds up and lowers the probability of a seed getting stuck. The only drawback is that seeds
might get blown out of the box if no cover is used. A lower closure is used for exchanging a
non-empty seedbox. If the operator wants to exchange a seedbox the lower closure can be
closed manually and the seedbox can be dismounted from the ground plate.
For the slider movement actuation, monostable electromagnets of the type Tremba
HMA 2622z.001, operated at 24VDC, are used. These magnets pull the lever in when
positive voltage is applied. Once it is in its retracted position it is held there by a permanent
magnet. If a negative voltage is applied, the lever is released and shifted out by the spring up
to an external mechanical blocking device. These magnets have a nominal shifting distance of
13mm, but only about 8mm are used in the selection mechanism as the retracting force
decreases dramatically with increasing shift. The movement range is determined by the slider
reaching the end of the guidance in the ground plate.
Electronical Design
For controlling the electromagnets and pressure valves an Arduino Mega 2560 Microcontroller with a custom built driver board is used. To provide the needed volt- ages for
the electromagnets, dpdt (double pole, double throw) relays were selected. Logic-MOSFET's,
resistors and diodes are needed for controlling and switching the relays. Each slider
mechanism of the seed selection system requires its own module of the above described
elements. One module of the driver board can be found in fig. 2.6. Further the control of the
air pressure valves and the optional vibrating motor intended for the prevention of slider
jamming are also integrated into the driver board.
2.3.2
Ejection
Mechanical Design
First experimental results showed that the direct ejection from the delivery tube
onto the surface is not very precise. If the gap between the tube and the surface is too high,
the ejected seeds get blown of by the pressure exhaust. On the other hand if the gap is too
small, the robot cannot move unhindered and is likely to dislocate already placed seeds.
Therefore the mechanism illustrated in fig. 2.7 was designed to slow down the seeds at the
ejection. In this mechanism, the ejection tube is
2.3. Engineering
11
Cover
Seedbox
Closure
85mm
Electromagnet
Control Tube
Slider
Delivery Tube
Upper Ground Plate

Lower Ground Plate
Figure 2.5: Seed delivery module for multiple seed types

closed by the blocker when the delivery pressure is applied. The seeds fall first onto the
blocker which is controlled by an electromagnet and the air pressure is reduced by the holes
in the tube in front of the blocker. Once the delivery pressure valve is closed and the pressure
in the tube is reduced, the blocker opens the ejection port and the seeds fall slowly onto the
surface through gravity. With this mechanism, the gap between the ejection tube and the
surface can be significantly higher.
Electronical Design
The control electronics of the ejection opening magnets are designed identically as
the control electronics of the seed dispenser magnets since the same magnet type is used. To
ensure the functionality of this magnet and to reduce inductive efects the corresponding
driver was not integrated into the general driver board. It is placed on a secondary board
located on the moving manipulator near the ejection subsystem as feedback couplings on the
Arduino board have been observed. To further improve the reliability of the seeding process,
an ejection sensor consisting of two pairs of an IR-photodiode and an IR-phototransistor each
as illustrated in fig 2.8 were added. These light-barriers detect the falling seeds and close the
loop by returning feedback to the planting mechanism. Thus, if no seed was planted, the
seeding process can be repeated until a seed gets detected. Since two slightly
shifted light-barriers are used in parallel, the full outlet diameter is supervised.
12
Figure 2.6: Relay driver
2.3. Engineering
13
From Delivery Tube
Air Pressure
Reduction Holes
Electromagnet
Blocker
Ejection Sensor Light Barriers
Seed Ejection
Figure 2.7: Seed ejection mechanism
14
Figure 2.8: Ejection sensor schematic
2.4
2.4.1
Extendability and Possible Improvements

More Seed Types
The seed delivery system is extendable to more seed types than in the current stage.
The latest selection mechanism already ofers the possibility to be adapted to seed types of a
diferent size class. If for any reason, a desired seed type does not comply with the slider
mechanism, but requires another selection principle, an additional handling mechanism can
be designed and added to the system by attaching it in series with the current design to the
delivery tube. Therefore, the commonality of the two mechanisms would be the delivery
through air pressure.
2.4.2
Sensors on the Sliders
A further improvement could be the addition of sensors at the end of the electromagnets to determine whether the single sliders reach their end positions or not. This would
make sure that the seeds in the seedboxes have the highest probability to fall into the hole of
the slider mechanism and get blown of to the outlet. Several tests showed that if the sliders
reached their end positions, a seed gets planted with a very high probability. Under certain
conditions the electromagnets are not strong enough to pull the sliders back and therefore do
not reach their optimal positions. This occurs when longish seeds fall into the hole of the
slider and get stuck there. Reasonable sensors would be circuit breakers or light barriers.
Chapter 3
XY-Table for Tiling

A variety of possible planting concepts has been discussed. The usage of a water soluble foil and spray glue
is proposed to create a tiled growing ower image. Further the mechanical engineering of a static planting
mechanism that integrates the seed delivery system developed in the previous chapter is presented. The design
of a
corresponding electronical system and a control software completes this chapter.
3.1
Concept
A static robot is to be built which plants five diferent kinds of seeds onto a substrate selected
below, using the seed delivery system proposed in the previous chapter. The substrate is
placed underneath the manipulator of the XY-Table. If the chosen substrate requires
additional adhesion, a small amount of dissolving and bio-degradable glue is applied to
prevent the displacement of the seeds. Finally a thin layer of soil or a second sheet of
substrate is placed on top of the seeds to allow transportation and storage.
3.1.1
Substrate
In table 3.1 a set of diferent possible materials, such as PVA 1, organic nonwoven
fabric, paper or soil, is discussed. The trade-of between handling simplicity, cost efectiveness
and ecological compatibility leaded to the selection of the PVA foil 2 available in small
quantities at [17]. Its major advantages are the easy handling and the good water solubility.
Further it is cheaper than the nonwoven fabric and biologically more compatible than the
paper solution. The limitation on the applicable glue types is insignificant as there exist
various adhesives suitable for PVA. Direct application on soil was discarded due to its
heavy weight and the difculties for storage without germinating seeds.
3.1.2
Glue
Depending on the substrate diferent glue types would be possible. The most important criteria are:
Bio-degradability
Harmlessness to the seeds
1Polyvinylalcohol
2Product name: Solvy from Sulky G utermann. This product is commonly used as a stabilizer for quilting
applications.
15
16
Chapter 3. XY-Table for Tiling
Method
Advantages
Soil
PVA foil
Water
soluble
nonwoven fabric
/ organic eece
Paper
Disadvantages
- Greenhouse growing
possible
- Plant the tiles as soon
as the owers are
grown
- No materials that
could harm the environment
- Needs a frame to hold

the soil together
- Heavy
- Difcult for transportation and storage
- Water soluble
- Very light and thus
easy to transport and
store
- Possibility for printing
endless sheets if the foil
is provided on a
roll
- No remaining parts in
the ground
- Limited glue types (no

water based glues applicable)
- Costly for small quantities
- Water soluble
- Very light and thus
easy to transport and
store
endless sheets if the foil
is provided on a
roll
- No remaining parts in
the ground
- Stronger than the PVA
foil
- Easy to transport and
store
endless sheets if the
paper is provided on a
roll
- Several glue types applicable
Table 3.1: Substrates
- Limited glue types (no

water based glues applicable)
- Dissolves slower than
the PVA foil
- Very expensive
- Needs more time to

dissolve
- Leaves chemicals on
the ground
- Seeds must be more robust to grow through
the sheets
- Possibly not entirely
bio-degradable
3.2. Engineering
17
Water solubility
Simple application
Fixing the seed position
Some ideas are: Water soluble glue spray, corn-starch-water-mix, corn-starch-olive oil-mix,
egg white, hairspray etc. An important point is that the glue dissolves in water such that the
seeds are not enclosed in a glue shell and the germination is suppressed therefore. Also there
shall not remain toxic deposit that contaminates the soil. As proposed in [10] corn-starch
would be a good solution.
During experimental phases, it turned out that the optimal ratio between water
and starch is hard to find. Too much water means easier application, but dissolving substrate
and reduced adhesion. More starch leads to difculties during application. Also the water
based glue seems not to be optimal for storing sheets since it dissolves the PVA foil over
time. An alternative could be to replace the water by oil since it does not attack the substrate,
but has a reduced adhesiveness without additional ingredients. Further, mould formation has
been observed that may harm or suppress the germination of the seeds.
Therefore, the water soluble glue spray3 was selected since it meets all the demands
and is readily available at [17]. This glue is especially made for the application with the PVA
foil, hence it overcomes the drawback of early dissolving of the substrate and has the same
properties as the substrate itself. For further improvements of the planting robot, it would be
possible to apply the glue automatically by the manipulator in order to increase the level of
automation and simplify the planting process.
3.2
Engineering
3.2.1
Mechanical Design
An overview of the platform is given in fig. 3.1. The static robot shall be able to
plant seeds on a quadratic area of 1m 2. There are several standard robot configurations to
achieve such a task: SCARA, delta or Cartesian manipulators would all fit the requirements.
For simplicity a Cartesian robot in an H-configuration was chosen. The H-configuration
ofers more stability than other Cartesian setups and can be built of relatively cheap standard
elements.
Actuation
For the actuated axes linear tooth-belted rails of the type IGUS DryLin ZLW 1040 Basic 02
are used. In X-direction, the two axes are connected by an aluminium rod to ensure
synchronous movements. The required torque was calculated using figure Abb. 07 on the
datasheet. The load was assumed to be around 10kg, which includes the ejection device as
well as the mass of the Y-axis for the X-axis-actuation. With an applied torque of 1.4Nm
accelerations up to 5ms2 can be achieved. For simplicity the same motor type shall be used for
both axes, thus the requirements on the X-axis are more restrictive as the load on the Y-axis
is lower. Further, the axes and the motors, as well as the axes and the aluminium rod are
connected with radial couplings. The axes have a shifting distance of 1.1m in order not to
drive into limits when using the specified shift of 1m.
3Product
name: Spr uhfixierer Gold Zack 250 ml
18
X-Axis connection
tin
Plan
X-Motor
Y-Axis
rea
gA
Ejection
2nd X-Axis
Limit Switches (Y)

Y-Motor
1st X-Axis
Limit Switches (X)
Seed Dispensers
Figure 3.1: XY-Table (Overview)
Pneumatic System
The pneumatic system consists of a pressure reduction device, two electromagnetically controllable valves and a pressure distributor. The valves are operated at 24VDC and
are controlled by the Arduino Mega microcontroller. The pressure reduction device allows
the regulation of the system pressure in order to adapt for diferent seed properties. A tube
that can be attached to any pressure source with a standard air pressure connector is used as
the input to the pneumatic system. Thus the system can be reused on the mobile platform
with an air pressure cartridge. All components are from Festo and were dismounted from an
unused project.
Structure
The structure is based on Item aluminium profiles of the size class 6. There are
two parallel longer profiles that compose the base structure for the shorter profiles mounted
across. On the first two crossing profiles the X-axes as well as the limit switches are attached
using a specially designed mounting structure. The Y-axis is connected to the sliding
carriage of the X-axes by an intermediate plate on which also a second pair of limit switches
is mounted. On the Y-Axis no supporting structure is used. Between the axes a wooden
plate is installed as a surface for the placement of the substrate. The remaining crossing
profiles serve as a mounting structure for the seed dispensers and all other auxiliaries. No
connections have been considered critical, so no calculations were conducted for the
described design. All mechanical parts have been designed in accordance with the industrial
standards as defined in [12].
3.2.2
Electronic System
The electronic system of the XY-Table is illustrated in 3.2. The system integrates
the motors, the corresponding controllers and its auxiliaries like the limit switches as well as
all custom made electronics as described in section 2.3. The system has two DC power links,
one at 5VDC and one at 24VDC, coming from an external source. All electronics, unless
required otherwise by the functionality, are placed in a row besides the planting area.
Computer
ElectroMagnet 3
ElectroMagnet 4
ElectroMagnet 5
ElectroMagnet 1
ElectroMagnet 2
USB
Delivery
Pressure Valve
EPOS2 (X)
EPOS2 (Y)
Emergency Stop
On moving platform
Figure 3.2: XY-Table system schematics
24VDC
5VDC
DIO
Arduino
Driver Board 1
DIO
Control
Pressure Valve
ElectroMagnet 6
Driver Board 2
Ejection
Sensor
min
Limit Switch
max
Limit Switch
Y -Motor
min
Limit Switch
max
Limit Switch
X-Motor
3.2. Engineering 19
USB CAN-Bus
20
Motors and Controller

The linear axes are driven by a Maxon RE30 60W DC motor each. To provide
the needed torque and precision, the two motors were supplied with planetary gear heads
(23:1 reduction). This gives a maximum torque of 85mNm 23 = 1.955Nm and a nominal
speed of 8500rpm/23 = 369.57rpm. Combined with the transmission ratio of the axes of
66mm/turn, this yields a nominal linear speed of 0.41m/s. Further the motors are equipped
with MR encoders with 1024 counts per turn. Since the motor controllers use quadcounts
(qc), this further increase the resolution by a factor of four. Thus leading to a positioning
resolution as calculated in formula 3.1, which is - even combined with the precision of the
axes themselves of0.35mm - significantly higher than required as the plants cannot be
expected to grow perfectly vertical. Therefore no linear position sensors are needed.
66 mm turn
23
red = 7.006 104 mm
4qc 1024
Inc
1
turnInc
(3.1)
The motors are each controlled by a Maxon EPOS2 24/5 position controller and
are equipped with several safety elements: Both axes are limited in the movement by an
inductive limit switch at each end. For the limit switches, inductive sensors of the type
Wenglor IH020BK41VD were integrated into the system. These sensors operate at 24VDC
and have a normally closed behaviour. This ensures that the movement is also blocked, if a
power failure occurs on the sensor circuit. As an additional safety element an emergency
stop button was added which immediately suppresses all movements.
3.3
Software
A graphical user interface (GUI) was designed in Microsoft Visual Studio 2010 in C++/CLI
to set several planting properties like the field size, ower distance or picture and colour
selection. To control the electromagnets and the valves as well as to read the ejection sensor
data, a script in the Arduino alpha environment, an IDE coming with the Arduino Mega
2560 microcontroller, was written. The programming language of this IDE is based on C++.
3.3.1
Graphical User Interface
In figure 3.3 the graphical user interface (GUI) is shown. The main screen is divided
into three parts. On the left hand side the most important buttons for controlling the
application and the seeding process are available. In the main window on the top right
position, the selected picture is displayed. Beneath, an image of the actual planting picture
can be previewed such that the settings like ower distance and field size might be adjusted
before planting. Also it is possible to add or delete single seeds of the final picture by
clicking into the picturebox and selecting the desired colour from the colour bar at the
bottom.
If the open button is clicked, an open file dialog appears. The software accepts all common
image formats and displays them in the upper picturebox. While opening the file, it is parsed
into a pixel-array according to the predefined settings selected in the panel on the left. Thus,
the open file dialog also accepts txt-files with the
following content which is already in the format of the pixel-array:
X-Position (mm), Y-Position (mm), ARGB-Colour-Value XPosition (mm), Y-Position (mm), ARGB-Colour-Value
3.3. Software
21
Figure 3.3: Graphical user interface

Further the user interface can read dat-files that list one planting point per line in the form
X;Y;T, where X and Y are the position in millimetres and T is the seedbox
index (1-5). The file may also contain comment lines, starting with a #. These
comments are displayed when loading the file and are intended to instruct the user,
which seeds to fill into the boxes.
3.3.2
Arduino Microcontroller
As written above the Arduino microcontroller can be easily programmed in the

SDK Arduino alpha. At startup of the Arduino, the initialization routine starts and defines
which pins are used for output or input. The startup procedure also initializes the interrupts
used to read the ejection sensor data. Thereafter a polling algorithm begins to monitor the
serial interface for incoming instructions.
3.3.3
Planting Process
As illustrated in fig. 3.4 the planting process consists of two parallel tasks. The
GUI software starts the process by sending the planting command consisting of position and
type of a single planting point. While the manipulator is moved to
22
the goal position, the Arduino task selects the seed and opens the valves to deliver it to the
ejection port. Once the position has been reached the Arduino opens the ejection and reads
the sensor data. If a seed has been detected the planting was successful and the GUI sends
the command for the next planting point. Otherwise the procedure is repeated. If the
repetition was not successful, the operator is requested to check the system and can decide
either to continue with the next seed or to repeat the current process.
The planting process takes 4.1 seconds per planting point under the assumption that the
position is reached in 2.6 seconds, which is the time that the process needs to select and
deliver a seed to the ejection. This time corresponds to a driving distance of 0.497m as
calculated in equation 3.3 which is unlikely to be used in ower images with a reasonable
resolution.
23
4000rpm 66 mm
turn
red 60sec
m 2.6sec
0.191 sec
0.191 sec
(3.2)
0.497m
(3.3)
The overall planting time is composed of the time needed to establish the magnetic
field of the slider actuators and the electromagnetic valves. Further, the relay may not be
switched too soon after the power has been turned of by the MOSFET, as the inductive
currents of the magnets need to be dissipated by the yback diode first in order not to damage
the electronics and to avoid feedback coupling. Other delays come from the time to reliably
shoot the seeds from the dispenser to the ejection by the air pressure, and from the ejection
mechanism itself. The latter includes the opening of the ejection port and the detection of
the seeds by the sensor system. Assuming a tile of a hundred seeding points (10cm spacing),
the process to plant a square meter takes approximately seven minutes when no failure on the
system occurred. Related to a growing time of about three months, this planting time is
insignificant.
3.3. Software
23
Position and Seed

Type Input
EPOS2
Position
Controller
Arduino
Open Slider
Motors
Electromagnets
(Seeddispenser)
Open Valve
Control Pressure
Valve
Close Valve
Close Slider
Open Valve
Delivery Pressure
Valve
Close Valve
Wait until
Position reached
Arduino
Open Magnet
Ejection Opening
Electromagnet
No
Close Magnet
Seed ejection
detected?
Yes
No
Yes
Nr of trials 2
Failure
Success
Figure 3.4: Planting process owchart
Ejection Sensor
24
Chapter 4
Testing
This chapter presents the results from a testing phase that has been conducted to verify the performance of
the XY-Table. The focus lied on the reliability, precision and repeatability of the system. All tests showed
that the specifications are completely satisfied.
4.1
Setup
The seed delivery system has been tested with diferent ower seed species. The placement
procedure as defined in section 3.3.3 has been used on a regular grid pattern consisting of 64
placements (See fig. 4.1). For each place, the number of ejected seeds was counted and their
distance from the reference point was divided into the classes separated at0.5cm, 1cm,
1.5cm, 2cm, 2.5cm. During the tests, the performance of the ejection sensor has also been
evaluated.
4.2
Number of Ejected Seeds
Two species with diferent seed sizes have been planted by the manipulator, using three
diferent diameters of the slider bore: 2mm (S), 3mm (M) and 4mm (L). As a first species,
Viola wittrockiana were tested with slider size S and M, while Phlox drummondii were planted
with sliders M and L. The number of ejected seeds have been counted and illustrated in fig.
4.2. For the Violas, the mean value was 1.63 seeds per ejection for the small slider and 4.55
for slider size M with a variance of 0.43 resp. 1.55. This result is as expected and shows that
the slider size has a major impact on the number of ejected seeds and its variance. The
Phlox experiment resulted in a mean value of 1.73 for the M slider and 4.05 for the large
slider with variances of 0.61 resp. 2.93, which verifies the conclusions of the Viola
experiment. The requirements as defined in section 1.1 are fulfilled by these results.
During the experiments with Violas and the small slider 3 ejection failures (no
ejection in two trials) have been observed resp. 3 failures with phlox (Slider M). In contrast,
no failures have been observed with the larger sliders. This comes from the statistical fact
that the larger slider holes have a higher probability of selecting at least one seed and is
therefore an expected result. To improve the reliability, the bore diameter could be adapted
more precisely to the seed size or a third trial in the planting procedure could be added.
25
26
Chapter 4. Testing
Figure 4.1: Test setup
4.3
Seeding Precision
In the precision tests no glue has been used in order to reuse the seeds and repeat the
experiments. From qualitative observations it is expected that the precision is increased
significantly when using glue. In fig. 4.3, 4.4 and 4.5 the positions of the ejected seeds for
each species are shown. In the left graph the spatial distribution of the ejected seeds and in
the right plot a corresponding histogram are presented.
It is expected that the precision does not vary significantly between species, but
small diferences were observed. This may be due to the seed form and size of the diferent
species. Since Impatiens and Phlox have a similar seed form, but diferent sizes (See section
1.1.2), and show similar results, it can be concluded that the seed form has a higher impact
on the precision than the seed size. This is verified by the comparison between Violas and
Impatiens that have a similar size, but diferent forms.
In fig. 4.6 the results from all experiments were combined and displayed. About
67% of the seeds were ejected in a circle with a radius of 1cm and about 85% within 1.5cm,
which complies with the specification (See section 1.1).
4.4
Ejection Sensor Performance
During the experiments the ejection sensor performance has been monitored and
the results have been divided into the following classes:
True positive (tp): Ejection & Detection

True negative (tn): No ejection & No detection
False positive (f p): No ejection & Detection
0
4
5
6
7
Number of Seeds
10
10
20
20
30
30
Viola, Slider M
4
5
6
7
Number of Seeds
40
40
10
20
30
40
50
50
Phlox, Slider L
Percentage Percentage
50
10
20
30
40
50
4
5
6
7
Number of Seeds
Viola, Slider S
4
5
6
7
Number of Seeds
Phlox, Slider M
4.4. Ejection Sensor Performance 27
Percentage Percentage
Figure 4.2: Number of ejected seeds per position for diferent species and slider sizes
28
Chapter 4. Testing
Distribution
Histogram
50
45
40
Number of Seeds [%]
cm
0
1
2
35
30
25
20
15
10
5
3
3
0
cm
Class
Figure 4.3: Position of ejected Viola seeds

Distribution
Histogram
50
45
40
2
Number of Seeds [%]
35
cm
0
1
2
30
25
20
15
10
5
3
3
0
cm
4
Class
Figure 4.4: Position of ejected Phlox seeds

Distribution
Histogram
50
45
40
2
Number of Seeds [%]
35
cm
0
1
2
30
25
20
15
10
5
3
3
0
cm
Figure 4.5: Position of ejected Impatiens seeds
4
Class
4.4. Ejection Sensor Performance
29
Distribution
Histogram
50
45
40
Number of Seeds [%]
cm
0
1
2
35
30
25
20
15
10
5
3
3
0
cm
Class
Figure 4.6: Position of ejected seeds from all measurements

Species
Slider
tp
tn
fp
fn
Viola
Viola
Phlox
Phlox
Impatiens
S
M
M
L
S
56
64
64
64
63
19
2
4
0
2
1
0
0
0
0
13
1
1
0
21
0.98
1
1
1
1
0.81
0.98
0.98
11
0.75
0.84
0.99
0.99
0.76
Table 4.1: Ejection sensor performance. tp: True positive. tn: True negative. f p: False
positive. f n: False negative. p: Precision. r: Recall. a: Accuracy.
(Definitions see text)
False negative (f n): Ejection & No detection

From these values a performance analysis has been conducted. For each experiment
the precision has been calculated using p = tp/(tp + f p) and the recall using r = tp/(tp + f n).
Precision and recall values are plotted in the fig. 4.7. Further the accuracy, defined as a = (tp
+ tn)/(tp + tn + f p + f n), has been computed for each experiment. As expected, larger seeds
and a higher amount of seeds lead to a better performance of the sensor. This is proved by the
high accuracy, precision and recall values as listed in tab. 4.1 in experiments with a high
number of seeds or by a large sized species. The efect of the number of seeds can especially
be observed in the two Viola experiments where the accuracy difers by 15% for the lower
number of seeds. With Phlox the number of selected seed is not important, since the size of
Phlox seeds is large enough to be reliably detected even for individual ones. In the Impatiens
experiment, the performance is worst, as the Impatiens have the smallest seeds and also are
ejected in low numbers.
30
Chapter 4. Testing
Precision/Recall of the Ejection Sensor

1.05
0.95
Recall
0.9
0.85
Viola, Small Slider

Viola, Medium Slider
Phlox, Medium Slider
Phlox, Large Slider
Impatiens, Small Slider
0.8
0.75
0.7
0.7
0.75
0.8
0.85
0.9
Precision
0.95
1.05
Figure 4.7: Precision and recall for the ejection sensor for violas with two diferent sliders.
Larger numbers of seeds or larger sized seeds lead to a better sensor performance.
Chapter 5
Mobile Platform
Based on the static solution a concept for an autonomous mobile robot was develo - ped. First, the
prerequisites and the underlying assumptions are discussed. Possible locomotion and localization solutions
were analyzed followed by a concrete proposition that would be able to accomplish the task of
autonomously planting ower images. This includes a size estimation, the integration of a manipulator and
a sensory system for controlling the robot.
5.1
Requirements
This section defines the requirements and constraints on the mobile platform of the gardening
robot.
5.1.1
Terrain Form Constraints
As illustrated in fig. 5.1 the terrain for the mobile platform is approximately rectangular and planar, but not necessarily horizontal. It is characterized by the length
A and width B, by the average steepness , the maximum steepness max, the
convexity C and the roughness D. The specified limiting values are listed below.
Size A, B: usually around 10m each, maximum 20m

Convexity C: As small as possible, maximal0.5m or 0.05 minA, B
Roughness D: As small as possible, maximal0.5m or 0.05 minA, B
Steepness max: maximum0.35rad
5.1.2
Obstacle Avoidance
The field is assumed to be obstacle free, but the robot should choose its path such
that it reduces the probability of displacing or destroying an already placed seed by not
driving into previously planted areas. This does not require obstacle detection sensors but
rather the possibility that the robot can mark seeds as placed on the list of positions.
Nevertheless, depending on the selected sensor system (e.g. LADAR) an obstacle avoidance
algorithm can be implemented to allow planting growing ower images on fields that
contain trees, rocks or obstacles of other kinds. For safety reasons, bumper switches can be
added to the robot.
31
32
Chapter 5. Mobile Platform
(A, B)
(0, B)
(A, 0)
(0, 0)
a) Terrain size A, B
max
b) Terrain steepness , max and convexity C
c) Terrain roughness D
Figure 5.1: Terrain constraints
5.2. Locomotion Concept
5.1.3
33
Precision
The robot needs - as the XY-Table - a seeding precision of a few centimetres to create
reasonable ower images. As the seeding manipulator should be able to correct an ofset
between the robot's actual and desired pose to ensure the seeding precision, the localization
accuracy is required to be higher or at least as good as the demanded seeding precision. This
implies a localization precision of a few millimetres up to maximally two centimetres for the
spatial positioning and an orientation error lower than 0.05 radians. Better localization is
desired as small heading errors lead to a large manipulator ofset, but to achieve this
precision, more expensive sensors are required.
5.1.4
Seeding Constraints
The robot can operate either in a single- or a multipass behaviour. In the latter
case, the robot passes the field multiple times with a diferent set of seeds each time, whereas
in the former procedure the robot passes the field only once with the entire set of seeds. The
multipass operation mode is especially used, when seeds of species with significantly diferent
germination and growing durations have to be planted or when the number of diferent seeds
exceeds the number of seedboxes that can be carried on the robot. The single pass mode is
useful e.g. when diferent coloured seeds of the same species have to be planted.
5.2
Locomotion Concept
There are miscellaneous locomotion concepts that are commonly used in robotics. These
possibilities have been evaluated for the suitability to an outdoor planting application. The
maneuverability in the specified terrain, the stability and the low impacts on the environment
are used as criteria.
5.2.1
Evaluation
Unless cited otherwise, details on all described locomotion principles can be found
in [8].
Wheels
Wheels are the most common locomotion system used in mobile robotics, since it is a very
simple approach and does not need complex low-level controls. For outdoor applications,
several ready-to-use platforms exist. Further the wheel size and tyre profiles can be adapted
to the terrain needs. Stability and relatively well-defined odometry measurements are
advantages. For multipass applications, a precise path planning and a sensor system is
needed in order not to damage or displace any previously placed seeds.
Tracks
The main advantage of tracks is the possibility to operate in difcult terrains and to maneuver
in limited space. The robot mass is distributed along the entire track thus increasing the
stability and being able to traverse small gaps. This comes at the price of weak odometry data
and slip while rotating. For multipass usage, tracks are not suitable as they are likely to harm
growing plants.
34
Legged Locomotion
Compared to wheels or tracks, legged locomotion needs a complex low-level control structure
and might be unable to move stable on lose ground in steep terrain due to the lower surface
contact area. In multipass operation legged locomotion would be advantageous, since the
legs could be placed precisely between two plants. This would require an accurate sensor
system to detect the plants.
Cable Suspended System

As seen in sports events, a camera system (See [18]) is mounted on several cables and can be
maneuvered over a playing arena in three dimensions. This principle could be adapted for the
planting of growing ower images by exchanging the camera with a seed delivery system.
The main advantages would be the planting precision, the possibility to add plants at any
growing state of the already placed seeds without harming them and the redundancy of an
external localization system. The major drawback is the cost and efort to setup the system
as it needs fixed poles at the corners of the field for the cable system.
Airborne
An airborne seed delivery system, e.g. with a coaxial helicopter or a quadrotor is not
considered as applicable for the gardening robot. The advantage that it could maneuver
independent of the terrain, where the seed are delivered, is almost negligible as the terrain is
considered to be more or less at and without steps. In a multipass operation, an airborne
system would not damage already germinated and growing plants. The disadvantage, i.e. the
fact that the seeds can be blown away by the rotor downwash is quite serious. If the
helicopter is ying at low altitudes to deliver the seed, this applies to already placed seeds
nearby the current position, and if the helicopter is ying at higher altitude in order to
prevent already placed seeds to be dislocated, the current seed is likely to be planted
inaccurately.
5.2.2
Selection
The wheeled locomotion concept is considered as most applicable for the use as specified.
The simplicity, the range of possible configurations and the stability are the main reasons for
this selection. Also the wide usage of wheeled locomotion and therefore the large amount of
public domain information was another decision point. Hence all other solutions are
discarded and the wheeled concept is further developed by proposing the wheel configuration
in the section below.
5.2.3
Wheel Configuration
The most common wheel configurations are platforms with two, three, four or six wheels.
From the usual arrangements as listed in [8], a diferential drive with additional castor
wheels for stabilization is considered to fit the requirements. In contrast to an Ackermann
steered configuration the diferential drive ofers a higher maneuverability and the possibility to
rotate in place. This property also simplifies the correction of yaw errors during the planting
procedure. Configurations with Swedish wheels are not suitable for outdoor tasks as the soil
could easily block the sensitive mechanics of the Swedish wheels.
5.3. Localization
5.3
35
Localization
For the localization and pose estimation of the robot there are several possibilities which are
all considered as standard solutions for robot localization. The applica- bility of these
solutions for the localization of the gardening robot is evaluated in this section. The criteria
are the complexity of the working principle, the required hardware equipment on the robot,
the hardware equipment that is to be set up for each deployment, and the required algorithms.
The solutions below are intended to be used in combination with odometry as given by the
driving motor or wheel encoders.
5.3.1
Evaluation
Unless cited otherwise, details on all described positioning methods can be found
in [8].
Diferential GPS
In a diferential GPS sensor (dGPS) the positioning signal from the satellite-based
Global Positioning System is extended with a signal from a local reference trans- mitter.
Thus the positioning accuracy is around one centimetre, but the system is quite costly,
because it requires to set up a local reference station besides the field and a corresponding
receiver on the robot. An advantage is that the receiver has the position as output data and
no other calculations are required. But there is no information about the orientation and
therefore a compass has to be added to the system.
External Laser System
In an external laser system a laser tracker is positioned at the origin of the local
coordinate frame. This tracker uses angular and time of ight measurements with a laser
beam that is reected by a mirror on the robot, to determine the robot's position. The only
way to estimate the heading of the robot with this method is to diferentiate the position, as
the tracker gives only the position. Alternatives with multiple reectors would be feasible,
but an algorithm to distinguish between the mirrors would be required as well as a method to
ensure that the reectors are not blocking the beam from each other. A compass can be
added to simplify the orientation estimate. Laser tracker systems are costly, but ofer a very
precise localization in the submillimetre range.
LADAR with external passive Beacons
A rotating laser sends out a beam which is thrown back by beacons with retroreective material (mirrors). The reected beam can be detected by a sensor next to the laser
source and the rotation of the module is measured by an encoder. If a beam is detected, the
relative angle of the beacon to the robot is stored. If enough angles are collected, the robot
pose can be calculated by simple triangulation. The system requires that the physical location
of the beacons and the map stored in the robot are highly correlated in order to minimize the
error. This can be achieved either by precise placement of the beacons and/or that the
positions of the beacons are measured accurately after placement by an operator e.g. using
geodetical metrology instruments. This increases the set-up time dramatically. An alternative
would be a SLAM1 approach which automatically builds and updates a map.
1Simultaneous
Localization and Mapping
36
Assuming that the rotating laser system is equipped with a 16bit angular encoder, a
measurement error of one encoder tick corresponds to a localization error of almost one
millimetre in a viewing distance of 10m, when assuming sin(x) = x for small x.
2rad = 2rad
65536
216
6
95.874 10 rad 10000mm
=
=
95.874 106rad
0.958mm
(5.1)
(5.2)
On-Board Vision System

Similar to the rotating laser solution an omnidirectional camera or a multiple camera
system which covers the full 360 degrees field of view, or a rotating camera is mounted on
the robot. In the corners of the field beacons with easily detectable tags, such as ARTags
[2], are placed. With standard computer vision algorithms the relative position of the tags to
the robot can be calculated and then the same calculations as in the rotating laser solution
are conducted to find the robot pose. The problem is that the resolution of the camera has to
be reasonably high to detect the tags over the entire field.
Assuming a setup of three full HD cameras with a 120 degrees field of view each is used, the
required tag size is about 30cm. This viewing angle per pixel is calculated
as
2/3 rad = 1.091 103 rad
1920px
px
On a distance of 10m the width w corresponding to 1px is derived (assuming
sin(x) = x for small x)
w = 1.091 103rad 10000mm = 10.91mm
(5.3)
(5.4)
With a marker size of 30cm, this corresponds to 27px, which is sufciently high
for a good detection rate according to Fiala [2]. The pixel width defines also the positioning
precision.
External Camera
A camera is mounted on a pole next to the field. This camera tracks the robot pose
using a marker on the top of the robot, which allows fast detection and localization. The
required resolution of the camera is quite high in order to ensure a correct localization. An
alternative to a high resolution camera would be a pan-tilt-zoom- camera, but which would
then need again a precise orientation sensor to determine the viewing direction of the camera.
Wireless Ethernet Localization
WiFi localization has been studied by Howard et al.[5] using the signal strength as
an indicator. For this purpose, a signal strength propagation model and a signal strength
map are used. The precision is stated to be around 10cm in the tested indoor environment. It
is expected that the precision could be increased in an obstacle free outdoor environment,
but it is not expected to reach the requirements without very fundamental research on WiFi
localization and an appropriate signal strength propagation model. The system calibration by
the preliminary creation of a signal strength map is considered inefcient, as the system is
intended to be deployed only once a year in the same place. This would imply that the
signal strength is measured by precisely localized measurement tools for each field which
would probably need more efort than the actual seeding process.
5.4. Required Components and Size Estimation
37
Ultrasonic Triangulation
In ultrasonic triangulation, time-of-ight measurements of ultrasonic waves are used to
determine the distance between a beacon and an object to be localized. Ultra- sonic systems
are available at quite low cost, but a major source of error are reections from any
surrounding object, also the ground. This is the main reason, why
ultrasonic localization is not considered very suitable for ground applications.
5.3.2
Selection
The solutions evaluated in the section above are summarized in tab. 5.1. Which solution is
most applicable is mainly defined by the terrain. If the terrain has low convexity and is
rather smooth (low roughness), the rotating laser, as well as an onboard vision system is
most suitable. The steepness of the terrain does not inuence this selection, as both proposed
onboard systems require a more or less planar, but not necessarily horizontal terrain. If the
terrain does not fit these restrictions, the onboard system has to be placed on a balancing
platform. An alternative for a non-planar terrain would be an external system, e.g. an
external vision system or a laser tracker. For further improvements of pose estimation a
compass can be added to the system which significantly enhances the orientation accuracy.
5.4
5.4.1
Required Components and Size Estimation

Overview
It is assumed that the mobile platform consists of the components listed in table 5.2. The robot
would be based on a diferential drive platform with auxiliary passive castor wheels for
preventing lateral and longitudal fall-over and guaranteed stability in rougher terrain. One
passive wheel would be mounted centred in front of the two active wheels and one castor
wheel would be attached to each end of the manipulator axis as showed in fig. 5.2. It has to
be noted, that the ground clearance is high enough to not damage growing plants in
multipass operations. Therefore, a wheel diameter of about 20cm is proposed to allow
contact-free movements of the robot over germinating seeds. Further, soft spiked tractor
tyres would be preferred to create enough traction on lose ground. The width and softness
property of those tyres lead to a large contact area of the wheels which is important for
outdoor locomotion.
In regular planting grids, the distance between single plants is typically set to 10cm
such that the distance between the active wheels has to be set to approximately 40cm.
Therefore the robot can move between grown owers without destroying them. The
LADAR system should be installed as high as possible in order to detect the reective
beacons at any situation, especially in environments with a high roughness. Optionally, it
could be mounted on a balancing platform to compensate for convexity and roughness.
For high level control like path planning, obstacle avoidance and the planting algorithm an onboard laptop would be preferred. This ofers the possibility to simplify the
interaction between operator and robot by using a graphical user interface. Fur- ther a remote
control over WiFi or Bluetooth could be useful for sending commands to the working robot. If
a laptop is installed on the platform, the required hardware would already be available.
Table 5.1: Localization methods summary
Precision depending
on
environment
(buildings,
trees)
and
weather
(clouds)
Draw-
Other
backs
High
Usual
solution,
very
simple
beacons
Communication Terrain limican imply de- tations, bealays
con position
calibration
Very high
-
Very high
-
Cost
Other Advantages
GPS receiver
driver
Rotating
Laser
Geometrical
transformation
Retroreective
marker
Communication
between
tracker
and
robot
dGPS receiver
Required HW
(robot)
Required SW,
Algorithms
Very high
x
x
Retroreective
beacons
Very high
x
Tracker
High
x
Reference station
Accuracy
Location
Orientation
Required HW
(external)
Rotating
Laser
Laser
Tracker
dGPS
Method
bea-
Terrain limitations,
camera resolution requirements, beacon
position calibration
Low
Simple
cons
High
x
x
Beacons
with
easy
detectable
markers
Camera system
Image
processing
and
geometrical
transformation
On-Board
Vision
sys-
Camera system resolution

requirements,
communication can imply
delays
Image
processing
and
communication
from
camera
to
robot
Low
-
Marker
High
x
x
Camera
tem
External Vision
Signal
strength
propagation
model, accuracy can sufer
from ground
reections
Accuracy
can
sufer
from ground
reections
Low
-
Data processing
Data processing
Low
HW already
intended
US receiver
Low
x
US sender
Ultrasonic
Triangulation
WiFi receiver
Low
x
WiFi sender
WiFi Triangulation
38 Chapter 5. Mobile Platform
5.4. Required Components and Size Estimation
39
Mass (per piece)
Volume (w l h)
Component
Count
Chassis
Driving motors
Active wheels
Passive wheels
Driving controller
1
2
2
3
1
5kg
1kg
0.5kg
0.5kg
0.3kg
60 60 150mm
200 200 30mm
100 100 100mm
60 110 90mm
LADAR
Compass
Bumpers
1
1
2
4.5kg
0.05kg
0.3kg
160 160 210mm

10 10 10mm
300 50 50mm
Manipulator
Selection mechanism
Pressure system
Air pressure cartridge
Manipulator controller
1
1
1
1
1
6kg
0.8kg
0.5kg
1.5kg
0.5kg
1300 250 60mm

150 120 90mm
120 170 90mm
100 100 300mm
180 100 100mm
Battery
Main controller (Laptop)
2
1
6kg
1kg
180 80 170mm
300 200 40mm
Total
36.5kg
Table 5.2: Required components
LADAR
Air pressure cartridge
Seed dispenser
Supporting castor wheels
Manipulator
Controllers and batteries

Drive wheels
Figure 5.2: Overview of the mobile platform
40
5.4.2
Manipulator
For the actual seeding process a manipulator is required. This manipulator is composed of one axis as developed in the stationary robot as well as the selection and ejection
mechanism with all the required electronics. As a pressure source an air pressure cartridge
can be attached to the pneumatic system. The use of an onboard air compressor is not
considered applicable due to the high power demands.
Depending on the kind of seed to be planted it has to be considered if the ejection mechanism shall be modified to ofer underground planting. Possibly, some seeds have to
be covered with soil or should be buried to germinate and to protect from environmental
efects like displacements by strong winds or birds. Hence, the drilling mechanism proposed
in 2.1.3 could be implemented.
5.5
5.5.1
Control
Pose Estimation
A simulated environment has been set up in ROS [16] with the gazebo simulator
to test several available localization algorithms. A diferential drive robot has been
equipped with one forward-looking laser scanner. The properties of the scanner were set to
those of a SICK LMS200, i.e. 30m ranging distance and a 180 degrees field of view. The
environment was composed of a horizontal ground plane and five retroreective poles of
15cm diameter at (1,1), (11,1), (11, 2), (11, 11) and (1, 11) meters. The planting field was set to the square
in (0, 0) (10, 10). By choosing a smaller diameter of the poles, the localization error is reduced.
But on the other hand, if the diameter is increased, the feature matching performance is
enhanced. Therefore the optimal pole size depends on the matching algorithm
used. The position estimate has been tested with the amcl (Adaptive Monte-CarloLocalization) ROS-package from the navigation stack with a hand-built map and with a
SLAM approach using the gmapping ROS-package from the slam gmapping
stack. For both tests a proportional control as stated in algorithm 5.1 has been
implemented. This controller first orients the robot to head the goal position, then drives in
this direction and finally rotates the robot in place to the defined orientation. A description
of the parameters is given in tab. 5.3.
Problems occurred for the gmapping approach as it uses polygon matching and
therefore prefers larger features such as non-straight walls to single poles. This led
to a blurred map, hence an imprecise localization. In environments with surrounding objects like trees or buildings, the gmapping algorithm could be suitable. For
general environments, triangulation in a known map using the laser data fused with
odometry and compass measurements would give the best results for both position and
orientation estimate.
5.5.2
Planting
As the mobile robot is controlled by ROS [16], a corresponding node has been
written. This node is composed (See fig. 5.3) of the service PlaceSeed, which can
place a seed of a specified type in a given position. The service has an object of the
class Arduino, which communicates with the microcontroller over the corresponding
character device file, and an object of the class XYTable, which accesses the EPOS2 motor
controllers and the linear axes over the libepos library [15].
As the node can also be used with the stationary robot for testing purposes, there
are two clients: PlaceSeedAt is a simple client, which formulates a single request
to the service, and PlaceTiles, which can place an entire pattern on the stationary
robot. A detailed software documentation for this node can be found in the source
5.5. Control
41
Algorithm 5.1 Position Control Algorithm. Goal Pose: , . Current Pose: x, y

x, y, . Speed Command (linear, rotational): v, . Tolerances:,. Tuning
Parameters: k, k, k, , (See tab. 5.3 for detailed description)
x x x
y y y
x 2 + 2y
atan2(y, x)

while AND do
Calculate the linear speed:
g exp(1/2 2/) p
v g k
Calculate the rotational speed:
h exp(1/2 2/)
if then
Position not yet reached. Orient the robot to head the goal:
k + h k
else
Position reached. Rotate the robot in place:
h 2 k
end if
Limit the output to vmax,max
Send speed command
Delay
Calculate errors:
x x x
y y y
x2 + y2
atan2(y, x)
o
end while
Parameter
k
k
k
Description
Proportional parameter for the linear speed v.
Proportional parameter for the rotational speed due to heading
ofset.
Proportional parameter for the rotational speed due to goal
orientation ofset.
Width of the Gaussian function for the intermediate parameter g,
which decreases the linear speed v for increasing heading ofsets.
Width of the Gaussian function for the intermediate parameter
h, which increases the weight of the goal orientation error in the
calculation of the rotational speed , when the robot is closer to the goal
position.
Numerical tolerance for the allowable goal position ofset.
Numerical tolerance for the allowable goal orientation ofset.
Table 5.3: Position control algorithm parameter description
42
ROS-Client PlaceSeedAt
ROS-Client PlaceTiles
ROS-Service PlaceSeed
Class Arduino
Class XYTable
fstream
libepos
Char Dev File
Dev File
USB
USB-to-RS232
Arduino
EPOS
Figure 5.3: ROS-Node structure for the control of the manipulator
codes. In contrast to the initial control software, this node connects to the EPOS2 over
RS232 (or with an USB-to-RS232-converter), and not over USB since the
libepos library does not yet support USB communication.
5.5.3
Path Planning
Depending on the image that has to be planted the task could be separated into two
subproblems. For multipass operations a regular planting grid is required, else the placed
seeds and germinating plants could be displaced or harmed. For singlepass tasks the image
can have an irregular plant arrangement. In both cases the mobile platform would follow the
algorithm 5.2 as illustrated in fig. 5.4.
Algorithm 5.2 Path planning algorithm. Field size: . Current position: x, y. x, y
while x do x
if > y then y
while y do y
Plant line of seeds
Detect obstacles
Move to next y position in positive y direction
end while
else
while y 0 do
Plant line of seeds
Detect obstacles
Move to next y position in negative y direction
end while
end if
Detect obstacles
Move to next x position and rotate for 180 deg
end while
43
5.5. Control
Figure 5.4: Path planning. Red: Robot movement. Green: Manipulator movement. Blue:
Laser measurement.
44
5.6
Available Platforms
The algorithms proposed above were briey tested on ASL's BIBA robot (diferential drive)
which would not be suitable for outdoor applications as it has small indoor wheels and a low
ground clearance. A more adequate outdoor solution could be ASL's six-wheeled CRAB
Robot or its antecessor CrabLi which is a smaller version of CRAB. However, a diferential
drive robot would be preferable as explained above. Several ready-to-use platforms of
wheeled and tracked type can be found in [19], [20] and [21].
Chapter 6
Conclusion and Outlook

6.1
Conclusion
The goal of this thesis was defined to develop and implement a mechanism that is able to
plant diferent seed types in order to create a growing ower image. The concept has been
proven to be feasible by the design of a seed delivery system and a stationary platform that
can either place seeds on an area of one square meter onto a substrate or onto ground
directly. This prerequisite for the development of an autonomous mobile platform that can
plant the seeds on a custom sized field in any outdoor location has successfully been satisfied.
From the extensive testing phase, it can be concluded that the built system fulfils
the requirements in precision, repeatability and reliability. Because of the exceeding diversity
of seed sizes and forms only a limited number of seed species has been considered. Although
already first tests gave good results, the mechanism has been optimized to reduce the
planting time and to increase the precision without losing performance in repeatability and
reliability.
For planting any kind of seeds, the built mechanisms can be further modified or
expanded. With the seed dispenser mechanism, a simple machinery has been built that can be
used to handle single seeds of diferent kinds. The ability of planting seeds onto a water
soluble substrate ofers a great exibility in the design of the process this system is integrated
into. The substrate can be prepared in another place than the final image shall be grown and
stored for later usage. This can be extended to a commercial application, where a customer
sends in an image of his choice and receives a sheet of prepared substrate that he can plant in
his own garden.
The robotic application of the glue to the substrate or the placement of the substrate (e.g. from a roll), as well as an automatic soil coverage on the manipulator for the
mobile platform have not been implemented as these aspects have not been considered to be
crucial for this early development stage. This would further increase the level of automation
and reduce the process time and cost. For a commercial application this is a point that has to
be taken into account.
6.2
6.2.1
Outlook and Future Work

Engineering of the Mobile Platform
As a next development stage the mobile platform can be realized based on the
conceptual ideas given in chapter 5. The selection of localization and locomotion methods
has to be tested in an environment closer to the final field properties. If
45
46
Chapter 6. Conclusion and Outlook
this was successful, either an existing platform can be chosen or a new one can be designed
from scratch to implement this concept. If needed, the concept can be extended to an
autonomous delivery of fertilizer, water or soil.
6.2.2
Grass seeding
To increase the plant density and contrast of the ower images, the idea of planting grass seeds emerged. Depending on the species of owers, the minimum seed distance
between two plants is defined from 10cm to 20cm, while grass does not have such
restrictions. But the seeding process of grass is diferent compared to the planting mechanism
of owers where only a small amount of seeds are dropped onto the ground. For dense grass
images one has to print patches of grass seeds containing hundreds of single seeds. Thus the
current dispenser mechanism is not suitable for this application and has to be redesigned. A
reasonable solution would be a modification of the vacuum pick-up system as described in
2.1.1. Instead of having a predefined number of holes at the tip of the vacuum tube, a thin
mem- brane and a variable aperture could be mounted. This would allow to adapt the radius
of the efective vacuum tip and therefore to change the amount of delivered seeds. The big
advantage is that this mechanism is applicable to any kind of seed form and size, hence also
for owers. It additionally lowers the required time for planting grass images since a large
number of seeds could be planted simultaneously. The drawback of the vacuum delivery
system (the need for a third axis) and legal concerns given by the patent of Keller et al. [6]
remain.
Nevertheless, a set of grass species has been tested and the following subset may
work with the current robotic planter:
Melica nutans: Work almost perfectly

Molinia gaerulea: Work well
Anthoxandum odoratum: Work
Melica ciliata: Work
Agrosstis rupestris: Work, but too small for reliable ejection detection
The following species do not work or need mechanical adjustments to the robot, due to size,
seed form or slider jamming.
Cynosurus cristatus: Too small (Slider Jamming)

Poa alpina: Oddly shaped form
Alopecurus pratensis: Oddly shaped form
Briza media: Too big, might work with a bigger slider hole
Appendix A
Datasheets
47
maxon DC
motor
RE 3030 mm, Graphitbrsten, 60 Watt
M 1:2
Lagerprogramm
Standardprogramm
Sonderprogramm (auf Anfrage)
Bestellnummern
gemss Massbild
Wellenlnge 15.7 gekrzt auf 8.7 mm
310005
310006
310007
310008
310009
268193
268213
268214
268215
268216
18.0
8170
212
7630
51.7
4.00
844
49.8
87
24.0
8590
164
7900
75.5
3.44
991
39.3
87
36.0
8810
106
8050
85.0
2.20
1020
23.5
87
48.0 2
8590
78.5 4
7810
83.4
1.72 7
936
19.0 9
88
0.362
0.0703
13.9
685
9.74
3.25
35.7
0.611
0.119
19.9
479
8.71
3.03
33.3
1.53
0.281
25.9
369
8.69
3.17
32.9
2.52 11
0.513 12
39.8
240
9.22
3.01 16
34.5
Motordaten
Werte bei Nennspannung
1 Nennspannung
V 12.0
Leerlaufdrehzahl
min-1
Leerlaufstrom
mA 300
Nenndrehzahl
min-1
Nennmoment (max. Dauerdrehmoment)
mNm
Nennstrom (max. Dauerbelastungsstrom)
A 4.00
Anhaltemoment
mNm
Anlaufstrom
A 60.5
Wirkungsgrad
% 86
Kenndaten
10 Anschlusswiderstand
0.198
Anschlussinduktivitt
mH 0.0345
Drehmomentkonstante
mNm A-1
Drehzahlkonstante
min-1 V-1
14 Kennliniensteigung
min-1 mNm-1
Mechanische Anlaufzeitkonstante
ms
3.42
Rotortrgheitsmoment
gcm2 33.5
Spezifikationen
Betriebsbereiche
Thermische Daten
17
18
19
20
21
22
Mechanische Daten (Kugellager)

Grenzdrehzahl
12000 min-1
Axialspiel
0.05 - 0.15 mm
Radialspiel
0.025 mm
Max. axiale Belastung (dynamisch)
5.6 N
Max. axiale Aufpresskraft (statisch)
110 N
(statisch, Welle abgesttzt)
1200 N
28 Max. radiale Belastung, 5 mm ab Flansch
28 N
Von der Standardspezifikation abweichende

Toleranzen.
Option
Vorgespannte Kugellager
1020 8
Max.
53.8 13
178
8.33 15
Legende
Dauerbetriebsbereich
Unter Bercksichtigung der angegebenen thermischen Widerstnde (Ziffer 17 und 18) und einer Umgebungstemperatur von 25C wird bei dauernder
Belastung die maximal zulssige Rotortemperatur
erreicht = thermische Grenze.
Kurzzeitbetrieb
Der Motor darf kurzzeitig und wiederkehrend berlastet werden.
23
24
25
26
27
Motordaten gemss Tabelle sind Nenndaten.

Erluterungen zu den Ziffern Seite 49.
7750 5
88.2 6
n [min-1]
Therm. Widerstand Gehuse-Luft

6.0 KW-1
Therm. Widerstand Wicklung-Gehuse 1.7 KW -1
Therm. Zeitkonstante der Wicklung
16.2 s
Therm. Zeitkonstante des Motors
714 s
Umgebungstemperatur
-30 ... +100C
Max. Wicklungstemperatur
+125C
Weitere Spezifikationen
29 Polpaarzahl
30 Anzahl Kollektorsegmente
31 Motorgewicht
8490 3
1
13
238 g
Typenleistung
bersicht Seite 16 - 21
maxonBaukastensystem
Planetengetriebe
32 mm
0.75 - 6.0 Nm
Seite 229 / 231 / 232
Koaxdrive
32 mm
1.0 - 4.5 Nm
Seite 235
Spindelgetriebe
32 mm
Seite 249 / 250 / 251
Encoder MR
256 - 1024 Imp.
3 Kanal
Seite 263
Empfohlene Elektronik:
ADS 50/5
Seite 282
ADS_E 50/5
283
EPOS2 Module 36/2
304
EPOS2 24/5
305 305
EPOS2 50/5
308
EPOS2 P 24/5
18
Hinweise
80
maxon DC motor
Ausgabe April 2011 / nderungen vorbehalten
Planetengetriebe GP 32 A32 mm, 0.75 - 4.5 Nm

Technische Daten
maxon
gear
Planetengetr iebe
Abtriebswelle
Wellendurchmesser als Option
Abtriebswellenlagerung
R a d i a l s p i e l , 5 mm a b Fl a n s c h
Axialspiel
M a x . z u l . R a d i a l l a s t , 1 0 mm a b Fl a n s c h
Max. zulssige Axiallast
Max. zulssige Aufpresskraft D r
e h s i n n , A n t r i e b zu A b t r i e b
Empfohlene Motordrehzahl
Empfohlener Temperaturbereich
geradeverzahnt
rostfreier Stahl
8 mm
Kugellager
max. 0.14 mm
m a x . 0 . 4 mm
140 N
120 N
120 N
=
< 6000 min- 1 40 ... +100C
M 1:2
Option: Geruschreduzierte Ausfhrung
Lagerprogramm
Standardprogramm
Sonderprogramm (auf Anfrage)
Bestellnummern
166155 166158 166163 166164 166169 166174 166179 166184 166187 166192 166197 166202
Getriebedaten
1 Untersetzung
2 Untersetzung absolut
3 Max. Motorwellendurchmesser
Bestellnummern
1 Untersetzung
Bestellnummern
1 Untersetzung
Bestellnummern
Untersetzung
Bestellnummern
1 Untersetzung
4 Stufenzahl
5 Max. Dauerdrehmoment
6 kurzzeitig zulssiges Drehmoment
Max. Wirkungsgrad
Getriebespiel unbelastet
10 Massentrgheitsmoment
11 Getriebelnge L1*
*fr EC 32 flat ist L1 + 2.0 mm
Gesamtlnge
3.7 : 1
14 : 1
26/ 7
676/ 49
mm
6
6
166156 166159
4.8 : 1
18 : 1
24/ 5
624/ 35
mm
4
4
166157 166160
5.8 : 1
21 : 1
23/ 4
299/ 14
mm
3
3
166161
23 : 1
576/ 25
mm
4
166162
28 : 1
138/ 5
3
mm
2
1
Nm
0.75
2.25
Nm
1.1
3.4
% 80 75
75
g 118 162
162
0.7
0.8
0.8
gcm2
1.5
0.8
mm
26.5
36.4
33 : 1
529/ 16
3
2
2.25
3.4
70
194
1.0
0.8
36.4
51 : 1
111 : 1
246 : 1
492 : 1
762 : 1 1181 : 1 1972 : 1 2829 : 1 4380 : 1

19044/25 10123776/8575 8626176/4375 495144/175 109503/25
4
3
3
4
4
3
3
166175 166180 166185 166188 166193 166198 166203
295 : 1 531 : 1 913 : 1 1414 : 1 2189 : 1 3052 : 1 5247 : 1
101062/343 331776/625 36501/40 2425488/ 1715 536406/245 1907712/625 839523/160
3
4
3
3
3
3
3
166176 166181 166186 166189 166194 166199 166204
318 : 1 589 : 1 1093 : 1 1526 : 1 2362 : 1 3389 : 1 6285 : 1
389376/ 1 2 2 20631/
9345024/
2066688/ 8 7 474513/
6436343/ 1024
6125
35 279841/ 256
140
5
5
4
3
3
4
3
3
3
166177 166182
166190 166195 166200 1
411 : 1 636 : 1
1694 : 1 2548 : 1 3656 : 1
1162213/ 686 7962624/ 3 1 2 457056/
359424/ 875 79488/ 125
125
5
4
3
3
4
3
166178 166183
166191 166196 166201
456 : 1 706 : 1
1828 : 1 2623 : 1 4060 : 1
2238912/
2056223/
3637933/
89401/ 196 158171/
1225
784
896
224
3
3
3
3
3
4
4
5
5
5
4
5
4.50
4.50
4.50
4.50
4.50
4.50
4.50
6.5
6.5
6.5
6.5
6.5
6.5
6.5 7
60
60
50
50
50
50 8 Gewicht
226
226
258
258
258
258 9 Mittleres
1.0
1.0
1.0
1.0
1.0
1.0
0.7
0.7
0.7
0.7
0.7
0.7
0.7
49.8
49.8
49.8
56.5
56.5
56.5
56.5
17576/ 343 13824/ 125 4 2 1 8 2 4/ 1 7 1 5 86112/ 175
6
166165
66 : 1
16224/ 245
4
166166
79 : 1
3887/ 49
3
166167
86 : 1
14976/ 175
4
166168
103 : 1
3588/ 35
3
3
4.50
6.5
70
194
1.0
0.7
43.1
4
166170
123 : 1
6877/ 56
3
166171
132 : 1
3312/ 25
3
166172
159 : 1
1587/ 10
3
166173
190 : 1
12167/ 64
3
3
4.50
6.5
60
226
1.0
0.7
43.1
Gesamtlnge
maxonBaukastensystem
+ Motor
RE 30, 60 W
W
W
RE 35, 90 W
RE 35, 90 W
RE 35, 90 W
A-max 32
A-max 32
EC 32, 80 W
EC 32, 80 W
EC 32 flat, 15 W
232
maxon gear
Seite
+ Sensor / Bremse Seite
Gesamtlnge [mm] = Motorlnge + Getriebelnge + (Sensor / Bremse) + Montageteile
80
94.6
104.5
104.5
111.2
111.2
117.9
117.9
117.9
124.6
124.6
MR
263
106.0
115.9
115.9
122.6
122.6
129.3
129.3
129.3
136.0
81
97.6
107.5
107.5
114.2
114.2
120.9
120.9
120.9
127.6
127.6
MR
263
109.0
118.9
118.9
125.6
125.6
132.3
132.3
132.3
139.0
81
HED_ 5540
266/268
118.3
128.2
128.2
134.9
134.9
141.6
141.6
141.6
81
DCT 22
276
115.7
125.6
125.6
132.3
132.3
139.0
139.0
139.0
81
AB 28
318
133.7
143.6
143.6
150.3
150.3
157.0
157.0
157.0
81
HEDS 5540 / AB 28
266/318
150.8
160.7
160.7
167.4
167.4
174.1
174.1
174.1
109/111 89.5
99.4
99.4
106.1
106.1
112.8
112.8
112.8
119.5
119.5
110/112 88.1
98.0
98.0
104.7
104.7
111.4
111.4
111.4
118.1
118.1
110/112 MR
263
99.3
109.2
109.2
115.9
115.9
122.6
122.6
122.6
110/112 HED_ 5540
266/268
108.9
118.8
118.8
125.5
125.5
132.2
132.2
132.2
153
86.6
96.5
96.5
103.2
103.2
109.9
109.9
109.9
116.6
116.6
HED_ 5540
267/269
105.0
114.9
114.9
121.6
121.6
128.3
128.3
128.3
153
Res 26
277
106.7
116.6
116.6
123.3
123.3
130.0
130.0
130.0
188
44.5
54.4
54.4
61.1
61.1
67.8
67.8
67.8
74.5
74.5
189
54.6
64.5
64.5
71.2
71.2
77.9
77.9
77.9
84.6
84.6
189
56.3
66.2
66.2
72.9
72.9
79.6
79.6
79.6
86.3
86.3
124.6
136.0
127.6
139.0
148.3
145.7
163.7
180.8
119.5
118.1
129.3
138.9
116.6
135.0
136.7
74.5
84.6
86.3
124.6 RE 30, 60 W 80
136.0
136.0 RE 35, 90
127.6 RE 35, 90 W 81
139.0
139.0 RE 35, 90
148.3
148.3
148.3
145.7
145.7
145.7
163.7
163.7
163.7
180.8
180.8
180.8
119.5 A-max 32
118.1 A-max 32
129.3
129.3
129.3
138.9
138.9
138.9
116.6 EC 32, 80 W 153
135.0
135.0
135.0
136.7
136.7
136.7
74.5 EC 32 flat, IE, IP 00
84.6 EC 32 flat, IE, IP 40
86.3
Ausgabe April 2010 / nderungen vorbehalten
MODELL MK1
TECHNISCHE INFORMATION
Eigenschaften:
E
ISO 4029
Bisherige
Lsungen:
Material:
Balg aus hochelastischem Edelstahl, Nabe AL

Balg: Serie 0,5 Tombak
Aufbau:
Naben mit radialen Klemmschrauben

DIN 916 und integrierter Demontagenut
Temperaturbereich:
-30 bis +120 C
Drehzahlen:
Bis 20.000 1/min. ber 20.000 1/min.

in ausgewuchteter Ausfhrung
Lebensdauer:
Bei Beachtung der techn. Hinweise sind die

Kupplungen dauerfest und wartungsfrei
Passungsspiel:
Welle-Nabeverbindung 0,01 - 0,08 mm
Sonderlsungen:
Wie andere Passungen, Passfedernuten,

Sondermaterial und Blge sind kurzfristig mglich
Bestellbeispiel
MK1 / 5 / 26 / 4 / 5 / XX
Modell
Serie
Gesamtlnge mm
Bohrungs D1 H7
Bohrungs D2 H7
Sonder z.B. Naben rostfrei
spielfrei und verdrehsteif

Ausgleich von Fluchtungsfehlern
durch integrierte Demontagenut
kann ein Abflachen der Welle entfallen
preiswerte Ausfhrung
niedriges Trgheitsmoment
Serie
Modell MK 1
0,5
10
15
20
45
100
1,5
2,0
4,5
10
(Nm)
TKN
0,05
0,1
0,5
1,0
(mm)
14
20
23
25
(mm)
6,5
10
15
15
19
25
32
40
(mm)
6,5
6,5
7,5
11
13
15
Sonderbohrung von bis H7 (mm)
D1/2
1-3
1-5
3-9
3-9
3-12
3-16
6-22
6-28
Standardbohrung H7
D1/2
6/10
6/10
10
10
1xM2
1xM2,5
1xM3
1xM3
2xM3
2xM4
2xM5
2xM6
0,35
0,75
1,3
1,3
1,3
2,5
Nenndrehmoment
Gesamtlnge
Auendurchmesser
Passungslnge der Nabe
(mm)
Klemmschrauben ISO 4029

Anzugsmoment
Abstand
Trgheitsmoment
Gewicht ca.
Torsionssteife
axial
(Nm)
20
26
22
28
24
29
26
31
35
37
45
43
53
(mm)
1,5
1,8
(gcm)
0,1
0,4
1,1
1,2
1,3
1,3
1,8
4,7
5,5
15
18
20
65
70
180
220
12
14
22
24
26
54
58
106
114
(g)
(Nm/rad)
CT
(mm)
2,5
3,5
50
70
280 210 170 510 380 320
750
700
5000
9050
8800
0,4
0,4
0,4
0,5
0,6
0,5
0,7
0,5
0,7
0,7
1,2
0,1
1
0,15
1
0,15 0,2 0,25 0,15 0,2 0,25

1 1,5 2
1 1,5
0,15
1,5
0,2
1,5
0,15 0,2 0,25

1,5 1,5 2
0,2
1,5
0,25
2
0,2
1,5
0,3
2
0,5
0,6
0,4
1200 1300 1200 7000
0,6
max.
lateral
angular
(mm) Werte
(Grad)
Integrierte Demontagenut ab Bohrungsdurchmesser 4 H7.
R+W
www.rw-kupplungen.de
DryLin
SHT
DryLin ZLW | Technische Daten | Zahnriemen

Technische Daten
ZLW-0630
Gewicht
Gewicht
ohne Hub [kg] 100 mm Hub [kg]
max.
ber-
Hublnge
setzung
[mm]
[mm/U]
Verzahnung
Zahnriemen
-material
-breite
-spannung
[mm]
[N]
Basic 02
0,38
0,08
1.000
54
HTD 3M
Neopren mit GF
75
Standard 02
0,43
0,08
1.000
54
MTD3
PU mit Stahl
100
m a x . R a d i a l b e l a s tu n g
U m le n k l ag e r
max. Geschwindigkeit
max. Positionsabweichung**
[N]
[m/s]
[mm]
Basic 02
100
Rillenkugellager
0,35
Standard 02
150
Rillenkugellager
0,3
ZLW-1040
Gewicht
Gewicht
ohne Hub [kg] 100 mm Hub [kg]
max.
ber-
Hublnge
setzung
[mm]
[mm/U]
Verzahnung
Zahnriemen
-material
-breite
-spannung
[mm]
[N]
Basic 02
0,9
0,14
2.000
66
RPP 3M
Neopren mit GF
15
150
Standard 02
1,0
0,14
2.000
70
AT5
PU mit Stahl
16
200
m a x . R a d i a l b e l a s tu n g
U m le n k l ag e r
max. Geschwindigkeit
max. Positionsabweichung**
[m/s]
[mm]
Basic 02
200
[N]
Rillenkugellager
0,3
Standard 02
300
Rillenkugellager
0,2
* Grere Hublngen bieten wir gerne auf Anfrage nach technischer Rcksprache und Klrung an.
** Effektiv gemessene Werte am Schlitten bei maximal zulssiger Belastung in horizontaler Einbaulage
Wagenplatte aus eloxiertem Aluminum
Lagergehuse aus blau chromatiertem

Zink, optional einstellbar
Linearlager aus iglidur J oder J200
Nutensteine
PU-Zahnriemen mit Stahlzugstrang

oder Neopren mit GF
Profil aus hartanodisiertem Aluminium

Klemmelemente aus Aluminium bzw. Kunststoff
Rillenkugellager
solides Kunststoffgehuse
igus GmbH 51147 Kln | Tel. +49 (0) 22 03/96 49-145 Fax -3 34 | info@igus.de | www.igus.de 929
DryLin
SHT
DryLin ZLW | Lieferprogramm | Zahnriemen
ZLW 1040 - Zahnriemenachse
Bestellschlssel
Die Zahnriemenachse DryLin ZLW-1040 ist die ideale

Lsung fr viele Positionieraufgaben. Die Einbauhhe
ZLW-1040-02-B-100-L-
betrgt nur 45 mm. Die Hublnge ist frei whlbar (maximal

2.000 mm). Der Schlitten ist in drei Lngen verfgbar.
xxx
DryLin ZLW-1040 ist in den Baureihen Basic 02" und
Hublnge in mm
Standard 02" lieferbar.
L=
Antriebszapfen
links
R=
Antriebszapfen
rechts
L/R = Antriebszapfen
beidseitig
l + Hub
Schlittenlnge in mm,
AI
100 mm
(optional 150/200 mm)

S: Baureihe Standard
B: Baureihe Basic
lt
Version 02 mit
E3
Rillenkugellagern
E
2
Baugre 1040
Zahnriemenachse
l
2
d2
lz
T2
A-A
T
1
s
g
Kernloch fr M6 metrisches Gewinde oder selbstschneidende Kunststoffschrauben, nicht im Lieferum-
h
a
fang enthalten
Abmessungen [mm]
Bestellnummer
AI
-0,3
ZLW-1040-02-
E2
hc
0,15
74
100
45
60
E3
0,15 0,15
204 22,5
87
40
lt
sg
ha
lz
l2
d2*
M6
22
27
20
10
0,3
52
* Basic Version: 6 mm Vierkant, Kunststoff-Adapter fr Zapfendurchmesser 10 mm im Lieferumfang enthalten
Abmessungen [mm]
Bestellnummer
Anschlussmae
x
0,2
Liefer- 2-3 Tage

zeit
AP
LP
dp
-1,0
0,25
0,25 ZLW-1040-02-
nb
Preise Online-Preisliste
www.igus.de/de/DryLinSHT
932 Lebensdauerberechnung, 3-D-CAD-Daten, PDF-Downloads
nw
nh
frei whlbar
T1
60
T2
78
d
40
Bestell- Bestellnummer
beispiel ZLW-1040-02
www.igus.de/de/DryLinZLW
6,4
DryLin
SHT
1,8
Antriebsmoment [Nm]
Antriebsmoment [Nm]
DryLin ZLW | Lieferprogramm | Zahnriemen

1,6
1,4
1,2
1,0
0,8
0,6
1,8
1,6 1,4 1,2
1,0 0,8 0,6
0,4 0,2
0,4
0,2
1
10
15
20
Nutzlast [kg]
10
Nutzlast [kg]
Abb. 07: Erforderliches Antriebsmoment*; Einbaulage

vertikal - ZLW-1040
horizontal - ZLW-1040, Version Basic 02
2,5
2,5
Antriebsmoment [Nm]
Antriebsmoment [Nm]
2,0
1,5
1,0
0,5
0,0
2,0
1,5
1,0
0,5
0,0
10
15
20
25
Nutzlast [kg]
10
15
Nutzlast [kg]
horizontal - ZLW-1040, Version Standard 02
vertikal - ZLW-1040, Version Standard 02
* Annahme: die zu bewegende Masse befindet sich max. in einem umschriebenen Kreis mit R = 100 mm zur Mitte
der Fhrungsschiene, max. zul. Moment, ZLW-1040 Basic 02: 1,75 Nm, a = 0 m/s2 , ZLW-1040 Standard 02: 2,4
Nm, a = 0 m/s2, Konstantfahrt, ohne nennenswerte Beschleunigung
Belastung [N]
Geschwindigk. d. Wagens [m/s]
5,0
4,5
4,0
3,5
3,0
2,5
350
300
0,2 mm
0,1 mm
0,5 mm1,0 mm
250
2,0 mm
200
150
2,0
100
1,5
50
1,0
0,5
0
0
50
100
150
200
250
300
Belastungen des Wagens [N]
500
700
900
1100
1300 1500
1700 1900 2100
Sttzlnge [mm]
Abb. 11: Maximale Belastung im Vergleich: ZLW-0630 und
Abb. 12: Durchbiegung in Abhngigkeit der Sttzweite
ZLW-1040, 100% ED (Einschaltdauer). Die Grafik berck-
ZLW-1040, Version Basic 02 und Standard 02. Maximal
sichtigt die Summe aller auf den Wagen wirkenden Krfte.
2 mm Durchbiegung zulssig - Einbaulage horizontal
igus GmbH 51147 Kln | Tel. +49 (0) 22 03/96 49-145 Fax -3 34 | info@igus.de | www.igus.de 933
In d u k tiv e r N h e r u n g s s c h a lte r
IH 0 2 0 B K4 1 VD
Bestellnummer
Technische Daten
Induktiv
Schaltabstand
2 mm
Korrekturfaktor V2A / Ms / Al
0.73/0.49/0.39
Einbauart
bndig
Einbau A/B/C/D in mm
0/16/6/0
Schalt-Hysterese
Elektrisch
< 15 %
Versorgungsspannung
10...30 V DC
Stromaufnahme (Ub = 24V)
< 6 mA
Schaltfrequenz
1500 Hz
Temperaturdrift
< 10 %
Temperaturbereich
-25...80 C
Spannungsabfall Schaltausgang
< 2,5 V
Schaltstrom Schaltausgang
100 mA
Reststrom Schaltausgang
< 100 A
kurzschlussfest
ja ja
verpolungs- und berlastsicher

Mechanisch
Gehusematerial
Kunststoff
ja
Vollverguss
IP67
Schutzart
Kabel
Anschlussart
2m
Kabellnge
50 V
Schutzisolierung, Bemessungsspannung
PNP ffner
206
Anschlussbild-Nr.
50 V
Maangaben in mm (1 mm = 0.03937 Inch)
1 = S c h a ltz u s ta n d s a n z e ig e
Symbolerklrung
Versorgungsspannung +
Versorgungsspannung 0V
Versorgungsspannung (Wechselspannung)
Schaltausgang (1,2,3,...)/Schlieer
Schaltausgang (1,2,3,...)/ffner
Verschmutzungs-/Fehlerausgang
Verschmutzungs-/Fehlerausgang
Eingang analog oder digital
Teacheingang
Zeitverzgerung (Aktivierung)
Schirm
RS-232 Empfangsleitung
RS-232 Sendeleitung
Bereit
Masse
Takt
Eingang/Ausgang programmierbar
(NO)
(NC)
(NO)
(NC)
Testeingang
Testeingang invertiert
Triggereingang
Analogausgang
Bezugsmasse/Analogausgang
Blockabzug
Ausgang Magnetventil/Motor
Ausgang Ventilsteuerung +
Ausgang Ventilsteuerung 0V
Synchronisation
Empfnger- Leitung
Sende-Leitung
Erdung
Schaltabstandsreduzierung
USB Daten +
USB Daten Schnittstellen-Bus A(+)/B(-)
Sendelicht abschaltbar
Adernfarben nach DIN IEC 757

schwarz
braun
rot
orange
gelb
grn
blau
violett
grau
wei rosa
grngelb
Einbau
A
B
C
D
Sensoren fr Ihren Erfolg www.wenglor.com Sensoren fr Ihren Erfolg www.wenglor.com
T e c h n is c h e n d e ru n g e n v o rb e h
a lte n
206
56
Appendix A. Datasheets
Appendix B
Test Results
Table B.1: Seeding test results for Impatiens, Slider S
Position
1
2
0-0.5cm
3
2
03
04
05
06
07
08
09
010
011
012
013
014
015
016
017
018
019
020
021
022
023
024
025
026
027
028
029
030
031
032
033
0
0.5-1cm
1
0
4
2
2
3
4
2
1
2
3
0
1
3
2
2
1
2
4
0
3
1
2
2
0
4
2
0
0
0
1
2
1
1-1.5cm
0
1
0
2
0
2
1
0
1
0
3
1
2
2
1
0
2
1
0
2
0
2
2
0
0
1
1
5
2
5
7
0
0
1.5-2cm 2-2.5cm Outside

0
2
0
0
2
2
0
0
1
0
0
0
1
0
0
0
3
0
0
0
0
1
4
0
1
0
1
1
0
0
2
1
1
0
0
0
0
0
0
0
1
0
0
1
1
0
0
0
1
0
0
0
2
0
1
0
0
0
0
0
2
0
0
0
0
0
Continued on the next page
57
0
0
0
1
0
0
1
1
0
2
0
0
1
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
58
Appendix B. Test Results
Position
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
0-0.5cm
1
0
0
2
3
1
2
3
1
1
2
1
3
2
1
4
0
3
3
1
0
2
2
1
2
2
4
3
0
1
2
1
0
0.5-1cm
0
5
0
2
1
1
0
0
2
1
3
3
2
2
0
0
0
0
2
0
3
1
2
1
2
2
1
0
0
0
1
0
1
1-1.5cm
2
0
0
1
0
0
0
0
0
1
0
0
0
1
0
0
2
0
0
0
2
2
0
1
0
1
0
0
1
0
0
0
0
1.5-2cm
0
1
1
0
0
1
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
1
0
1
0
1
1
1
0
1
0
0
0
2-2.5cm
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Total
115
84
39
16
Outside
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
59
Table B.2: Seeding test results for Violas, Slider S

Position
1
2
0-0.5cm
2
1
03
14
05
06
07
08
09
010
011
112
013
114
015
016
017
118
019
020
021
022
023
024
025
026
027
028
029
030
031
032
033
034
035
036
037
038
039
040
041
042
043
044
045
046
0
0.5-1cm
0
0
1
2
2
0
1
1
0
1
1
0
1
1
0
1
0
0
0
1
1
0
1
0
0
0
1
0
0
2
0
2
0
1
0
1
0
1
1
0
1
1
0
0
1
1
1-1.5cm
0
0
0
0
0
1
0
0
1
0
0
1
0
1
2
0
0
0
0
0
0
1
0
2
0
1
0
1
1
0
1
1
0
0
1
1
0
2
1
1
1
0
0
1
1
0

0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
0
1
1
0
0
0
0
1
0
1
0
0
2
0
1
0
0
0
1
0
0
0
0
0
0
1
0
0
1
0
1
0
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
60
Position
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
0-0.5cm
0
1
0
0
0
0
1
1
1
1
1
0
0
0
0
1
1
0
0.5-1cm
1
0
0
1
1
1
0
0
0
0
1
1
0
1
0
0
0
0
1-1.5cm
0
0
1
1
0
1
0
0
0
0
0
1
0
0
1
1
0
2
1.5-2cm
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
2-2.5cm
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
Total
39
30
19
Outside
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
5
61
Table B.3: Seeding test results for Violas, Slider M

Position
1
2
0-0.5cm
6
3
03
04
05
06
07
08
09
210
011
012
013
114
015
016
017
018
019
020
021
022
023
124
125
026
027
028
029
030
031
132
033
034
035
036
137
038
039
040
041
042
043
044
045
046
0
0.5-1cm
0
1
0
1
0
3
3
1
1
3
1
1
0
0
1
0
0
0
3
3
1
1
0
2
2
0
3
0
0
1
2
2
2
2
3
2
2
2
0
1
2
1
2
3
1
5
1-1.5cm
0
0
3
2
0
2
0
1
1
1
3
1
1
2
0
1
0
3
0
0
3
1
1
1
0
1
0
0
2
3
0
0
3
3
1
0
2
1
1
1
1
0
1
1
0
0

0
0
1
0
2
1
2
0
0
0
1
3
1
1
2
1
0
2
0
0
1
0
1
1
1
0
1
0
2
1
2
0
3
0
1
0
0
0
0
0
2
0
1
0
0
2
0
1
1
1
1
0
0
0
2
1
1
0
1
0
0
1
1
0
1
0
0
0
0
0
1
0
1
0
2
0
2
1
3
0
1
0
2
0
0
0
0
0
2
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
2
0
0
0
0
0
2
0
1
0
0
0
2
0
0
1
0
0
0
0
0
0
1
0
0
0
0
62
Position
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
0-0.5cm
0
2
1
2
3
1
2
3
1
2
1
2
4
0
2
1
4
2
0.5-1cm
0
1
2
2
2
1
2
0
1
2
0
4
1
2
2
3
1
0
1-1.5cm
3
0
1
0
0
1
0
1
1
0
2
1
0
2
2
0
1
0
1.5-2cm
0
0
1
1
0
1
2
2
0
1
1
0
0
0
0
1
0
0
2-2.5cm
1
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
Total
105
75
60
28
15
Outside
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
8
63
Table B.4: Seeding test results for Phlox, Slider M

Position
1
2
0-0.5cm
1
1
03
04
05
06
07
08
09
010
011
012
013
014
015
016
017
018
019
020
021
022
023
024
025
026
027
028
029
030
031
032
033
134
035
036
037
038
039
040
041
042
043
044
045
046
0
0.5-1cm
0
1
1
1
3
2
1
2
0
0
1
0
1
1
1
0
0
3
2
2
1
1
0
1
0
0
1
1
1
0
0
0
1
1
1
1
1
0
0
0
2
2
0
0
1
1
1-1.5cm
0
0
0
1
0
0
0
0
0
0
0
2
0
1
1
1
0
0
1
0
0
0
0
0
1
0
1
0
0
0
2
2
0
0
0
2
1
0
1
1
0
0
0
2
0
0

0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
1
0
0
2
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
1
0
0
0
0
0
0
0
1
1
0
0
0
1
1
0
1
0
1
0
0
0
1
0
0
0
0
0
0
0
0
0
1
0
0
0
0
1
2
0
0
0
0
0
0
0
0
0
0
1
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
64
Position
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
0-0.5cm
0
1
0
0
1
1
0
0
0
0
0
1
3
0
0
1
1
1
0.5-1cm
0
0
2
1
0
0
0
1
1
1
0
0
0
1
0
2
0
0
1-1.5cm
1
0
0
0
0
0
0
0
0
0
2
0
0
0
1
0
1
0
1.5-2cm
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2-2.5cm
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
1
0
Total
50
30
19
Outside
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
65
Table B.5: Seeding test results for Phlox, Slider L

Position
1
2
0-0.5cm
2
1
03
04
05
06
07
08
09
010
011
112
013
014
015
016
017
018
019
020
121
022
023
224
025
026
027
028
029
030
131
032
033
034
035
036
037
038
039
040
041
042
043
044
045
046
0
0.5-1cm
2
2
2
4
2
1
3
4
2
0
1
0
4
1
4
1
1
0
5
1
4
3
0
3
2
1
1
1
1
0
1
3
3
1
1
2
1
1
3
1
4
2
0
0
3
0
1-1.5cm
0
0
0
0
0
0
0
1
0
1
2
0
1
1
1
0
2
3
2
1
0
0
2
0
0
0
0
0
2
1
2
1
0
2
1
2
2
2
1
4
0
2
1
0
0
2

0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
2
1
0
1
1
0
1
0
1
0
1
0
0
0
2
0
0
0
0
0
0
0
0
0
0
1
1
1
0
0
1
0
0
0
1
1
0
1
0
0
0
1
2
0
0
1
0
0
0
1
1
0
0
0
0
1
1
1
0
0
2
1
0
0
2
0
0
0
3
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
1
0
1
0
1
0
0
1
1
0
1
0
0
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
66
Position
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
0-0.5cm
0
1
1
0
1
1
1
2
1
0
1
2
2
2
0
4
1
2
0.5-1cm
2
1
3
0
2
2
1
1
0
0
2
0
2
0
1
4
2
2
1-1.5cm
1
4
0
0
3
1
0
0
2
3
0
2
2
3
0
0
0
0
1.5-2cm
0
0
0
1
0
0
0
1
1
1
0
0
0
0
0
1
0
0
2-2.5cm
0
0
0
0
0
2
0
0
0
1
0
0
0
0
0
0
1
1
Total
103
71
45
20
14
Outside
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
6
Appendix C
Content of the DVD

A DVD containing all project data is attached to this report. Its contents are
structured as listed below:
./admin/: the project description and agreements

./cad/: all CAD data
./datasheets/: Datasheets of all components that were bought externally
./documentation/: This Report and the corresponding LTEXsource files A
./electronics/: Altium Designer files of all custom-made electronic parts
./orders/: Drawings, 3D-Data and forms used to order components
./presentation/: Powerpoint files of the intermediate and final presentation
./references/: All cited papers and patents
./software/: Source code and binaries of all software written in this project
./tests/: Test measurements data
./video/: Some videos documenting the system
67
68
Appendix C. Content of the DVD
Bibliography
[1] L. Comba, P. Picarolo, D. Aimonino:Robotics and Automation for
Crop Management: Trends and Perspective. International Conference Ragusa
SHWA2010, Ragusa, Italy, 2010
[2] M. Fiala: Artag revision 1, a fiducial marker system using digital techniques.
National Research Council Canada Publication 47419/ERB-1117, 2004
[3] W. Balachandran, S. A. Thompson, S. E. Law, W. Machowski: Electrical Characteristics of an Electrostatic Valve Used for Bulk Transport of Agri cultural Seeds. IEEE Transactions on Industry Applications, Vol. 35, No. 2,
1999
[4] J. Hollinghum: Robots in agriculture. Industrial Robot: An International
Journal, Volume 26, Number 6, pp. 438-445, 1999
[5] A. Howard, S. Siddiqi, G. S. Sukhatme: An Experimental Study of Localization Using Wireless Ethernet. Field and Service Robotics, STAR24, pp.
145-153, Springer Verlag, Berlin, 2006
[6] D. O. Keller, J. Lightner, T. M. Swartwood, R. Wagner, J. V. Win kle: Robotic Seed Handling Apparatus and Methods. US Patent 6 705 827, 2004
[7] J.-P. Maury: Seeding Machine with a Pneumatic Seed Ejecting Means, Especially for Vegetable Culture. US Patent 4 399 757, 1983
[8] R. Siegwart, I. Nourbakash, D. Scaramuzza: Introduction to Autonomous Mobile Robots. ISBN 0-262-01535-8, The MIT Press, Cambridge,
Massachusetts, 2011
[9] E. L. Tonus: Needle Seeder. US Patent 4 480 765, 1984
[10] S. Trebbin: Verfahren und Vorrichtung zum Aufbringen von Motiven auf
Landschaften. German Patent Application DE 10 2008 031 315 A1, 2010
[11] S. Trebbin, M. Klimke: Verfahren und Vorrichtung zum Aufbringen von
Bildern und Texten auf Landschaften. German Patent Application (withdrawn)
DE 10 2008 031 314 A1, 2010
[12] Verein Schweizer Maschinen-Industrieller (VSM): Normenauszug f ur
technische Ausbildung und Praxis. ISBN 3-905430-03-7, 12th Edition 2002
[13] P. Wirz:Verfahren zur Herstellung von Saatmatten. German Patent
DE 10 155 703 C1, 2003
[14] Y. Yaji, N. Ito, K. Motobayashi, K. Nishiwaki, S. Mujinazawa,
K. Kudo, H. Tanaka: Seeding Machine with Controlled Planting at Predetermined Depth. US Patent 5 787 825, 1998
69
70
[15] R. Kaestner, A. Krebs, M. Hoepflinger: Driver library for the Maxon

EPOS motor control. https://aslforge.ethz.ch/projects/libepos/ (Retrieved
June 27, 2011)
[16] ROS: Robot Operating System. http://www.ros.org (Retrieved July 13, 2011)
[17] Fadenshop. http://www.fadenshop.ch (Retrieved March 9, 2011)
[18] Skycam Camera System. http://www.skycam.tv (Retrieved July 26, 2011)
[19] Robotshop. http://www.robotshop.com (Retrieved July 26, 2011)
[20] Active-Robots. http://www.active-robots.com (Retrieved July 26, 2011)
[21] SuperDroid Robots. http://www.superdroidrobots.com (Retrieved July 26,
2011)
Bibliography

Gardening Robot

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Gardening Robot

Загружено:

Авторское право:

Доступные форматы

Autonomous Systems Lab

Prof. Roland Siegwart

Spring Term 2011

1.2 Related Work . . . . . . . . . . . .

Creation of Flower Images .

6 Conclusion and Outlook

C Content of the DVD

Ejection sensor performance . . . . . . . . . . . . . . . . . . . . . . .

Localization methods summary . . . . . . . . . . . . . . . . . . . . . Required

Project Goals and Requirements

Dahlia hybrids: Flat, longish, about 1x4mm

1.2. Related Work

As agriculture is extensively supported by technical means like seeding, mowing or

Seed Delivery Systems

Creation of Flower Images

Seed Delivery System

Chapter 2. Seed Delivery System

2.1. Basic Principles

Chapter 2. Seed Delivery System

As a basic concept, a combination of the following principles was chosen:

Select a seed using gravity through a controllable downward opening in the

Deliver the seed to the ejection port by applying air pressure

Seed Selection Mechanism

Air Pressure Tube

Air Pressure Tube

Figure 2.3: Slider with limited opening volume

Chapter 2. Seed Delivery System

Upper Ground Plate

Figure 2.5: Seed delivery module for multiple seed types

Chapter 2. Seed Delivery System

Figure 2.6: Relay driver

From Delivery Tube

Ejection Sensor Light Barriers

Chapter 2. Seed Delivery System

Figure 2.8: Ejection sensor schematic

Extendability and Possible Improvements

Sensors on the Sliders

XY-Table for Tiling

Chapter 3. XY-Table for Tiling

- Needs a frame to hold

- Limited glue types (no

Table 3.1: Substrates

- Limited glue types (no

- Needs more time to

name: Spr uhfixierer Gold Zack 250 ml

Chapter 3. XY-Table for Tiling

Limit Switches (Y)

Figure 3.1: XY-Table (Overview)

Figure 3.2: XY-Table system schematics

Chapter 3. XY-Table for Tiling

Motors and Controller

Graphical User Interface

Figure 3.3: Graphical user interface

As written above the Arduino microcontroller can be easily programmed in the

Chapter 3. XY-Table for Tiling

Position and Seed

Figure 3.4: Planting process owchart

Chapter 3. XY-Table for Tiling

Number of Ejected Seeds

Figure 4.1: Test setup

Ejection Sensor Performance

True positive (tp): Ejection & Detection

4.4. Ejection Sensor Performance 27

Figure 4.3: Position of ejected Viola seeds