Field and Service Robotics - Corke P. and Sukkarieh S.(Eds) Part 15 potx

condi-As part of the project, the following main components were designed andtested for their feasibility and fulfillment of the requirements:• Carrier system motion kinematics, robot for

condi-As part of the project, the following main components were designed andtested for their feasibility and fulfillment of the requirements:

• Carrier system (motion kinematics, robot) for positioning along the sewerline;

• Sensor and measuring systems for inspecting pipe condition above andbelow the water line as well as for detecting deposits;

• Sewer cleaning system;

• Media supply (power, data communication, water, etc.);

• Control system, operation;

• System navigation and positioning in the sewer;

• Handling systems for positioning sensors and cleaning tools on and alongthe sewer wall

A large test station with various reinforced concrete pipes and different types

of damage (e.g cracks or spalling) was set up at the Fraunhofer Institute IFF

in Magdeburg The sensors for inspection were mostly new developments

In consultation with the Emschergenossenschaft, the Fraunhofer IFF hasdeveloped and built a prototype of the rough inspection system as well as atest prototype of the cleaning system and a test prototype of the inspectionsystem in order to test these in a comparable, already existing sewer system(diameter 2300 mm)

2 Inspection Strategy and Systems

The strategy for automatically inspecting and cleaning the Emscher sewersystem incorporates a three-stage approach

In the first stage, a small swimming system called the Spy is employed in thesewer for rough inspection It inspects and measures the entire sewer line andconducts camera inspections, recording major abnormalities such as erosion,deposits, obstacles and leaks in the gas space At the same time, it checkswhether the cleaning and inspection systems detailed below can be deployed.The Spy must be able to position itself centrically even in curved pipes in thesewer covering a length of 600 m

Automated Inspection System for Large Underground Concrete Pipes 569

In the second stage, if necessary, the wheel-driven cleaning system eliminatesdeposits detected by the spy in the bed area and cleans the sewer wall beforethe inspection system is deployed

In the third stage, the inspection system (wheel-driven for diameters

< 2000 mm, swimming for diameters > 2000 mm) inspects the sewer pletely, measuring the sewer (joint widths, pipe offsets, cracks) with greateraccuracy than the Spy For all three systems, concepts were designed and de-

com-Fig 1 Spy prototype and test of the swimming inspection system in a real sewer

veloped for their control, operation, introduction into and extraction from thesewer line as well as for their energy and water supply and data transfer.2.1 Rough Inspection System (Spy)

The Spy (Fig 2) is an easy-to-operate, cable-guided swimming system forrough screening of the sewer line The Spy detects corrosion, obstacles, de-posits and cracks The Spy detects sewer conditions with little operating effortbut with less precision than the inspection system

Fig 2 Sensors on Spy

570 N Elkmann et al.

Using a camera system, the Spy can visually inspect the gas space It isequipped with several flashlights for illumination Flashlights are used becausethey yield more light while consuming less power than floodlights The Spy isadditionally equipped with ultrasound sensors for sewer measurement in thewater space The Spy prototype must have smooth swimming behavior andlie stably in the sewer line even at higher flow velocity The objective of thesuccessful navigation tests was to position the Spy centrally in the current inorder to create good conditions for geometry measurement It was possible todetermine the position of the Spy in the sewer and measure the sewer crosssection in the gas space

2.2 Inspection Systems

In contrast to the Spy, the distinctive feature of the inspection system is itsability to achieve greater accuracy of measurement with its measuring sensors.Sensor systems are additionally integrated Various concepts were developedfor the carrier system The two favored carrier systems are:

1 Floating systems for large sewer diameters and

2 Wheeled chassis for smaller sewer diameters

The floating systems are convincing because of the high certainty of recovery.Their operational range is limited by the required water level though Wheel-guided car systems are used when filling level is low or nominal diametersare smaller The test prototype is modularly constructed and represents the

Fig 3 Swimming inspection system

Automated Inspection System for Large Underground Concrete Pipes 571two favored carrier system concepts: Swimmer and Car The Car test proto-type consists of the Swimmer and the additional wheeled chassis subsystem.Sensors for determining position in the sewer (laser ranging sensors and incli-nation sensors) and sensors for damage detection (laser scanners, ultrasoundscanners, camera system, ultrasound crack sensor) were installed on the in-spection system These sensors are either rigidly connected directly with thecarrier system or they are moved by additional sensor kinematics The rota-tion arm on the stern of the carrier system moves the ultrasound crack sensoralong the sewer wall Ultrasound scanners, laser scanners and camera systemare mounted on a linear axis and can precisely measure the pipe profile over

a length of approximately 1.5 m

3 Positioning, Pipe Axis Measurement

Momentary position and orientation in the sewer have to be known at the time

of any measurements with the Spy and with the inspection system Therefore,pipe axis measurement is an essential prerequisite for exactly representing andanalyzing the sensor data To this end, an algorithm was developed, which,taking a model of a complete pipe as its starting point, measures the pipe axisexactly The total error of pipe axis measurement is arrived at by adding upthe accuracy of the laser ranging sensors, the tolerance and the pipe’s surfacecondition as well as the systematic error caused by the Spy system’s motion.For the laser and ultrasound scanners to detect damage, a positional value

of the pipe axis has to be assigned for every individual reading Accordingly,when the accuracy of measurement is being assessed, the superposition ofposition detection and the measuring method for detecting damage has to beassessed

The inspection system achieves a greater accuracy of measurement of itsposition in the pipe because it is stationary during measurement and, as such,only the accuracy of the laser ranging sensors themselves plays a role in theoverall measurement accuracy A sensor system was conceived, which uses

15 laser ranging sensors (5 aligned vertically and 10 horizontally) to stantly record position The sensor distance data is converted into the sensorcoordinate system

con-To measure the pipe axis, a cylinder with an elliptical surface area is used

to model the real pipe with its surface quality and tolerances This is clearlydescribed in the Spy’s coordinate system by the cylinder axis, the radius andthe diameter The interpretation of the measuring data would be easy if theSpy were exactly in the center of the pipe without any deviation in its angles

of alignment In reality, the systems tilt at yaw and pitch angles are out ofline with the center of the pipe The measuring points are not on a straightline but rather on a segment as the green measuring points in Fig 4 indi-cate Hence an exact model of the measurement has to be made, which allowsfor the curvature of the pipe The model is based on the correlation between

572 N Elkmann et al.

Fig 4 Coordinate system, difference between the readings in the pipe model (greenmeasuring points) and the model with straight walls (red measuring points)

the measured distance, pipe radius and displacement to the pipe axis as well

as the yaw and pitch angles Since this non-linear dependence is known, thealignment of the pipe axis can be determined from the distance measurement.This alignment then makes it possible to transform the measuring points ontothe circular projection of the pipe and consequently to determine the posi-tion of the pipe axis in relation to the Spy and the inspection system Thepipe position is determined first by mathematically resolving the non-linearcorrelations The determination of the position of the pipe axis is based onusing the pipe axis alignment to plot the measuring points on the circularprojection After applying this transformation, a circle with an offset center is

fit to the measuring points The displacement of the axis of the pipe vis-`a-visthe axis of the Spy is obtained from this fit

In principle, this approach opens a method for measuring the pipe axis, which

is independent of the pipe diameter as well as of the orientation and position

of the measuring system The method was modified to the effect that the sor alignment is compensated for by parable approximations Correspondingcalibrating measurements are taken on a calibration rig

sen-The system’s position along the axis of the sewer line is determined by suring the lengths of cable uncoiled In addition, the camera system referencesthe current position at all joints with the camera system This way, inaccura-cies caused by cable sag and slippage can be compensated for and every singlepipe can be approached with an accuracy of ±50 mm

mea-4 Types of Damage and Selected Sensor Systems for Damage Detection

One focus of the project was the development of the sensor systems, whichhave the required accuracy of measurement under difficult conditions in thesewer and make it possible to take comparative measurements throughout thesewer’s period of operation (120 years)

Automated Inspection System for Large Underground Concrete Pipes 573Minimum requirements for sewer inspection in Germany are stipulated inself-monitoring regulations issued by the states Legal requirements, techni-cal specifications and negotiations with local authorities have produced theinspection tasks displayed in Table 1 The requirements of a one-pipe line arefar more demanding than those in the technical guidelines.

(*) Cracks caused by mechanical stress can be located throughout the entirepipe Cracks detected in the upper section of the pipe can be used to calculatethe extent of cracks in the lower section

Table 1 Inspection tasks for the interceptors parallel to the Emscher sewer system

4.1 Chemical Corrosion

Optical measuring methods detect surface corrosion of the concrete in thegas space and represent possible developments of damage A semiautomaticprocedure consisting of automatic and manual analysis by an operator is fa-vored for corrosion detection and classification The option of mapping theconcrete wall comparatively with previous inspections is important in order

to be able to map any possible development of damage Several cameras areused to image the sewer wall completely

An image processing algorithm with a short runtime is used to detect normalities immediately The appearance of individual structural elements ofthe surface is inspected for abnormalities, the measured number being moreimportant than its precise characteristic When a variable limit value is ex-ceeded, surface corrosion may be likely

ab-574 N Elkmann et al.

Direct statements can be made about the possible occurrence of corrosion bycomparing the distribution of the various proportions of gray tones in thereadings with calibrated values or values already ascertained from previousinspections Fig 5 presents characteristic gray tone distribution curves for dif-ferent surfaces Fig 5 clearly shows the various curves for differently corroded

Fig 5 Image of a corroded concrete surface with superimposed gray scale curvefor subareas

surfaces Clearly, differently damaged subareas can be identified individually.The total assessment of potential corrosion would be obtained by averagingthe entire image space

In addition, laser scanners, which measure the cross section of the pipe,are used to detect corrosion with an accuracy of ±4 mm

4.2 Obstacles, Sediments, Incrustations, Mechanical CorrosionNewly developed ultrasound scanners with an accuracy of measurement of

±2 mm are being used to detect obstacles, deposits and mechanical erosion inthe water space

Fig 6 shows the test setup for geometry measurement in the water space withultrasound scanners and a scan image

4.3 Crack Detection in Concrete Pipe

First, digital image processing systems are used to detect cracks in the gasspace Several cameras are used to identify cracks in the gas space

Automated Inspection System for Large Underground Concrete Pipes 575

Fig 6 Geometry measurements with ultrasound scanner (obstacles, sediments)

In accordance with the requirements, cracks with a width of 0.5 mm and ward have been positively identified and logged While cracks can reach along length, their frequently very narrow width makes great demands on themeasuring system mapping them Other measures such as comparisons withprevious inspections and images of other cracks with known width as well

up-as the superimposition of scales help make it possible to more closely termine crack width and thus more closely detect the type of damage Animportant analysis module is automatic crack detection It employs methods

de-of image processing and pattern recognition in order to determine whether one

or more cracks are possibly visible on a particular image or not Particularlywhen there are small cracks, which an operator could overlook on the moni-tor, this automatic system constitutes a considerable advantage and increasesthe quality of the inspection results Fig 7 illustrates how different analysismodules identified a crack In addition, each crack was graphically marked as

a recognized structure for the purpose of presentation The entire crack figuration was never identified However, only the information of whether aFig 7 Details of result images when different crack detection methods are employed

con-576 N Elkmann et al.

crack may be present in a particular image or not is important for supportingthe user It follows from this that the automatic analysis module can alreadyterminate the processing of the current image and inform the user once anycrack segment has been found

Additionally, new acoustic methods (ultrasound, impact-echo) have beendeveloped or adopted to detect cracks in the concrete in the gas and the waterspace These acoustic systems are able to provide information on crack depth.The acoustic methods for crack detection additionally allow the following:

• With the right sensor system, cracks can be detected in the water spacetoo

• Cracks can be roughly classified (crack depth)

• Spallings can be detected and wall thickness can be determined

Fig 8 Acoustic sensor systems for crack detection

The use of these sensors sensor systems for crack inspection is completely new.4.4 Deviation of Pipe Position

Horizontal and vertical deviations of position and joint gaps have to be sured Laser scanners, aligned laterally or on the apex of the sewer, are used

mea-to detect and record the horizontal and vertical deviations of position

In the gas space, cameras measure the joint gap Differences in joint widthcompared with earlier inspections indicate an axial displacement Inconstantjoint width along the pipe circumference indicates a deformation

Automatic measurement requires exact identification of the joint edges Tothis end, image processing methods (segmentation, contour-finding) determinethe pixels on the edges of the joint

Fig 9 (a) shows a detail of the identified pixels When the parameters havebeen suitably selected, the joint edges can be identified with an accuracy of

a few pixels If these pixels are used to apply ellipse approximations, whichoptimally approximate the number of points, the joint edges are obtained,which support automatic measurement of the joints

Automated Inspection System for Large Underground Concrete Pipes 577

Fig 9 (b) shows a detail of the joint image with such ellipse approximations

A Hough transformation can be used to determine the ellipse mations Since positioning and joint identification already identify the jointedges, the parameters of the corresponding ellipses are also approximatelyknown Thus, the search area of the Hough transformation can be restrictedgreatly, making efficient implementation possible

approxi-Fig 9 Image detail with detected joint edges: When measurement is manual thejoint width can be marked by hand (a) When joint measurement is automated,ellipse approximations are placed through the joint edges (b)

5 Summary and Outlook

Since 2002, the Fraunhofer IFF as general contractor has developed a prehensive concept for inspection and cleaning systems for the Emscher sewersystem Not only have all the relevant subsystems been identified but theyhave also been designed in detail and subjected to all necessary tests in order

com-to be able com-to provide reliable information about their feasibility Feasibilitywas fully demonstrated Foci of development were the carrier systems formovement along the sewer line guaranteeing maximum recovery certainty, thepipe axis measurement and position sensing of the systems in the sewer andthe sensor systems for detecting the condition of the sewer’s gas and waterspaces Different sensor systems have been developed and tested in the teststation as well as in a real sewer Erosion, incrustations and corrosion of con-crete are detected with great accuracy Cameras detect axial displacement andlaser scanners detect offsets in pipe joints in the gas space Apart from thecameras, different sensors for crack detection in the gas and water space weredeveloped on an acoustic basis (e.g ultrasound)

578 N Elkmann et al.

Along with the sensors, all systems were designed for the favored inspectionand cleaning concept This involved a system for rough inspection of thesewer (Spy) as well as cleaning systems and inspection systems The control,the operation, the introduction into and extraction from the sewer and themanhole as well as the energy and water supply were engineered and thecertainty of recovery in case of breakdown was guaranteed

In consultation with the Emschergenossenschaft, the Fraunhofer IFF veloped and built a prototype of the spy and test prototypes of the cleaningsystem and the inspection system in order to acquire more experience un-der real conditions in the sewer The swimming behavior of the Spy and thefloating test prototype for the inspection system were studied The sensorbehavior for crack detection and sewer cross section measurement was alsotested The collected findings and insights will now enter into the engineeringand development for final prototypes

de-The feasibility of automatic inspection and cleaning systems for the scher sewer system and the fulfillment of the legal requirements for inspectionand cleaning have been demonstrated The research on and tests of the in-spection systems, the sensor systems and the cleaning technology guaranteethe inspection and cleaning required by law in a one-pipe sewer

Em-References

1 Hertzberg, Christaller, Kirchner, Licht, Rome: ”Sewer Robotics”, In: Proc.From Animals to Animats 5, 5th Intl Conf On Simulation of Adaptive Behav-ior (SAB-98), R Pfeifer and B Blumberg and J.-A Meyer and S.W Wilson(eds), MIT Press, P 427-436, 1998

2 Kuntze H.-B., Haffner H.: Experiences with the Development of a Robot forSmart Multisensoric Pipe Inspection ICRA 1998: 1773-1778

3 Rome E., Hertzberg J.,Kirchner F., Licht U., Streich S., Christaller Th.: wards Autonomous Sewer Robots: the MAKRO Project Urban Water 1, 1999,

To-P 57-40

4 Kirkham R:, Kearney, P Rogers K and Mashford J.: PIRAT - A System forQuantitative Sewer Pipe Assessment International Journal of Robotics Re-search, Vol 19, No 11, November 2000

5 Elkmann N., Althoff H., Saenz J., B¨ohme T.: Kinematics Systems for Inspectionand Cleaning of Sewer Canals 6th International Conference on Climbing andWalking Robots CLAWAR, Catania, 2003

6 Elkmann N., Althoff H., B¨ohme T., Felsch T., Kutzner S., Saenz J., St¨urzeT.: Entwicklung von Robotersystemen f¨ur die Inspektion und Reinigung vonAbwasserkan¨alen, Robotik 2004, M¨unchen, 17–18 June 2004

7 Elkmann N., Althoff H.: The emscher:kanal - Development of an AutomatedInspection System for Underground Concrete Pipes, 22th International NODIG Conference, 15.–17 November 2004, Hamburg, Germany

An Autonomous Weeding Robot

for Organic Farming

Tijmen Bakker1, Kees van Asselt1, Jan Bontsema2, Joachim M¨uller3andGerrit van Straten1

1 Wageningen University, Systems and Control Group, P.O Box 17, 6700 AAWageningen, The Netherlands, tijmen.bakker@wur.nl

2 Agrotechnology and Food Innovations BV, P.O Box 17, 6700 AA Wageningen,The Netherlands

3 University of Hohenheim, Institute for Agricultural Engineering, 70593

Stuttgart, Germany

Summary The objective of this research is the replacement of hand weeding inorganic farming by a device working autonomously at field level The autonomousweeding robot was designed using a structured design approach, giving a goodoverview of the total design A vehicle was developed with a diesel engine, hydraulictransmission, four-wheel drive and four-wheel steering The available power and thestability of the vehicle does not limit the freedom of research regarding solutions forintra-row weed detection and weeding actuators To fulfill the function of navigationalong the row a new machine vision algorithm was developed A test in sugar beet

in a greenhouse showed that the algorithm was able to find the crop row with anaverage error of less than 25 mm The vehicle is a versatile design for an autonomousweeding robot in a research context The result of the design has good potential forautonomous weeding in the near future

Keywords: Systematic design, machine vision, GPS, robotics, intra-rowweed control, autonomous weeding robot, organic farming

1 Introduction

Weeds in agricultural production are mainly controlled by herbicides As inorganic farming no herbicides can be used, weed control is a major problem.While there is sufficient equipment available to control the weeds in betweenthe rows, weed control in the rows (intra-row weeding) still requires a lot ofmanual labour This is especially the case for crops that are slowly growingand shallowly sown like sugar beet, carrots and onions In 1998, on average

73 hours per hectare sugar beet were spent on hand weeding in the lands [4] The required labour for hand weeding is expensive and often not

Nether-P Corke and S Sukkarieh (Eds.): Field and Service Robotics, STAR 25, pp 579–590, 2006.

© Springer-Verlag Berlin Heidelberg 2006

580 T Bakker et al.

available An autonomous weeding robot replacing this labour, could mean

an enormous stimulus for organic farming This paper presents the design ofsuch an autonomous weeding robot currently being developed at WageningenUniversity

2 The Design Procedure

2.1 Method

The autonomous weeding robot is designed using a phase model as the designmethod[3] In this phase model the design of a product is represented as aprocess consisting of a problem definition phase, alternatives definition phaseand a forming phase (figure 1) The results of the different phases are solutions

on different levels of abstraction

The problem definition phase starts with defining the objective of the design

In the problem definition phase also the set of requirements is established.The requirements can be split into fixed and variable requirements A designthat does not satisfy the fixed requirements is rejected Variable requirementshave to be fulfilled to a certain extent To what extent these requirements arefulfilled, determines the quality of the design The variable requirements arealso used as criteria for the evaluation of possible concept solutions The lastpart of the problem definition phase consists of the definition of the functions

of the robot A function is an action that has to be performed by the robot toreach a specific goal In our case, important functions are ’intra-row weeding’and ’navigate along the row’ The functions are grouped in a function struc-ture, which represents a solution on the first level of abstraction

The function structure consists of several functions Every function can beaccomplished by several alternative principles, e.g mechanical and thermalprinciples for weed removal In the alternatives definition phase, possible al-ternative principles for the various functions are presented in a morphologicalchart (fig 3) The left column lists the functions and the rows display the

Fig 1 The design process

An Autonomous Weeding Robot for Organic Farming 581alternative principles By selecting one alternative for each function and bycombining these alternatives, concept solutions can be established These con-cept solutions are represented by lines drawn in the morphological chart Thebest concept solution can be selected using a rating procedure In the formingphase this selected concept solution is worked out into a prototype.

2.2 Application for the Weeding Robot

The objective of the research is formulated as ’replacement of hand weeding inorganic farming by a device working autonomously at field level’ Starting fromthis objective, the first step in the problem definition phase was to establishthe set of requirements For this purpose interviews were held with potentialusers, scientists and consultants related to organic farming The resultingrequirements are as follows:

Fixed requirements:

• Replacing hand weeding in organic farming

• Applicable in combination with other weed control measures

• Manual control of the vehicle must be possible for moving the vehicle overshort distances

• Weeding a field autonomously

• Ability to work both day and night

• The weeding robot should not cross the field boundaries

• The weeding robot must be self restarting in absence of emergency

Fig 2 The function structure

582 T Bakker et al.

• The weeding robot informs the farmer when the weeding robot stoppeddefinitely (e.g due to security reasons) or when it is ready

• The weeding robot sends its operational status to the user at request

• The weeding robot must function properly in sugar beet

Variable requirements:

• Removing more than 90 percent of the weeds in the row

• The costs per hectare may be at least comparable to the costs of handweeding

• Damage to the crop is as low as possible

• The soil pressure under the weeding robot must be comparable or less thanfor hand weeding

• Energy efficient

• Safe for people, animals and property

• Suitable as research platform

• Limited noise production

• Reliable functioning

• Easy to use

After establishing the set of requirements the functions of the the weedingrobot were identified These functions were grouped into a function blockscheme This scheme is represented in figure 2 The lines in the scheme in-dicate flows of energy, material or information Functions located in parallellines can be performed simultaneously

The navigation system consists of four functions Firstly, the weeding robotshould constantly determine if it is located in- or outside the field Secondly,

if within the field, it should determine if it is on one of the headlands or not.Thirdly, in case it is not on the headlands, it should navigate along the rowand perform the intra-row weeding Fourthly, if the weeding robot arrives onthe headland, it should stop the intra-row weeding and start to navigate tothe next crop rows to be weeded This sequence repeats until the whole field,except the headlands, is weeded Weeding of the headlands is left out of con-sideration An increasing number of farmers in the Netherlands do not growsugar beet at the headlands because they think it is not cost-effective

In the alternatives definition phase possible alternative principles for the ous functions are listed in a morphological chart (fig 3) Four people involved

vari-in the project drew lvari-ines vari-indicatvari-ing possible concept solutions vari-in the chart.These concept solutions were then weighed against each other using the vari-able requirements listed before The concept solution indicated by the line infigure 3 is the final concept solution

In the forming phase described in section 3 the concept solution was workedout into a prototype

Fig 3 Morphologic chart

An Autonomous Weeding Robot for Organic Farming 583

584 T Bakker et al.

2.3 Results of the Design Process

Determine where intra-row weeding has to be performed

To determine where intra-row weeding has to be performed, pattern nition of plant locations is going to be used From earlier research [2] it isexpected that the quality of detection of this method is at least as good asthe quality of detection of other methods Though combinations of methodslike recognition of pattern, shape and colour are expected to have a potentialfor higher quality of detection, just pattern recognition is chosen because it isexpected to be sufficient

recog-Positioning of weeding

To position the actuator at the location indicated by the detection systemdead reckoning is going to be used A wheel with encoder, giving a precisedistance measurement, will be available already because it is also needed forthe pattern recognition system

Intra-row weeding

Intra-row weeding will be performed by a mechanical actuator It is expected

to be difficult to remove weeds growing close to a crop plant by air, flaming,electricity, hot water, freezing, microwaves or infrared without damaging thecrop plants In that respect laser would be an excellent solution However,laser can not work under the ground surface, and has therefore less effect oncertain weed species On the other hand, not moving the soil prevents buriedseeds from germinating A greater disadvantage of laser is its high price Highpower laser is needed to reach reasonable performance, and this involves highcosts Water-jet could also probably be a good solution for intra-row weeding,but this needs much more investigation than a mechanical solution

Determine if within field

GPS is selected to determine wether the weeding robot is within the field ornot The determination if the weeding robot is located within the field or not,needs to be guaranteed correctly at any time A combination of vision anddead reckoning can not give this guarantee as good as a solution in whichGPS is used Dead reckoning could improve the position determination byGPS However, if a GPS is selected with sufficient accuracy, additional deadreckoning is not needed

Navigate along the row

Machine vision is selected for navigation along the row Machine vision makes

it possible to navigate along the row by relative positioning to the row fore the weeding robot can work in any field without requiring absolute co-ordinates of a path to be followed Absolute positioning by means of GPS,

There-possibly combined with other sensors, requires knowledge of the absolute sition of crop rows in a field Navigation along the row by relative positioning

po-to the row could be done also using tactile, ultrasonic or optical sensors bined with dead reckoning Tactile sensors are not going to be used because

com-in case of sugar beet they could harm the crop Machcom-ine vision is preferredover ultrasonic or optical sensors, because of the ability to look forward, whichcontributes to a more accurate control of the position of the weeding robot rel-ative to the crop row It is not clear wether dead reckoning could substantiallycontribute to the navigation accuracy feasible with machine vision

Determine if on headland

GPS is selected to determine if the weeding robot is located on the headland.Using GPS requires some labour for recording the border of the headlands inadvance, but will result in a correct headland detection If a high accuracyGPS is selected, accuracy does not have to be improved by dead reckoning.Tactile, ultrasonic or optical sensors in combination with dead reckoning couldalso be used to determine wether the robot is on the headland, by detectingthe end of the row, i.e if over some predefined distance no row is detected.However, another crop may grow on the headland (seeded to prevent germi-nating of weeds) or crop rows seeded at the headland can cross the crop row

to be followed In these situations the latter solutions can not guarantee acorrect detection of the end of row, and therefore also not a correct head-land detection Machine vision could give more reliable results, but it is stilldifficult because headland to be detected is not so structured

Navigate on headland

For navigation on the headland GPS is selected On the headland the weedingrobot has to make a turn and position itself in front of the next rows to beweeded At the moment the robot arrives at the headland, a virtual path isplanned to a position in front of the next rows to be weeded Navigating overthis path is going to be done by GPS

Locomotion related functions

A diesel engine with a hydraulic transmission was selected for the tion related functions For weeding quite some power could be required andthe available power should not be limiting for realizing the objective of au-tonomous weeding of a field A diesel engine with an hydraulic transmission

locomo-is a proven concept in agriculture A gearbox limits the possible combinations

of the number of engine revolutions and driving speed and shuffling is difficult

to automate A continuously variable or hydraulic transmission is thereforepreferred over a gearbox Hydraulics makes it possible to design a compactwheel construction preventing damage to the crop

A design with four wheels is preferred over one with three because of stability

An Autonomous Weeding Robot for Organic Farming 585

586 T Bakker et al.

It was decided that four wheels is also preferred over two or four tracks Themost important advantages of tracks in practice are the better traction andthe less soil compaction But it is expected that if four wheels are used forsuch a light-weight vehicle (not more than 1500 kg) soil compaction will be ac-ceptable Traction when using wheels is expected to be good enough because

of the limited weight and the limited need of traction for intra-row weeding.Four wheel drive and four wheel steering were chosen to have the possibility

to investigate all kinds of driving strategies

Communication with the user

Specific settings for a field will be defined by a board computer Any moment

a user wants to know the status of the weeding robot, the weeding robotstatus will be accessible via the internet A website gives good opportunities

to represent information in an orderly way and it is easily accessible fromeverywhere In case the weeding robot needs help from its user, the weedingrobot notifies its user by sending an SMS (Short Message Service) message bythe GSM network In the Netherlands any place is covered by the GSM net-work From the alternatives listed, SMS is the solution that gives the highestassurance that the user really receives the message shortly after it is sent.Detect unsafe situations

Detecting unsafe situations will be done super canopy all around the weedingrobot Situations in which this solution is not sufficient are hardly imaginable.Ideally the weeding robot should detect every unsafe situation, at every leveland direction Even if somebody is lying in between the crop rows belowcanopy level this should be detected Because of the research effort involved

in reaching the ideal objective mentioned and the possible high costs for such

a solution, detecting around and only super canopy is preferred

3 The Vehicle

The size of the vehicle was determined by the standard track width used inagriculture in the Netherlands which is 1.50 m This track width also makesthe design versatile in the sense that it is suitable for crops grown in beds likecarrots an onions See figure 4 for the resulting vehicle

Sugar beets are grown at a row distance of 50 cm so the weeding robot ers three rows The engine power is selected so that it has enough power fordriving and steering under field conditions and for driving three actuators.The required power for the actuators was calculated based on an actuatorspecially designed for intra-row weeding by Bontsema et al [2] The engine is

cov-a 31.3 kW Kubotcov-a V1505-T

The ground clearance is about 50 cm to prevent the crop from being damaged

by the vehicle The vehicle is 2.5 m long to have enough space for mounting

Fig 4 The weeding robot

actuators under the vehicle in the middle between the front and rear wheels.The tyre width of 16 cm leaves enough space for steering in between crop rowswhile soil compaction is expected to be acceptable The weight of the vehicle

is about 1250 kg

The engine drives two hydraulic pumps One supplies the oil for steering anddriving, and the other for driving the actuators The oil for driving and steer-ing flows to a electrically controllable valve block with eight sections Fourare used for steering and four are used for controlling wheel speed, so wheelspeeds and wheel angles can be controlled individually The wheels are driven

by radial piston motors The required driving speed range for intra-row ing is 0.025 m/s - 2 m/s continually variable A desired top speed of 5.6 m/swas specified for fast moving of the robot within a field It appeared thathydraulics could not be designed to have a variable work speed from 0.025m/s to 5.6 m/s A solution was found by designing the hydraulics so thattwo speed ranges exist The working speed ranges up to 3.2 m/s A maximumtravel speed of 6.4 m/s is realized by changing to two wheel drive by combin-ing the oil flows of four wheels into two flows

weed-Each wheel is steered by an hydraulic motor with a reduction gear The imum steering speed is 360 degrees per second The angles of the wheels aremeasured by angle sensors The oil for driving the wheels flows via a turnableoil throughput This makes it possible to turn the wheels in any angle from0-360 degrees

max-The weeding robot electronics consists of 6 units connected by a CAN bus

An Autonomous Weeding Robot for Organic Farming 587