4th IFAC Conference on Modelling and Control in Agriculture,

Horticulture and Post Harvest Industry

August 27-30, 2013. Espoo, Finland

A Row Crop Following Behavior based on Primitive Fuzzy Behaviors for

Navigation System of Agricultural Robots
Rafael V. de Sousa*. Rubens A. Tabile*. Ricardo Y. Inamasu**. Arthur J. V. Porto***.

*Department of Biosystems Engineering, University of São Paulo, Brazil (rafael.sousa@usp.br;rubens.tabile@usp.br).
**Embrapa Instrumentation, Brazilian Agricultural Research Corporation - Embrapa, Brazil (ricardo.inamasu@embrapa.br)
***Department of Mechanical Engineering, University of São Paulo, Brazil (ajvporto@usp.br)

Abstract: The work presents a methodology for generating and implementing a complex row crop
following robotic for a mobile agricultural robot. Fuzzy rules are used to compose and coordinate the
simple behaviors using the specific agricultural contexts. Also, is proposed a method for coordinating the
simple behaviors based on multiples arbiters that operate in different stages on the robotic architecture to
perform the path tracking. A commercial robotic platform is adapted to simulate and support the behavior
based on architecture to allow the experiment performed in order to evaluate the implemented behaviors
and to evaluate the operational ability of the robotic platform on a simulated agricultural environment.
The modularity of the architecture using decentralized arbiters simplifies the implementation of the
robotic architecture. The results show the feasibility method to implement the behavior coordination and
to compose complex behaviors based on fuzzy and non-fuzzy simple context dependent behaviors.
Keywords: Mobile agricultural robot, behavior based architecture, autonomous navigation, fuzzy robotic
behavior, automatic guidance.

for mobile robots navigation (Huq et al. 2008). The fuzzy
1. INTRODUCTION
logic theory provides a formal methodology, which can be
Agricultural practices such as Precision Agriculture and the applied to transfer the human experiences to a robotic system.
practices that take environmental protection into account Linguistic variables are used to compose sets of simple and
need accurate and large sampling scale systems. One of the intuitive conditional statements (fuzzy rules). These heuristic
trends in the area is the application of semi-autonomous or rules associate perceptions to actions which generates
autonomous vehicles and robots on agricultural tasks behaviors to compose architectures in order to allow robot
(Blackmore et al. 2005; Pedersen et al. 2005; Blackmore tasks, such as robot navigation (Seraji et al. 2002).
2007; Bak et al. 2004). Feasible solutions are proposed for
A hierarchical architecture is proposed to compose complex
the development of AAV (Agricultural Autonomous Vehicle)
behavior based on primitive behaviors in Abreu et al. (2001).
and MAR (Mobile Agricultural Robot) of autonomous
A behavior-based architecture developed and implemented
agricultural vehicles and robots and positive advances on
on a Khepera basic module robot is discussed. Sub-modules
research are noticed in recent years (Barawind et al. 2007;
based on fuzzy rules compose the behaviors, and activation
Cariou et al 2009; Darr et al. 2005; Gan-Mor et al. 2007;
functions are defined for the behaviors. The physical location
Johnson et al. 2009; Kaizu et al. 2008; Kise et al. 2005;
of the sensors is used to create the low-level behaviors. Fuzzy
Nomark et al. 2012; Nagasaka et al. 2009). However, a
logic approaches are also proposed to integrate or coordinate
limited number of works have developed reliable systems
low-level behaviors in order to compose a complex behavior.
based on a robotic architecture which is able to perform
multiple independent operations and to self-adapt in order to An approach to develop and coordinate behaviors based on
change environmental conditions (Bakker et al. 2011; Cheein fuzzy rules is described in Saffiotti et al. (1999). Contexts of
et al 2011). behavior application are defined and fuzzy rules are applied
to blend those behaviors. Another approach in which a
On the other hand, in other areas of research, a considerable
method to design a fuzzy controller for navigation uses two
number of behavior-based architectures have been proposed
levels of abstraction is presented in Aycard et al. (1997). In
for mobile robot of autonomous guidance and navigation in
the first level, the sensors are grouped according to physical
unstructured and/or in unexplored environments. Important
location, and behaviors are developed for each sensor group.
research results have been achieved with techniques based on
In the second level, the low level behaviors are blended
artificial intelligence algorithms, such as fuzzy logic, to
according to an average of the actions or according to a pre-
implement behavior-based architectures for mobiles robots
defined behavior hierarchy. The researches in Tunstel et al.
(Huq et al. 2008).
(1997) proposed a similar concept of fuzzy behavior blending
Several research works have been focusing on the where a set of fuzzy rules is used to weigh the output fuzzy
development of methodologies to compose fuzzy behaviors sets of each behavior to control the robot.

978-3-902823-44-1/2013 © IFAC 91 10.3182/20130828-2-SF-3019.00020

IFAC AGRICONTROL 2013
August 27-30, 2013. Espoo, Finland
Later on, the approaches described in Abreu et al. (2001),
Aycard et al. (1997), Saffiotti (1999), Seraji et al. (2002) and
Tunstel et al. (1997) are applied for the development of
intelligent control systems with appropriate control actions
executed according to uncertainties related to unstructured
and dynamic environments. The work presented in
Vadakkepat et al. (2004) combines the approaches present in
Abreu et al. (2001), Saffiotti (1999) and Tunstel et al. (1997)
for the development of a team of three soccer robots. The
authors Yang et al. (2005) use the early results and presented
an architecture based on two layers with fuzzy behaviors
similar to those defined in Seraji et al. (2002). A layer is
responsible for determining navigation waypoints to reach the
final target. Another layer uses reactive behaviors to guide
the robot between waypoints. In the same way Remondini et
al. (2006) proposes a high-level behavior which is
responsible for performing a navigational plan by evaluating
and selecting sub-behaviors. Similarly, the authors in
Maaloufa et al. (2006) develop a hierarchical fuzzy controller
that combines inputs generated by a kinematic model of a
robot to control outputs to guide the robot between waypoints
to a desired end point.
Besides the research for methods of development and
coordination, the correct definition of the behavior is another
focus of current research in robotics architectures based on
fuzzy systems. It is discussed earlier in Murata et al. (1993)
and Willians et al. (2001) the definition of specific behaviors
to the navigation task in the own sphere of action of the
robots, respectively, the undersea environment and the
irregular and rocky surface of the mars planet. Recently, the
authors in Blackmore et al. (2005), Pedersen et al. (2005),
Blackmore (2007) and Bak et al. (2004) have discussed and
presented some specifications and methods for defining robot
behaviors to compose architectures related to agricultural
tasks which are currently performed, especially in the context
of precision agriculture.
In the work presented here, an approach to implement a
complex navigational behavior is proposed based on two
strategies: the composition of simple fuzzy behaviors
according to the physical location of the sensors and the
integration of these behaviors using fuzzy behavior
arbitration. Considering the physical location, three simple
behaviors have been developed: ‗straight in line‘, ‗follow
path‘ and ‗avoid obstacles‘. The complex navigational
behavior is implemented on a platform based on the mini-
robot Khepera. A relative locating system is developed based
on a robot motion model and odometric measurements for
mapping path and evaluating the navigational behavior on a
wooden labyrinth and on a simulated agricultural
environment.
The paper is organized as follows. The developed mini-robot
platform is presented in Section 2, where the relative locating
system for path mapping is also described. In Section 3, the
methodologies for fuzzy behavior composing and arbitration
are described as well as theirs implementation process in the
mini-robot platform. The experiments and results are
presented and discussed in Section 4. The conclusions are
given in Section 5.
92
2. SIMULATED MOBILE ROBOT PLATFORM
The platform used in this work has been implemented using
the commercial robot Khepera. It includes a Khepera basic
module, Khepera IO turret and a perceptual circuit mounted
on the IO turret based on VT935G LDR (light dependent
resistor) sensors. The LDR sensors are located in front of the
bottom circuit board of the base module to read the light
reflected by the floor following or looking for a path (dark
line). Both LDR sensors are placed side by side, about 4
millimeters above the ground and connected to a signal
conditioning circuit on the IO turret. The six front infrared
(IR) sensors of the basic module are grouped in three pairs of
adjacent sensors composing three perception areas: front, left
and right. Fig. 1 shows a top view of the platform.
Fig. 1. A top view of simulated robotic platform.
All routines are developed and executed on an external
computer connected to the Khepera platform by a serial link.
Both encoders of the Khepera platform are used to supply
odometric measurements for the relative locating system. The
incremental information (rotational displacements) over the
time are integrated by a motion model described in Murata et
al. (1993) for two wheeled vehicle with differential steering
(controlled independently).
The odometric measurement presents some well-known
problems that are due to accumulative errors. However, the
main aim of the experiments is to evaluate the implemented
behaviors and the ability of the platform to explore an
environment. The main causes of the accumulative errors can
be controlled by the experiments and the accuracy of the
relative localization system is enough for this work. Some
Methods to improve the locating system accuracy could be
applied for navigational control based on the odometry, such
as sensor fusion Chung et al. (2001).
3. FUZZY BEHAVIORS
The perceptual circuit based on LDR pair is used to the
‗follow path‘ behavior to simulate a vision system for the
identification of crop lines. The three groups of IR sensors
are used to implement the ‗avoid obstacle‘ based on three
related sub-modules defined according to the physical
location: ‗avoid left obstacle‘, ‗avoid right obstacle‘ and
‗avoid front obstacle‘. A deterministic process is chosen to
execute one of the ‗avoid obstacle‘ sub-modules. A fuzzy
IFAC AGRICONTROL 2013
August 27-30, 2013. Espoo, Finland

arbitrator controller defines the applicability of the ‗follow Follow Path

path‘, ‗avoid obstacle‘ and ‗straight in line‘ for each sample DIST

instant (all sensors are read). FAR MEDIUM CLOSE

NEG FLBR FLSR FLFR

3.1 Follow Path Behavior DIF ZERO SLSR FLFR FLFR

A calibration procedure is applied to minimize the effects of
different light conditions. Before starting, the navigation, the
Khepera platform is placed in two different ways in order to
expose the LDR pair to the reflected light by path painted
strip (major reading value) and to expose the LDR pair to the
reflected light by the cleared floor (minor reading value). The
difference between the maximum and the minimum readings
define a range for the readings. The crisp input values are
calculated based on normalized sensor readings according to
the range by (1).
R S -R SMIN
LDR S = (1)
R SMAX -R SMIN
Where:
S: indicates the left or right LDR sensor;
LDRN: left or right normalized sensor reading;
POS BLFR SLFR FLFR
Fig. 2. Knowledge base for the ‗follow path‘ behavior.
3.2 Avoid Obstacle Behavior
Two crisp input values, DIF and DIST, are defined for each
IR pair grouped according to the ‗avoid obstacle‘ behavior
sub-modules. However, the DIST inputs for IR pairs are
evaluated through the major value between the IR readings
(IRN) instead of the minor value. The DIST value indicates
the distance from a detect object and which group is closest
from the object (intensity and direction). The DIF values
denote which sensor of the group is closer from the object,
that means, the free and occupied spaces (direction).
Equations (4) and (5) are using to evaluate the crisp values
related to the IR pair grouped (G).

R: sensor reading during the navigation; 

DISTG =max IR GL ,IR GR  (4)

RMAX: major sensor reading in the calibration DIFG =IR GL -IR GR (5)
procedure (path painted strip); Six fuzzy controllers were developed for implementing the
‗avoid obstacle‘ behavior, two for each sub-module (one each
RMIN: minor sensor reading in the calibration for motor driver). Fig. 3 shows the knowledge base for the
procedure (clear floor). ‗avoid obstacle‘ behavior sub-modules.
For the LDR pair, two crisp inputs are defined: distance from Avoid Front Obstacle
path (DIST) and difference between the sensor readings
DIST
(DIF). The crisp values related to the LDR pair are evaluated
FAR MEDIUM CLOSE
from (2) and (3).
NEG FLFR SLFR BLFR

DIST=min LDR L ,LDR R  (2) DIF ZERO FLFR SLSR BLBR
DIF=LDR L -LDR R (3) POS FLFR FLSR BLFR

The DIST crisp value indicates how distant the LDR pair is Avoid Left Obstacle
becoming out of the path (intensity) and the DIF crisp value
DIST
denotes which sensor is more distant from the path
FAR MEDIUM CLOSE
(direction). The crisp inputs for ‗follow path‘ behavior are
composed of three fuzzy terms: far (FAR), medium NEG FLFR FLSR SLBR
(MEDIUM) and close (CLOSE) for DIST inputs, and DIF ZERO FLFR FLSR FLBR
negative (NEG), zero (Z) and positive (POS) for DIF inputs. POS FLFR FLFR FLSR
The behavior outputs are composed of two values obtained
by the centroid defuzzyfication method. Both outputs values Avoid Right Obstacle
are referred to each motor, left (L) or right (R). Three fuzzy DIST

terms are applied to describe the outputs: forward (F), stop FAR MEDIUM CLOSE
(S) and backward (B). Fig. 2 shows the knowledge base for NEG FLFR FLFR SLFR
the simple behaviors. Two fuzzy controllers were developed DIF ZERO FLFR SLFR BLFR
for the ‗follow path‘, one for each motor driver. The input POS FLFR SLFR FLBR
and output membership functions are implemented with
triangular and trapezoidal shapes.
Fig. 3. Knowledge base for the ‗avoid obstacle‘ behavior sub-
modules.

93
IFAC AGRICONTROL 2013
August 27-30, 2013. Espoo, Finland

A simple arbitration (non fuzzy arbiter) process applied the
major value among the values evaluate from (4) for choosing
which sub-module is operate in the sample instants.
The crisp inputs and outputs were composed applying the
same fuzzy terms as ‗follow path‘ behavior. Also, the input
and output membership functions are implemented with
triangular and trapezoidal shapes and the behavior outputs are
obtained by the centroid defuzzyfication method.
3.3 Fuzzy Behavior Arbitration
touching the walls and passes around the obstacles without
collision (Fig. 5).
The run executed where the robot platform starts the
navigation on the (0,0) point and follows the path until a lap
is complete. The discrete points are real sample points of the
closed path collected previously. The continuous line is
composed by 2030 sample points acquired in a way based on
the motion model. The results show that the robot is able to
perform several runs following the strip, which is
approximately 0.8 centimeters thick.

Contexts of application for which simple behavior were

defined according to the states where the behaviors are
competent: obstacle presence, path presence or free way.
Equations (2) and (4) are applied to identify those states in
the navigation and it was defined the crisp inputs
OBSTACLE, for the IR major reading, and PATH for the
LDR minor reading. Then, fuzzy rules were created to
associate the context to the behaviors and a fuzzy controller
was implemented to arbitrate them. The knowledge base for
the fuzzy behavior arbitration is presented in Fig. 4.
Behavior Blending
OBSTACLE

FAR MEDIUM CLOSE

P FAR STRAIGHT AVOID AVOID Fig. 5. Exploration Path.

A
T
MEDIUM STRAIGHT AVOID AVOID
H CLOSE FOLLOW FOLLOW AVOID Fig. 5 shows that when the robot finds the strip, it starts
following the strip until the end or until the strip comes
closed to the wall. In both situations, the robot restarts its
Fig. 4. Knowledge base for the for the fuzzy behavior exploration. Positive results have been achieved thus far for
arbitration. speeds up to 25% of rated vehicle maximum speed of the
Khepera platform.
The same linguistic terms as the ones used for the composing
behavior are defined for crisp inputs for behavior arbitration. A third kind of experiment was carried out on a simulated
The fuzzy terms applied to the outputs are STRAIGHT, agricultural environment with three lines of plants (crop
FOLLOW and AVOID, related to ‗straight in line‘, ‗follow lines). Fig. 6 illustrates the experiment and shows a mapping
path‘ and ‗avoid obstacle‘ behaviors respectively. Triangular acquired by using the relative positioning system.
and trapezoidal shapes are used and the output is obtained by
the centroid defuzzyfication method.
The knowledge base presented in Fig. 4 defines a hierarchy
where the ‗avoid obstacle‘ behavior has the highest priority,
‗follow path‘ has an intermediate priority and the ‗straight in
line‘ has the lowest priority. This hierarchical control
structure coordinates the behavior applicability and composes
the navigational complex behavior.

4. EXPERIMENTS AND RESULTS

Different experiments are conduced in a wooden labyrinth in
order to evaluate the abilities of robot exploration, obstacle
avoidance and looking for a path to follow. First, a closed
path painted strip (dark line) is used to evaluate the follow
path behavior. The path was painted on a 0.1-centimeter grid
paper. Others experiments are performed painting a strip in a
region of the labyrinth in such a way that part of this strip is
very close to a wall to test the arbitration of the behaviors. Fig. 6. Trajectory of the robot between the plants on a
Different runs are executed starting the robot in different simulated agricultural based the odometric measurement.
points and the odometric measurement systems was used for
data positioning estimation. The robot navigates without One of the deals of the robot on this experiment was
navigating based on the orientation of the ‗avoid obstacle‘

94
IFAC AGRICONTROL 2013
August 27-30, 2013. Espoo, Finland
behavior and the ‗straight in line‘ without the orientation of
spotted lines that would simulate the vision system. Thus, the
navigation strategy is restricted to straight in line and to self-
guide between two lines of plants (―wall plants‖) oriented by
the ‗avoid obstacle‘ recommendations. Some strips were
painted only in the end of the lines to guide the robot on the
end line maneuver.
Fig. 6 shows the forward and backward trajectory executed
by the robot which was guided by plants and the lines painted
on the floor. The discrete points in black correspond to the
points collected by odometry, and the starting point is marked
by a star. The discrete points in gray represent the return
trajectory. The points of all journeys count 410 points
acquired by odometric measurement system. The results of
the experiment illustrated in Fig. 5 and Fig. 6 shows that the
robot architecture allows the robotic platform to navigate
based on two behaviors only (except for the maneuver
operation) and it shows the ability of the architecture to
operate on adverse situations, as when it loses a sensorial
reference (strip painted).
Fig. 7 shows the detailed guidance and activation behavior
between A and E at the central street of the experiment also
shown in Fig. 6. In addition the Fig. 7 details regions B, C
and D between A and E.
(a)
(b)
Fig. 7. Detail of the trajectory of the robot (a) and the
illustrating chart of the behaviors activation (b) in the task of
navigating between rows of plants.
Fig. 7a shows a part of the backward trajectory of the robot
between A and E region. Fig. 7b illustrates the activation of
the behavior on the trajectory. The points form a discrete
trajectory composed of sixty (60) points with the coordinates
obtained by odometry. In region A, the robot navigates on the
lines painted on the floor that leads from the first street to the
95
second street between the rows of plants. The behavior
―following line‖ is used for guiding the robot through the end
line maneuver. The ―straight in line‖ behavior is activated to
keep the route to region B, which is identified as the next row
of plants at the right side of the robot. At that moment, the
arbiter deactivates the ―straight in line‖ behavior and
activates the ―avoid right obstacle‖ behavior. Than the arbiter
alternates the activation of the ―straight in line‖, ―avoid right
obstacle‖ and ―avoid left obstacle‖ according the current
context, which promotes a smooth positioning of the robot in
a safe region between the lines of the plants. At region C, the
―avoid left obstacle‖ is activated and at the D region, the
―avoid right obstacle‖ is activated. After this placement, the
robot behavior is driven by maintaining the route to region E,
where the painted line is identified and the robot is driven by
the behavior following line to the next row of the plants.
5. CONCLUSIONS
Generating simple fuzzy behaviors based specially on
physical sensor location reduces the development complexity
of the behavior-based control. The modularization of the
architecture by using decentralized fuzzy controllers
simplifies the implementation of the robotic architecture and
the fuzzy behavior arbitration is an easy and a feasible
method to implement the behavior cooperation and compose
a complex behavior.
The external processing to control the platform has proven to
be feasible for development. Also, the perceptual circuit
based on LDR sensors allows the implementation of a
simulated guidance vision system that could be extended and
used for tasks such as path mapping.
Some adaptations will be necessary for implementation of the
navigational behavior on a real agricultural robot, such as
replacing the IR sensor by sonars, laser scanners or
mechanical feelers, and replacing the LDR sensors by a
system for detecting the crop lines and crop maps. Also, it is
necessary generate other behaviors to increase the robustness
the architecture in other some situations such as when the
plants are missing. However, the results show the feasibility
of the approach on the platform and the behaviors that could
be used to improve a behavior based architectures for MARs.
REFERENCES
Abreu, A., Correia, L. (2001). Fuzzy Behavior-Based
Architecture for Decision Control in Autonomous Vehicles.
2001 IEEE Joint International Conference on Control
Applications (CCA) & International Symposium on
Intelligent Control (ISIC), Mexico-City, Mexico, 5-7.
Aycard, O., Charpillet, F. and Haton, J. (1997). A new
approach to design fuzzy controllers for mobile robots
navigation. In IEEE International Symposium on
Computational Intelligence in Robotics and Automation -
CIRA'97, Monterey, CA, IEEE Computer Society Press, 68-
73.
IFAC AGRICONTROL 2013
August 27-30, 2013. Espoo, Finland
Bak, T., Jakobson, H. (2004). Agricultural Robotic Platform
with Four Wheel Steering for Weed Detection. Biosystems
Engineering, 87 (2), 126-136.
Bakker, T.; Asselt, K. Van.; Bontsema, J.; Muller, J.; Straten,
G. Van. (2011). Autonomous navigation using a robot
platform in a sugar beet field. Biosystems Engineering, 109
(4), 357–368.
Barawid, O. C., Jr., Mizushima, A., Ishii, K., and Noguchi,
N. (2007). Development of an autonomous navigation system
using a two-dimensional laser scanner in an orchard
application. Biosystems Engineering, 96 (2), 139– 149.
Blackmore, B.S. (2007). A systems view of agricultural
robots. Precision Agriculture 7. ed. J. Stafford, V.
Wageningen Academic Publishers, 23-31.
Blackmore, S., Stout, B., Wang, M., & Runov, B. (2005).
Robotic agriculture—The future of agricultural
mechanisation?. In 5th European Conference on Precision
Agriculture, Uppsala, Sweden, 621–628.
Cariou, C., Lenain, R., Thuilot, B., Berducat, M. (2009).
Automatic Guidance of a Four-Wheel-Steering Mobile Robot
for Accurate Field Operations. Journal of Field Robotics,
26(6–7), 504–518.
Cheein, A.F.; Steiner, G.; Paina, G.P.; Carelli, R. (2011).
Optimized EIF-SLAM algorithm for precision agriculture
mapping based on systems detection. Computers and
Electronics in Agriculture. 78, 195–207.
Chung, H., Ojeda, L. and Borenstein, J. (2001). Accurate
Mobile Robot Dead-reckoning With a Precision-calibrated
Fiber Optic Gyroscope. IEEE Transactions on Robotics and
Automation, 17 (1), 80-84.
Darr, M. J., Stombaugh, T. S., & Shearer, S. A. (2005).
Controller area network based distributed control for
autonomous vehicles. Transactions of the ASAE, 48 (2), 479–
490.
Gan-Mor, S., Clark, R. L., and Upchurch, B. L. (2007).
Implement lateral position accuracy under RTK-GPS tractor
guidance. Computers and Electronics in Agriculture, 59, 31–
38
Huq, R., Mann, G.K.I., Gosine, R.G. (2008). Mobile robot
navigation using motor schema and fuzzy context dependent
behavior modulation. Applied Soft Computing, 8, 422–436
Johnson, D.A., Naffin, D.J., Puhalla, J.S., Sanchez, J. and
Wellington, C. K. (2009). Development and Implementation
of a Team of Robotic Tractors for Autonomous Peat Moss
Harvesting. Journal of Field Robotics, 26 (6–7), 549–571.
Kaizu, Y., Imou, K. (2008). A dual-spectral camera system
for paddy rice seedling row detection. Computers and
Electronics in Agriculture, 63, 49–56.
Kise, M., Zhang, Q., Rovira-Ma‘s, F. (2005). A stereovision-
based crop row detection method for tractor-automated
guidance. Biosystems Engineering, 90 (4), 357–367.
96
Maaloufa, E., Saada, M., Saliahb, H. (2006). A higher level
path tracking controller for a four-wheel differentially steered
mobile robot. Robotics and Autonomous Systems, 54, 23-33.
Murata, S. and Hirose, T. (1993). Onboard Locating System
Using Real-Time Image Processing for a Self-Navigation
Vehicle. IEEE Trans. Industrial Electronics, 40 (1), 145-154.
Nagasaka, Y., Saito, H., Tamaki, K., Seki, M., Kobayashi,
K., Taniwaki, K. (2009). An Autonomous Rice Transplanter
Guided by Global Positioning System and Inertial
Measurement Unit. Journal of Field Robotics, 26 (6–7), 537–
548.
Norremark, M.; Griepentrog, H.W.; Nielsen, J.; Sogaard,
H.T. (2012). Evaluation of an autonomous GPS-based system
for intra-row weed control by assessing the tilled area.
Precision Agriculture, 13, 149–162.
Pedersen, S.M., Fountas, S., Have, H.; Blackmore, B.S.
(2005). Agricultural robots: an economic feasibility study.
Precision Agriculture 7, 5, 589-595.
Remondini, D., Saffiotti, A. (2006). A modular, hierarchical,
configurable controller for autonomous robots. In: 12th
International Conference on Methods and Models in
Automation and Robotics, Miedzyzdroje, Poland, 585-590.
Saffiotti, A., Ruspini, E.H. and Konolige, K. (1999). Using
Fuzzy Logic for Mobile Robot Control. Chapter 5 in H. J.
Zimmermann (ed) Practical Applications of Fuzzy
Technologies, Handbooks of Fuzzy Sets series, 6, 185-205.
Kluwer Academic, MA.
Seraji, H., Howard, A. (2002). Behavior-Based Navigation on
Challenging Terrain: A Fuzzy Logic Approach. IEEE
Transactions on Robotics and Automation, 18 (3), 308-321.
Tunstel, E., Lippincott, T., Jamshidi, M. (1997). Behavior
hierarchy for autonomous mobile robots: fuzzy-behavior
modulation and evolution. Int. J. Intel. Automat. Soft
Comput., 3 (1) 37–49.
Vadakkepat, P., Miin, O.C., Peng, X., Lee, T.H. (2004).
Fuzzy behavior-based control of mobile robots. IEEE Trans.
Fuzzy Syst. 12 (4) 559–565.
Willians, S. B., Newman, P., Rosenblatt, J., Dissanayake, G.,
Durrant-Whyte, H. (2001). Autonomous Underwater
Navigation and Control. Robotica, 19 (5), 481-496.
Yang, X., Moallem, M., Patel, R.V. (2005). Layered Goal-
Oriented Fuzzy Motion Planning Strategy for Mobile Robot
Navigation. IEEE Transactions on Systems, Man, and
Cybernetics - Part B: Cybernetics. 35 (6), 1203-1213.