You are on page 1of 6

5th IFAC Conference on Sensing, Control and Automation for

Agriculture
Agriculture
Agriculture
5th IFAC
August
August
5th Conference
IFAC14-17,
14-17, 2016. on
on Sensing,
2016. Seattle,
Conference Seattle, Control
Washington,
Washington,
Sensing, Control and
and Automation
USA
USA Automation for
for
August Available
14-17, 2016. Seattle, Washington,
Agriculture USA online at www.sciencedirect.com
Agriculture
August
August 14-17,
14-17, 2016.
2016. Seattle,
Seattle, Washington,
Washington, USA
USA
ScienceDirect
Steering Control Strategies for a Four-Wheel-Independent-Steering
IFAC-PapersOnLine 49-16 (2016) 039044 Bin Managing
Steering Control Strategies for a Four-Wheel-Independent-Steering
Robot Bin Managing
Steering Control Strategies for a Four-Wheel-Independent-Steering
Steering Control Strategies for a Four-Wheel-Independent-Steering
Robot Bin
Bin Managing
Managing
Robot
Robot
Yunxiang Ye*. Long He**. Qin Zhang***Yunxiang Ye*. Long He**. Qin Zhang***
Yunxiang Ye*. Long He**. Qin Zhang***
*Center
*Center for for Precision
Precision & & Automated Yunxiang
Yunxiang Ye*.
Automated Agricultural
Agricultural Ye*.
Systems,
Systems,
Long
Long He**.
He**. Qin
Washington
Washington
Zhang***
QinState
Zhang***
State University, Prosser,
University, Prosser, WA WA 99350,
99350, USA USA (e-mail:
(e-mail:
*Center for Precision & Automated Agricultural yunxiang.ye@ Systems, Washington
wsu.edu). University, Prosser, WA 99350, USA (e-mail:
State
yunxiang.ye@ wsu.edu).
*Center
*Center
**Center for
for Precision
Precision
for Precision
Precision &
& Automated
Automated
& Automated
Automated Agricultural
Agricultural
Agricultural Systems,
yunxiang.ye@
Systems,
Systems,Washington
wsu.edu).
Washington
Washington State
State University,
University,
State University Prosser,
Prosser,
(e-mail: WA 99350,
99350, USA
WAlong.he@
long.he@ USA (e-mail:
(e-mail:
wsu.edu).
**Center for & Agricultural Systems, Washington State University (e-mail: wsu.edu).
**Center
***Center for for Precision
for Precision
Precision & & Automated
& Automated Agricultural
Automated Agricultural
Agricultural yunxiang.ye@
Systems,
yunxiang.ye@
Systems, Washingtonwsu.edu).
Washington
wsu.edu).State
Washington State University
State University (e-mail:
University (Corresponding long.he@
(Corresponding author, wsu.edu).
author, Tel:
Tel: 509-
509-
***Center Systems,
**Center
***Center
**Center for
forfor Precision&&
Precision
Precision & Automated
Automated
Automated Agricultural
Agricultural
786-9360,Systems,
Agricultural Systems,
Systems,
e-mail: Washington
Washington
Washington
qinzhang@ State
State
State
wsu.edu). University(Corresponding
University
University (e-mail: long.he@
(e-mail: long.he@ wsu.edu).
author, Tel: 509-
wsu.edu).
786-9360, e-mail: qinzhang@ wsu.edu).
***Center
***Center for for Precision
Precision & & Automated
Automated Agricultural 786-9360,Systems,
Agricultural Washington
e-mail: qinzhang@
Systems, Washingtonwsu.edu).State
State University
University (Corresponding
(Corresponding author, author, Tel:
Tel: 509-
509-
Abstract: Automatic steering 786-9360,
systems
786-9360, are e-mail:
a
e-mail: qinzhang@
standard
qinzhang@ wsu.edu).
function for
wsu.edu). robotic agricultural equipment.
Abstract: Automatic steering systems are a standard function for robotic agricultural equipment.
Abstract:
Ackerman Automatic
Ackerman steering is
steering steering
is the
the most systems
most common are
common typeaof
type ofstandard
steering function
steering mechanism
mechanism forononrobotic agriculturalmaking
such equipment,
such equipment, equipment.
making them
them
Abstract:
Ackerman
Abstract:
perform as Automatic
steering
asAutomatic is
car-like vehicles.steering
the most
steeringBecause
vehicles. systems
common
systems
Because are
type aof standard
are akinematic
of their
their steering function
mechanism
standardconstraints,
kinematic function for
constraints, for on robotic
such
isrobotic
it is agricultural
equipment,
agricultural
quite difficult
difficult equipment.
making
to maneuver
maneuver them
equipment. car-
perform car-like of it quite to car-
Ackerman
perform
Ackerman
like vehicles as steering
vehicles car-like
steering
effectivelyis
is the inmost
vehicles.
thein most common
Because
common
orchards due type
of their
type
to of
of steering
kinematic
confined steering
working mechanism
constraints,
mechanism on
space, itconstrained
on such
is quite
constrained equipment,
such difficult
equipment,
by physical making
to maneuver
physical making them
car-
them
boundaries
like effectively orchards due to confined working space, by boundaries
perform
like
suchvehicles
perform as car-like
as
as tree
tree car-like
effectively
rows andvehicles.
vehicles.
other Becausedue
in orchards
Because
obstacles. of
of their
toremove
Totheirremovekinematic
confined
kinematic
such constraints,
working space,
constraints,
technical itconstrained
it is quite
is
difficulties, quitemoredifficult
difficult to maneuver
bysophisticated
physical
to
sophisticated maneuver
boundaries car-
car-
steering
such as rows and other obstacles. To such technical difficulties, more steering
like
such
like vehicles
as tree
vehicles
mechanisms effectively
rows and
effectively
and steering in
other
in orchards
obstacles.
orchards
control due
due To
strategies to
to confined
remove
confined
are such
requiredworking
technical
working
for space,
space,
robotic constrained
difficulties,
constrained
agricultural more by
by physical
sophisticated
physical
equipment. boundaries
steering
boundaries
In order to
mechanisms and steering control strategies are required for robotic agricultural equipment. In order to
such
mechanisms
such as
as
conveniently tree
tree rows
and
rows
manage and
steering
and other
other
fruit obstacles.
control
obstacles.
bins in confined To
strategies
To remove
are
remove
tree such
required
such
aisles technical
for robotic
technical
constrained difficulties,
agricultural
difficulties,
by high more
more
density sophisticated
equipment.
sophisticated
tree rows, a In steering
order
steering
robotic to
bin
conveniently manage fruit bins in confined tree aisles constrained by high density tree rows, a robotic bin
mechanisms
management and
conveniently
mechanisms manage
and steering
fruit
steering
system, control
binsbin-dog
control
called bin-dogstrategies
in confined
strategiessystem, are
tree required
are aisles
required
implementable for
for robotic
constrained
robotic agricultural
inbytypical
high density
agricultural
typical equipment.
Washingtontree rows,
equipment. Statea In order
robotic
In order
tree to
bin
to
fruit
management system, called system, implementable in Washington State tree fruit
conveniently
management
conveniently
orchards has manage
system,
manage
been fruit
fruit
developed. bins
called
bins in
bin-dog
in
This confined
confined system,
bin-dog tree
tree aisles
aisles
system constrained
implementable
constrained
adopted a in by
by high
typical
high density
Washington
density tree
tree
four-wheel-independent-steering rows,
rows, Statea
a robotic
tree
robotic bin
fruit
(4WIS) bin
orchards has been developed. This bin-dog system adopted a four-wheel-independent-steering (4WIS)
management
orchards
management
mechanism hasas system,
asbeen
system,
the solution
solutioncalled
developed.
called bin-dog
This bin-dog
bin-dog
to achieve
achieve thesystem,
system,
necessary implementable
system adopted aand
implementable
drivability in typical
typical Washington
four-wheel-independent-steering
in maneuverability.
and Washington
maneuverability. To provideState
State tree
provide tree fruit
(4WIS)
fruit
adequate
mechanism the to the necessary drivability To adequate
orchards
mechanism
orchards
controllabilityhas asbeen
has beenthe developed.
solution
developed.
to this
this to This
achieve bin-dog
This bin-dog
4WIS, aa four-mode
four-mode the system
necessary adopted
drivability
systemstrategy,
steering a
adoptedincluding
strategy, four-wheel-independent-steering
and maneuverability.
a four-wheel-independent-steering
including Ackermann steering, To provide
steering, active (4WIS)
adequate
active(4WIS)
front
controllability to 4WIS, steering Ackermann front
mechanism
controllability
mechanism
and rear as
as the
rear steering
steering the solution
to (AFRS),
this 4WIS,
solution
(AFRS), to
to aachieve
crab four-mode
achieve
steering, the
theand
andnecessary
steering
necessary
spinning, drivability
strategy,
drivability
were designedand
and maneuverability.
including
designed forAckermann
maneuverability.
this bin-dog To
bin-dogsteering,
To provide
provide
to manage
manage adequate
active front
adequate
fruit bins
and crab steering, spinning, were for this to fruit bins
controllability
and rear steering
controllability
effectively in in theto
to this
(AFRS),
this
the confined 4WIS,
4WIS,crab
confined orchard a
a four-mode
steering,
four-mode
orchard space and
space .. This steering
spinning,
steering
This control strategy,
were
strategy,
control system including
designed
including
system makes
makes itfor Ackermann
this
Ackermann
it possiblebin-dog
possible for steering,
to manage
steering,
for thethe bin-dog active
fruit
active
bin-dog system front
bins
front
system
effectively
and rear
rear steering
effectively
andswitch
to switch in the(AFRS),
steering
between confined
(AFRS), crab
crab
four steering
steeringsteering,
orchard
steering,spaceand
modes and spinning,
. using
This control
spinning,
using were
were
the most
most designed
system makesfor
designed
appropriate forit this
this bin-dog
possible
bin-dog
steering for to
mode tothemanage
tobin-dog
manage fruit
fruit
complete bins
system
bins
all
to between four modes the appropriate steering mode to complete all
effectively
to switch
effectively
maneuvering in
in the
between
the
tasks confined
four
confined
in a most orchard
steering
orchard
effective space
modes
spaceway. .
. This
using
This
To control
the
control
design most
suchsystem
system
a makes
appropriate
makes
control it
it
system, possible
steering
possible
it is for
mode
for the
the
important tobin-dog
complete
bin-dog
to system
system
understand all
maneuvering tasks in a most effective way. To design such a control system, it is important to understand
to
the switch
maneuvering
to switch
influence between
tasks
between infour
of major a most
four
major steering
effective
steering
factors, modes
modes
such way. using
To design
using
as longitudinal the
the most
longitudinal such
most aappropriate
speed control
appropriate system,
and control
control steering
steering
gain mode
it is important
mode
on to
to to
performance complete
understand
complete underall
all
the influence of factors, such as speed and gain on performance under
maneuvering
the influence
maneuvering
different tasks
of
tasks
steering in
major
in a
a
modes most
most effective
factors, such
effective
when tracking way.
as
way. To
To design
longitudinal
design
various such
such
paths. a
speed
a
Incontrol
and
control
this system,
control
system,
paper, a it
gain
it is
is
pure important
on performance
important
pursuit to
to understand
under
understand
method was
different steering modes when tracking various paths. In this paper, a pure pursuit method was
the
different
the influence
steering
influence
implemented of
of major
using modes
major factors,
factors, such
when navigation
tracking
such as
as longitudinal
various
longitudinal paths. speed and
and control
In this
speed paper,
control again
pureon
gain performance
pursuit
onwith bothmethod
performance under
was
under
implemented using aa GPS-based
GPS-based navigation to evaluate
to evaluate auto-steering
auto-steering performance
performance with both Ackermann
Ackermann
different
implemented
different
steering and steering
using
steering
and AFRSmodesmodes
a GPS-based
modes. when
when tracking
navigation
Fieldtracking
tests were various
to
various
were paths.
evaluate
paths. to
conducted In this
auto-steering
In assess
to paper,
this paper,
assess a
performance pure pursuit
with
a pure pursuit
the navigation
navigation both
performance method
Ackermann
methodusing was
was
using
steering AFRS modes. Field tests conducted the performance
implemented
steering
implemented and
different steering using
AFRS
using
steering a GPS-based
modes.
a GPS-based
strategies Field navigation
tests
when navigation were
tracking different to evaluate
conducted
to evaluate
different paths. auto-steering
to assess
auto-steering
The results performance
the
results indicated navigation
performance
indicated with
thatwith both
performance Ackermann
both Ackermann
by properly
properly using
selecting
different strategies when tracking paths. The that by selecting
steering
different and AFRS
steering modes.
strategies Fieldtracking
when tests were
differentconducted
topaths. toa assess
The assess the navigation performance using
steering
steeringand
aa steering AFRS
strategy
strategy formodes.
for Field
the situation,
the situation, tests
it is
it were
is possible
possible conducted
to achieve
achieve toa results
satisfactory
satisfactoryindicated
the path that
navigation
path byperformance
tracking
tracking properly
performance
performanceselecting
using
for
for
adifferent
steering
different
tracking steering
strategy
steering
curvy strategies
for the
strategies
paths or when
situation,
when
completing tracking
it
tracking
tasks is different
possible
different
such
tracking curvy paths or completing tasks such as merging and cornering. as topaths.
achieve
paths.
merging The
The
and a results
satisfactory
results
cornering. indicated
path
indicated that
that by
tracking
by properly
performance
properly selecting
selectingfor
aa steering
tracking
steeringcurvy strategy
pathsfor
strategy fororthe situation,
completing
the situation, it
tasks is possible
is such
itglobal possible to
to achieve
as mergingachieve andaacornering.
satisfactory
satisfactory path
path tracking
tracking performance
performance for
for
2016, IFAC
Keywords:
Keywords:
tracking curvy (International
Steering
Steering
paths control
control
or Federation
systems,tasks
systems,
completing of such
global Automatic
positioning
positioning
as Control)
merging Hosting
systems,
systems,
and by Elsevier Ltd. All rights reserved.
four-wheel-independent-steering
four-wheel-independent-steering
cornering. system.
system.
tracking curvy paths or completing tasks such as merging and
Keywords: Steering control systems, global positioning systems, four-wheel-independent-steering system.cornering.
Keywords: Steering
Keywords: Steering control
control systems,
systems, global global positioning
positioning systems,
systems, four-wheel-independent-steering
four-wheel-independent-steering system. system.
1. INTRODUCTION
1. INTRODUCTION bounded driving
bounded driving paths paths and and obstacles
obstacles in in confined
confined job job sites
sites are
are
1. INTRODUCTION bounded
taken into driving
into paths andpath
account, obstacles
path planningin confinedand job sites are
equipment
The taken account, planning and equipment
The application
application of of 1.auto-steer
1.
auto-steer technology
technology on
INTRODUCTION
INTRODUCTION on agricultural
agricultural taken bounded
boundedinto
maneuvering driving
driving paths
account,
for paths
automaticand
andpathobstacles
obstacles planning
nonholonomic in
in confinedand
confined
systemsjob sites
equipment
job sites
become are
are
The
equipment application
in North of auto-steer
America can technology
be traced on to
back agricultural
the maneuvering for automatic nonholonomic systems become
equipment in North America can be traced back to the early
early taken maneuvering
taken
difficult. into
into account,
for automatic
account, path
path planning
nonholonomic
planning and
systems
and equipment
become
equipment
The
1920s.application
equipment
The in North
application
A patent
patent of
of auto-steer
America
auto-steer
depicted cantechnology
be traced
technology
automatic steering on
back agricultural
to
on systems the early
agricultural
systems that difficult.
maneuvering
1920s.
equipment
1920s.
A
A in North
patent
depicted
America
depicted
automatic
can
automatic be
steering
traced
steering back to
systems the
that
early
that An maneuvering for
difficult. for automatic
automatic nonholonomic
nonholonomic systems systems become
become
equipment
could guide
could guideinaaNorth tractorAmerica
tractor to follow
to followcanfurrows
be traced
furrows backaatofarm
across
across the early
farm field
field An auto-steered vehicle
difficult.
auto-steered vehicle for for tree
tree fruitfruit orchards
orchards use use needs
needs to to
1920s.
could A
guide patent
a tractordepicted
tothe automatic
follow furrows steering
across systems
a farm units that
field difficult.
1920s.
(Willrodt,
(Willrodt, A 1924).
patent
1924). depicted
With
With the automatic
development
development steering
of computing
of systems
computing that An
units auto-steered
effectively
effectively and safely
and vehiclemaneuver
safely for tree fruit
maneuver orchards
in aisles
in aisles use needs
formed
formed by tree
by treeto
could
(Willrodt,
could
and guide
guide
sensor aa tractor
1924). With
tractor
technologies, to
tothe follow
follow
auto-steer furrows
development
furrows across
of
across
technology aa farm
computing
has farmgainedfield
units
fielda An
An
rows, auto-steered
effectively and
auto-steered
as well as vehicle
safely
vehicle
quickly for
for
and tree
maneuver
tree
preciselyfruit
in
fruit orchards
aisles
orchards
steer in use
formed
use
and needs
by
needs
out tree
those to
to
and sensor technologies, auto-steer technology has gained a rows, as well as quickly and precisely steer in and out those
(Willrodt,
and
(Willrodt,
much sensor
wider 1924).
1924). With
technologies,
With
application the
the in development
auto-steer
development
agriculture. of
technology
of computing
has
computing
Numerous gainedunits
units
reported a effectively
rows, as
effectively
aisles well
(Freitasand
and aset safely
quickly
safely
al., maneuver
and
maneuver
2012; precisely
Subramanian in
in aisles
steer
aislesand formed
in and
formed
Burks, outby
by tree
those
tree
2007).
much wider application in agriculture. Numerous reported aisles (Freitas et al., 2012; Subramanian and Burks, 2007).
and
much sensor
wider technologies,
application auto-steer
in agriculture.technology
Numerous has
has gainedreported a aislesrows, as well
(Freitas as
aset quickly
al.,of and precisely steer in and
and out those
and
studies
studies sensor technologies,
verified
verified that auto-steer
that auto-steering
auto-steering technology
could
could help
help achieve
achieve gainedgooda One
good rows, as
One unique
uniquewellfeature
feature of2012;
quickly andSubramanian
working
working precisely
in
in orchard
orchard steerandin Burks,
environments
environments out2007).
those
is
is its
its
much
studies wider application in agriculture. Numerous reported aisles (Freitas et al., 2012; Subramanian and Burks, 2007).
driving accuracy while maintaining a high driving speed and confined space which makes correcting poses on an aisle its
much
driving verified
wider
accuracy that
application
while auto-steering
in
maintaining could
agriculture.
a high help
Numerous
driving achievespeed good
reported
and One
aisles
confinedunique
(Freitas feature
space et of
al.,
which working
2012;
makes in orchard
Subramanian
correcting environments
and
poses Burks,
on an is
2007).
aisle or
or
studies
driving
studies
at the verified
accuracy
verified
same time that
while
that auto-steering
maintaining
auto-steering
releasing the could
a high
could
operator help
driving
help
from achieve
achievespeed
monotonous good
and
good One
confined
One
steeringunique
unique
the feature
space
featurewhich
vehicle of
of working
makes
working
into an aislein
in orchard
correcting
orchard
without environments
poses on
environments
multiple antrials is
aisle
is
andits
or
its
at the same time releasing the operator from monotonous steering the vehicle into an aisle without multiple trials and
driving
at accuracy
the same timewhile maintaining
releasing andtheimprove aa high
operator driving
from speed
speed and
monotonous and confined
steering space
the which
vehicle intomakes correcting
an aisle poses on an aisle or
driving
driving accuracy
to reduce
to reduce while
fatigue
fatiguemaintaining
and improve high driving
driving
driving performance
performance confined
corrections
corrections space
very
very which makes
challenging
challenging if
if thewithout
correcting
the equipment
equipment multiple
poses on
uses
uses antrials
aisleand
only
only or
the
the
at
at the
driving same
to
the sameet
(OConnor time
reduce
et time releasing
fatigue
releasing
al., 1996;
1996; and
Zhang the
theand
andoperator
improve
operator
Qiu, 2004; from
driving
fromRoberson
2004; monotonous
performance
monotonous
Roberson steering
corrections
and steering
Ackerman the vehicle
very
the steering
vehicle into into
challenging an aisle
if the
an aisle without
mechanism. without
equipment multiple
uses trials
only
multiple trials and
Four-wheel-independent- and
the
(OConnor al., Zhang Qiu, and Ackerman steering mechanism. Four-wheel-independent-
driving
(OConnor to
to reduce
driving 2014).
Jordan, 2014). reduce fatigue
et al., 1996;
fatigueZhangand
and improve
and Qiu, driving
improve 2004; Roberson
driving performance
performance and corrections
Ackerman
corrections
steering very
steering
very challenging
mechanism.
challenging if
if thethe equipment uses
uses only
Four-wheel-independent-
equipment onlyboththe
the
Jordan, steering (4WIS)
(4WIS) is
is one
one solution
solution capable
capable of
of improving
improving both
(OConnor
Jordan,
(OConnor 2014). et al.,
et al., 1996;
1996; Zhang
Zhang and and Qiu,
Qiu, 2004;
2004; Roberson
Roberson and and Ackerman
steering
Ackerman
the drivability steering
(4WIS)
steering
andis one
the mechanism.
solution
mechanism.
maneuverability Four-wheel-independent-
capable of improving
Four-wheel-independent-
of orchard both
vehicles
Most the drivability and the maneuverability of orchard vehicles
Most auto-steer
Jordan, 2014).
auto-steer technologies for
2014). technologies for conventional
conventional agriculturalagricultural steering the (4WIS)
drivability andis
is one solution
thespace.
maneuverability capable ofof improving
of orchard both
Jordan,
Most auto-steer
equipment were technologies
based on for conventional
on Ackerman
Ackerman steering, agricultural
steering
operating
operating (4WIS)
in
in confined
confined one solution
space. When
When capable
designing
designing roboticvehicles
improving
robotic both
mobile
mobile
equipment were based steering, aa car-like
car-like the
the drivability
operating in
drivability
orchard equipment, and
confined
and the
the
the maneuverability
space. When
maneuverability
development of
designing
of of orchard
robotic
orchard
capable vehicles
mobile
vehicles
auto-steer
Most
Most auto-steer
equipment
vehicle auto-steer were technologies
steering based
technologies
mechanism for
on Ackerman conventional
for which steering,
conventional
which agricultural
is typically a car-likeaa operating
agricultural
typically orchard equipment, the development of capable auto-steer
vehicle steering mechanism is orchard in
in confined
equipment, space. When designing robotic mobile
equipment
vehicle were
steering
equipment were
nonholonomic based on
mechanism
basedAdmissible
system. Ackerman
on Ackerman
Admissible which steering,
is
steering,
directions a
typically
of motion car-like
a car-like
motion a operating
systems
systems
are systems
for
for 4WIS the
confined
4WIS development
space.
vehicles
vehicles When is of capable
is designing
essential.
essential. roboticauto-steer
Research
Research mobile on
on
nonholonomic system. directions of are orchard equipment, the development of capable auto-steer
vehicle steering
nonholonomic
vehicle
restricted steering
for system.mechanism
Admissible
mechanism
nonholonomic systems which
directions
whichdue tois their
is of typically
motion
typically
kinematic area developing the needed technologies has been reported on
a
orchard
developing for the 4WIS
equipment, needed vehicles
the development
technologies is essential.
of
has capable
been Research
auto-steer
reported in
in
restricted for nonholonomic systems due to their kinematic systems developing
systems
recent for
for
years. the4WIS
4WISneeded
For vehicles
technologies
vehicles
example, is
is
Bak essential.
has been
essential.
and Research
reported
Research
Jakobsen (2004) on
in
on
nonholonomic
restricted
nonholonomic
constraints. for Thus system.
nonholonomic
system.
when Admissible
systems
Admissible
environmental directions
due
directions to of
their
of
constraints motion
kinematic
motionsuch are
are
as recent years. For example, Bak and Jakobsen (2004)
constraints. Thus when environmental constraints such as developing recent
developing years.the
the needed
For example,
needed technologies
technologies Bak and has
has been
Jakobsen
been reported
(2004)
reported in
in
restricted
constraints.for
restricted forThus nonholonomic systems
systems due
when environmental
nonholonomic to
to their
their kinematic
dueconstraints such as recent
kinematic years. For example,
recent years. For example, Bak and Jakobsen (2004) Bak and Jakobsen (2004)
constraints.
constraints. Thus
Thus when environmental constraints
when environmental constraints such as 39 such as
Copyright 2016
Copyright 2016 IFAC IFAC 39
2405-8963 2016,
Copyright IFAC (International Federation of Automatic Control)
2016 IFAC 39 Hosting by Elsevier Ltd. All rights reserved.
Peer reviewunder
Copyright 2016 responsibility
IFAC of International Federation of Automatic
39 Control.
Copyright 2016 IFAC
10.1016/j.ifacol.2016.10.008 39
IFAC AGRICONTROL 2016
40 Yunxiang Ye et al. / IFAC-PapersOnLine 49-16 (2016) 039044
August 14-17, 2016. Seattle, Washington, USA

reported the development of a robotic platform with 4WIS wheels and four hydraulic rotary actuators (L10-9.5, HELAC,
system for weed detection. Its navigation controller utilized WA, USA) to steer those wheels. Eight bidirectional
two control points (front end and rear end) independently to proportional electrohydraulic valves (Series VPL, Parker
minimize the distance needed to reference a trajectory. Hannifin, OH, USA) were used to control those actuators
Oksanen and Backman (2007) used kinematic and dynamical individually. The rotational speed of each wheel was
models to steer both front and rear axles of their 4WIS measured using four optical incremental encoders (TR2
platform. Nagasaka et al. (2004) introduced an autonomous TRU-TRAC, Encoder Products, ID, USA) and their steering
field watching-dog robot with 4WIS, with different steering angles were measured using four absolute encoders (TRD-
modes to implement appropriate steering strategies for NA720NWD, Koyo Electronics Industries, Tokyo, Japan).
driving between rows and turning around at headland of a All steering wheels have a maximum steering angle of 90
field. which makes it possible for the 4WIS system to place its ICR
at any location.
To create a solution for efficiently and safely managing fruit
bins, researchers from Washington State University (WSU) 2.2 Steering Modes of 4WIS
Center for Precision and Automated Agricultural Systems
In order to simplify the control strategy and at the same time
(CPAAS) have developed a robotic bin-dog system with a
maximize maneuverability of the bin-dog system in confined
4WIS system for transporting bins in typical Pacific
working space, four steering modes, namely Ackermann
Northwest (PNW) tree fruit orchards (Ye et al., 2016). A
steering, active front and rear steering (AFRS), crab steering,
4WIS system is able to turn each wheel independently to
and spinning steering (as illustrated Figure 2), were used in
increase the maneuverability to omnidirectional by properly
this study. When Ackermann and AFRS steering are
coordinating the turning of all wheels. Such a capacity allows
implemented, the bin-dog will change its orientation and
a robot to turn at any instantaneous center of rotation (ICR)
driving direction and keep moving in longitudinal directions.
and provides for the ability to drive the bin-dog in confined
In spinning, it changes the bin-dog orientation without
orchard environments. A capable bin-dog steering control
changing its position and crab steering just does the opposite.
system plays a role in achieving required path tracking and
Thus spinning and crab steering could be effective when only
bin maneuvering performance, which is influenced by
orientation or position correction is required.
longitudinal driving speed, steering modes, control gain, and
path patterns. The goal of this study was to investigate the
influence of the identified factors, including robot
longitudinal driving speed, steering modes, control gain, and
path patterns, on the bin-dog steering performances on path
tracking and bin maneuvering in orchard environments. The
outcomes from this study will provide guidelines for
developing adequate steering strategies for the bin dog robot
to perform required bin management operations.
2. MATERIALS & METHODS
2.1 Bin-Dog research Prototype and 4WIS System
A bin-dog research prototype (Figure 1) for bin management Fig. 2. Four steering modes of 4WIS
in orchard environment was designed and fabricated as a Our previous work (Ye et al., 2016) has described the
research platform to investigate the influencing factors for detailed information of a 4WIS control system for this bin-
different auto-steering strategies. dog. Laboratory tests on paved ground showed that under the
maximum longitudinal speed of 1.2 ms-1, this bin-dog 4WIS
system could achieve a maximum steering speed of 30s-1
with a maximum error of 3.0. To avoid excessive
mechanical stress caused by large steering angle, the
minimum turning radiuses of Ackermann steering and AFRS
were set at 1.7 m and 2.3 m respectively.
2.3 GPS-Based Navigation System
As one of the simple geometric path trackers, pure pursuit
method showed a robust performance (Snider, 2009; Rains et
Fig. 1. A 4WIS robotic bin-dog research prototype al., 2014). By pure pursuit method, a bicycle model is used to
simplify vehicle models. For a robot working in orchard
The 4WIS system on this bin-dog prototype was made of an environment, its longitudinal speed generally is low (less
electrohydraulic system driven by a 9.7 kW gas engine than 10 ms-1), thus both Ackermann steering and AFRS
(LF188F-BDQ, LIFAN Power, China). Driven by a load could be simplified using bicycle models without losing
sensitive pump (Series P1, Parker Hannifin, OH, USA), the much accuracy. Mostly commonly used in Ackermann
drive chain used four low-speed high-torque hydraulic steering analysis as shown in Figure 3(1), the pure pursuit
motors (TL series, Parker Hannifin, OH, USA) to drive the

40
IFAC AGRICONTROL 2016
August 14-17, 2016. Seattle, Washington, USA
Yunxiang Ye et al. / IFAC-PapersOnLine 49-16 (2016) 039044 41

method could be easily modified for AFRS applications shape of a figure of eight as shown in Figure 4, were used as
(Figure 3(2)). the test courses. Because it consists of changing radius,
Lemniscate curve is often used to assess maneuverability of a
steering system.

Fig. 3. Pure pursuit geometries for Ackermann and AFRS


steering Fig. 4. Lemniscate curves for testing maneuverability of a
In Figure 3, R is the turning radius; ld is the look-ahead steering system
distance; is the angle between look-ahead vector and Such Lemniscate curve can be described using polar equation
heading vector of the vehicle; is the vehicle steering angle below:
and eld is the distance from goal point to heading vector. The
pure pursuit method connects a control point on vehicle and a = cos(2) (5)
goal point on path using a circular arc. The goal point (gx, gy)
is found by calculating a look-ahead distance from the In which both r and are its polar coordinates. The value of
control point to the path. For Ackermann steering, the control a equals the distance between origin to the right most point of
point is typically set at the mid-point of its rear axle, while the curve. Thus the size of Lemniscate curve which closely
for AFRS, the control point is often set at its geometric relates with its curvatures can be controlled by changing the
center. The curvature of the circular arc for Ackermann value of a. The minimum radius of curvature of a Lemniscate
steering could be calculated using following equation curve can be calculated using equation below:
(Snider, 2009):
= (6)
3
2
= 2 (1) Figure 4 also illustrates the curvatures of test courses with

different values of a. As the value of a increases, the size of
Based on the geometry of bicycle model for Ackermann Lemniscate curve as well as the minimum radius of curvature
steering, by defining vehicle steering angle as the steer angle also increase and the rate of change of curvature becomes
in the bicycle model, it could then calculated: milder.
2
= tan1 ( ) = tan1 ( 2 ) (2) In this study, three sets of tests were conducted on uneven

soft soil ground in natural environment using both
The curvature of the circular arc and steering angle for AFRS Ackermann and AFRS steering modes. In the first set tests,
can be calculated using a similar approach: the bin-dog was guided using an auto-steer system to track
2 four Lemniscate curves of different sizes (a = 5, 6, 7.5, and
4 = 2 (3) 10 m) with a fixed longitudinal driving speed of 0.6 ms-1 and

fixed look-ahead distance of 1.0 m for Ackermann steering
4
4 = tan1 ( ) (4) and 0.75 m for AFRS steering. In the second set tests, five
2
different longitudinal speeds (0.4, 0.5, 0.6, 0.8, and 1.0 ms-1)
Once desired driving speed and vehicle steering angle are were used with a fixed look-ahead distance of 1.0 m for
determined, the speed and steering angle of each wheel can Ackermann steering and 0.75 m for AFRS steering to track a
be calculated accordingly. Lemnicate curve with a value of 10 m. In the third set of
A RTK-GPS receiver (AgGPS432, Trimble, Sunnyvale, CA, tests, five different look-ahead distances (0.8, 0.95, 1.0, 1.1,
USA) with claimed 3 cm positioning accuracy was used for and 1.2 m) for Ackermann steering, (0.65, 0.7, 0.75, 0.85,
tracking the positioning of the bin-dog in all tests. The and 0.95 m) for AFRS steering were used with a fixed
heading angle of bin-dog was measured using an IMU longitudinal speed of 0.6 ms-1 to track the same Lemniscate
(AHARS-30, Xsens, Netherlands). curve as used in first set of tests. The range of look-ahead
distances for a steering mode was determined so that it could
2.4 Experiment Design cover a look-ahead distance which gave a satisfactory path
2.4.1 Path Tracking on Curvy Paths tracking performance. Therefore, the test could reveal the
influence of look-ahead distances around a satisfactory
Capable of following curvy paths accurately is one of the setting over path tracking performance. All the combinations
basic requirements for an auto-steer system. To investigate were conducted with three replications.
the influence of different factors on path tracking
performance on curvy paths, Lemniscate curves, having a 2.4.2 Merging, Straight Line Tracking and Cornering

41
IFAC AGRICONTROL 2016
42 Yunxiang Ye et al. / IFAC-PapersOnLine 49-16 (2016) 039044
August 14-17, 2016. Seattle, Washington, USA

As mentioned above, steering a bin-dog into an aisle from were used with the look-ahead distance of Ackermann
headland is frequently needed during bin management. This steering being fixed at 1.2 m.
process involves actions including merging which allows
3. RESULTS AND DISCUSSION
platform to reach desired path without changing its original
heading angle and cornering which allows platform to turn 90 3.1 Path Tracking on Curvy Paths
to enter aisles from the headland. In order to design a steering
Figure 6 and Table 1 provide the results obtained from
strategy for tasks of merging and cornering, it is important to
Lemniscate curves tracking performance from field tests. Two
investigate influence of factors such as longitudinal driving
speed, control gain, and steering modes on work efficiency important parameters that influence path tracking performance
and spatial requirements. Figure 5 shows a test course used for are minimum radius of the curvature and rate of curvature.
The minimum turning radius of Ackermann steering system
testing this function.
for this bin-dog research platform was 2.3 m while the
minimum radius of curvature of Lemniscate curve with a-
value of 5 m was 1.7 m. Thus it is impossible for Ackermann
steering to perfectly track this curve. AFRS has a minimum
turning radius of 1.7 m which gave it better performance when
the radius of curvature of the path is small. The path tracking
trajectory showed in Figure 6(1) revealed that the lateral error
of the bin-dog in tacking the smaller curve was noticeably
larger. As an AFRS system has much smaller minimum
turning radius, and it could achieve a much better path
tracking performance on the small curve than Ackermann
steering (as shown in Figure 6(2)).
Fig. 5. Test course for merging, straight line tracking and
cornering
The test path consists of two segments connected by a 90
turning point as illustrated in Figure 5. In the series of tests,
the bin-dog started from point A, which was 1 m to the right
of the beginning of headland path (represented by the vertical
line in the figure), merged to the desired headland path, took
a 90turn to the left entering the aisle, then followed the path
in the aisle (presented by the horizontal line) until reaching
the target location. As the testing path has a right turn which
is mathematically discontinuous, it is impossible for
Fig. 6. Path tracking of Lemniscate curves of different radius
nonholonomic systems such as Ackermann and AFRS
systems to follow this path precisely. This study planned and Table 1 Path tracking performances on Lemniscate curve
conducted all three sets of tests on uneven soft soil ground to of different sizes with Ackermann and AFRS steering
investigate the performance of different steering modes on
Lateral error
such a path. In the first set tests, four different longitudinal a-
Ackermann Steering 4WCS
speeds (0.4, 0.5, 0.6, and 0.8 ms-1) were used with a fixed value
Mean Std-dev Max Mean Std-dev Max
look-ahead distance of 1.1 m for Ackermann steering and 1.1 (m)
(m) (m) (m) (m) (m) (m)
m for AFRS steering to track the desired path. In the second 5.0 0.35 0.33 1.39 0.13 0.13 0.58
set, five look-ahead distances each for Ackermann steering 6.0 0.10 0.09 0.41 0.15 0.06 0.38
(0.95, 1.0, 1.1, 1.2 and 1.5 m) and for AFRS (0.8, 0.9, 1.0, 7.5 0.26 0.16 0.57 0.07 0.06 0.23
1.1 and 1.2 m) were used with a fixed longitudinal speed of 10.0 0.05 0.04 0.60 0.10 0.08 0.44
0.6 ms-1 to track the path. In the third tests, three steering
modes were used in combination to complete the entire test Table 2 lists the results for tests in different longitudinal
course. Crab steering was firstly used to merge with the speeds in detail. Both Ackermann steering and AFRS
headland path from the start point A. The navigation system successfully completed the test course under all speed levels
kept calculating steering angle using the location information (expect for AFRS under 1.0 ms-1 due to frequently engine
of bin-dog control point. Once the control point was close stalls). As expected, a lower longitudinal speed could always
enough to point B on the desired path, the steering mode results in a higher tracking accuracy (with a lower mean
automatically switched to Ackermann steering. The Bin-dog lateral offset, standard deviation, and maximum offset). That
then kept tracking the headland path till its control point is because a lower longitudinal speed allows the bin-dog
reached point C which was 0.47 m (half of the wheel base) steering actuating system to have sufficient time to respond
below the corner. The bin-dog would stop at point C, turn the to input steering command to correct any observed path
steering mode into spinning and, rotate 90on its geometry tracking error. It could also help to reduce lateral wheel slip
center to set its orientation in parallel to aisle path. The bin- and allowed pure rolling on all four wheels. Thus a bin-dog at
dog then switched steering modes back to Ackermann low longitudinal speed could stay on the reference curve and
steering to track in-aisle path till it reached the target point. maintain a satisfactory and stable path tracking performance.
Five longitudinal speeds (0.4, 0.5, 0.6, 0.8, and 1.0 ms-1) This finding revealed a need that the bin-dog steering system

42
IFAC AGRICONTROL 2016
August 14-17, 2016. Seattle, Washington, USA
Yunxiang Ye et al. / IFAC-PapersOnLine 49-16 (2016) 039044 43

have good control performance to satisfactorily support high Figure 7 shows the trajectories of the bin-dog turning 90into
speed operations as it would require more frequent an aisle using AFRS only mode and multi-modes. In the
corrections to make the vehicle track precisely on its desired Figure, D1 and L1 represent for the width and length for
path under all disturbances. The steering system tends to completing merging while D2 and L2 represent for the width
oscillate in large amplitude when the system response and length for cornering. For both Ackermann and AFRS
frequency is low resulting in large maximum lateral offset steering, some noticeable overshoots can be observed during
and standard deviation. merging and cornering on those places. When the auto-steer
system tries to complete merging and cornering, it needs to
Table 2. Path tracking performances of Ackermann and
correct both offset error and heading angle error. However, it
AFRS steering at different longitudinal speeds
is impossible for Ackermann steering or AFRS steering to
Lateral error eliminate both errors at the same time, and some overshoots
Speed Ackermann Steering AFRS Steering will unavoidably occur.
(ms-1) Mean Std-dev Max Mean Std-dev Max
(m) (m) (m) (m) (m) (m)
0.40 0.06 0.03 0.19 0.03 0.02 0.16
0.50 0.13 0.06 0.39 0.06 0.04 0.37
0.60 0.05 0.04 0.31 0.10 0.08 0.44
0.80 0.19 0.10 0.60 0.10 0.10 0.66
1.00 0.21 0.23 1.12 / / /
Table 3 summarized the obtained path tracking performance
of the bin-dog tested when different look-ahead distances
were applied. The look-ahead distance is the only control
gain for tuning auto-steer system using the pure pursuit
method.
Fig. 7. Path tracking result for merging and cornering using
Table 3. Path tracking performances of Ackermann
active-front-and-rear-steering only and multi-mode.
steering and AFRS steering with different look-ahead
distances Tables 4 and 5 provide a summary of the required space for
Lateral error Lateral error
correcting merging and cornering overshoots (D1 and D2).
Ackermann Steering AFRS Steering Table 4. Merging and cornering performance of different
ld ld
Std- Std- steering modes at different longitudinal speeds
(m) Mean Max (m) Mean Max
dev dev
(m) (m) (m) (m) v T1 T2 D1 D2 L1 L2
(m) (m) Mode
0.80 0.06 0.04 0.23 0.65 0.20 0.14 0.82 (ms-1) (s) (s) (m) (m) (m) (m)
0.95 0.05 0.04 0.31 0.70 0.06 0.06 0.31 0.40 29.70 21.83 2.71 2.70 7.05 0.23
1.00 0.07 0.06 0.42 0.75 0.10 0.08 0.44 0.50 22.63 15.83 2.73 2.84 7.80 0.92
Multi-
1.10 0.23 0.16 0.68 0.85 0.12 0.08 0.36 0.60 19.63 15.10 2.76 3.03 10.50 0.50
mode
1.20 0.30 0.13 0.80 0.95 0.16 0.16 0.63 0.80 14.17 13.17 2.79 2.78 9.37 3.04
1.00 11.07 14.87 2.69 2.99 11.17 4.24
Typically, the look-ahead distance is tuned for different 0.40 16.70 35.27 3.13 4.33 3.84 7.13
longitudinal speeds. Based on equations (2) and (4), a longer 0.50 12.20 38.10 3.13 5.13 3.67 9.69
look-ahead distance will generate a smaller steering angle to AKMN
0.60 8.47 33.53 3.16 5.29 3.38 10.01
correct the same level of lateral errors and cause an under- 0.80 7.40 / 3.28 / 4.73 /
correction. Because of that a long look-ahead distance 0.40 17.53 34.70 3.16 4.13 3.03 2.24
usually responds slowly in correcting a lateral error. A long AFRS 0.50 10.90 24.73 3.22 4.15 2.91 5.23
look-ahead distance may also cause overlooking some 0.60 8.37 30.87 3.20 4.62 3.15 6.50
signature points just in front of the moving vehicle and result In table, v: Longitudinal speed; T1, T2: Time used for
in a poor tracking performance. Therefore when tracking merging and cornering; D1, L1: space and length required for
curvy paths with changing radius, it should control the look- merging; D2, L2: space and length required for cornering;
ahead distance carefully to avoid a too long distance to ignore AKMN: Ackermann steering.
the just-in-front signatures and result in large lateral This space is increased as the longitudinal speed increases for
offsets. A short look-ahead distance is usually more sensitive
both Ackermann and AFRS steering. As discussed earlier, a
to lateral errors which makes it often generate relatively
long look-ahead distance tends to ignore some just-in-front
larger steering correction signal than needed and cause over-
path signature points such as the discontinuity point for
correction. The steering tends to oscillate and the system
merging and cornering. So while it could help reduce
becomes unstable leading to poor tracking performance when overshoots and smooth the trajectories, it will have some
look-ahead distance is set too short as in the case of AFRS difficulty making precise 90turns.
steering with ld being set at 0.65 m. In general, long look-
ahead distance is suitable for high speed operation and short The length required for merging and cornering (L1 and L2)
look-ahead distance is suitable for low speed. will decrease as the longitudinal speed goes down. However
there is no clear relationship with look-ahead distance.
3.2 Path Tracking on Merging and Cornering

43
IFAC AGRICONTROL 2016
44 Yunxiang Ye et al. / IFAC-PapersOnLine 49-16 (2016) 039044
August 14-17, 2016. Seattle, Washington, USA

Generally, a short look-ahead distance tends to make dynamically choose optimal steering strategy for the situation
oscillation in path tracking and a long look-ahead distance is in terms of the longitudinal speed, look-ahead distance, and
slow to steer the vehicle back on track. Thus both could result path curvature.
in long L1 and L2.
ACKNOWLEDGEMENTS
The time required to complete merging and cornering (T1 and
This research was partially supported in part by United States
T2) is often related to robot longitudinal speed. Normally, a
Department of Agriculture (USDA)s Hatch and Multistate
higher longitudinal speed could substantially reduce the time
Project Funds (Accession No 1005756 and 1001246), USDA
needed for merging the robot back to track, but may also
National Institutes for Food and Agriculture competitive
require a longer distance to complete cornering and turning in
grant (Accession No 1003828), and Washington State
a tree aisle. The data acquired from field tests revealed that the
University (WSU) Agricultural Research Center
differences in time required to complete a turn were rather
(ARC). Any opinions, findings, conclusions, or
small within all tested longitudinal speeds.
recommendations expressed in this publication are those of
Table 5. Merging and cornering performance of different the author(s) and do not necessarily reflect the view of the
steering modes with different look-ahead distances U.S. Department of Agriculture and Washington State
University.
ld T1 T2 D1 D2 L1 L2
Mode
(m) (s) (s) (m) (m) (m) (m) REFERENCES
0.95 12.97 31.60 3.33 5.20 3.39 13.24
1.00 10.10 40.30 3.27 5.78 3.38 10.01 Bak, T., H. Jakobsen. (2003). Agricultural robotic platform
AKMN 1.10 8.47 33.53 3.16 5.29 3.19 7.27 with four wheel steering for weed detection. Biosystems
1.20 9.50 25.05 3.10 4.29 3.72 6.09 Engineering. 87(2): 125-136.
1.50 10.97 23.83 2.96 4.04 4.71 6.87
Freitas, G., B. Hamner, M. Bergerman, S. Singh. (2012). A
0.80 15.90 37.20 3.90 5.14 4.45 7.90
practical obstacle detection system for autonomous orchard
0.90 17.85 33.70 3.72 5.14 2.89 6.60
AFRS 1.00 10.25 25.90 3.24 4.79 3.15 6.50 vehicles. IEEE International Conference on Intelligent
1.10 8.37 30.87 3.20 4.62 3.19 5.55 Robots and Systems. Vilamoura, Algarve, Portugal.
1.20 11.40 22.60 2.27 3.89 3.39 13.24 O'Connor, M., T. Bell, G. Elkaim, B. Parkinson. (1996).
As afore discussed, compared to Ackermann and AFRS Automatic steering of farm vehicle using GPS. The 3rd
steering, crab steering is more effective in translating which International Conference on Precision Agriculture.
allows correcting the position error without changing vehicle Minneapolis, Minn.
heading orientation. Thus it is more effective in merging. As
shown in Table 4 and 5, it required minimum space to merge Nagasaka, Y., Q. Zhang, T. E. Grift, Y. Kanetani, N. Umeda,
back on track. The long length for merging in multi-mode T. Kokuryu. (2004). An autonomous field watching-dog
steering was because the correction was targeted at setting robot for information collection in agricultural fields.
location of point B. If the B location was moved closer to ASABE Paper No. 043092. St. Joseph, Mich.: ASABE.
point A, it could substantially reduce the required merging Oksanen, T., J. Backman. (2013). Guidance system for
length. In cornering, spinning mode is the most effective agricultural tractor with four wheel steering. The IFAC
method to correct vehicle heading orientation without Bio-Robotics Conference, Osaka, Japan.
changing its location. Thus it took least space to complete
cornering. Rains, G.C., A.G. Faircloth, C. Thai, R.L. Raper. (2014).
Crab steering can be smoothly switch to or back from either Evaluation of a simple pure pursuit path-following
Ackermann or AFRS steering without stopping the vehicle, algorithm for an autonomous, articulated-steer vehicle.
but stops are required for switching to or back from spinning. Applied Engineering in Agriculture, 30(3): 367-374.
If continuous driving is required, even though spinning Roberson, G.T., D. L. Jordan. (2014). RTK GPS and
requires the least space for turning, Ackermann or AFRS automatic steering for peanut digging. Applied
steering is still preferred. Engineering in Agriculture, 30(3): 405-409.
4. CONCLUSIONS Subramania, V., T. Burks. (2007). Autonomous vehicle
In this study, with the influence of different longitudinal turning in the headlands of citrus groves. ASABE Paper
speeds, look-ahead distances, and path curvatures of different No. 071015. St. Joseph, Mich.: ASABE.
steering modes were investigated in terms of their auto-steer Willrodt, F.L. (1924). Steering attachment for tractors. US
performances when tracking a few specially designed test Patent No. 1506706.
courses. The results indicated that both Ackermann and
AFRS steering could achieve satisfactory path tracking Ye, Y., L. He, Q. Zhang. (2016). Bin-Dog A Self-Propelled
performance within the permissible limits of corresponding Platform for Bin Management in Orchards. Journal of
physical configurations. Obtained results also verified that Field Robotics. (In peer-review)
crab steering and spinning could effectively reduce the spatial Zhang, Q., H. Qiu. (2004). A dynamic path search algorithm
requirement for merging and cornering compared to for tractor automatic navigation. Transactions of the
Ackermann or AFRS steering. Based on this work, the future ASAE, 47(2): 639-646.
work will focus on developing a dynamic control system to

44