MOCONT LOCATION MODULE: A CONTAINER LOCATION

SYSTEM BASED ON DR/DGNSS INTEGRATION

Joseba Landaluze, Victoria del Río, Carlos F. Nicolás, José M. Ezkerra, Ana Martínez

IKERLAN Research Centre, Arizmendiarrieta 2, 20500 Arrasate/Mondragon, The Basque Country (Spain)

Keywords: DR/DGNSS, DR, GPS, DGPS, inertial navigation system, location system

Abstract: The problem of identifying and adequately positioning containers in the terminal yard during handling with

Reach Stackers still remains to be solved in an appropriate manner, while this is extremely important in

making the identification and positioning operations automatic. A precise knowledge in the Terminal

Operating System (TOS) of such data in Real Time would have considerable economic impact on the

logistic treatment of operations. The MOCONT system sets out to provide a solution to this lack. In

particular, the MOCONT Location Module establishes the position of the container in the yard while it is

being handled by a Reach Stacker. This system is based on the integration of the Differential Global

Navigation Satellite System (DGNSS) with a Dead Reckoning (DR) inertial system. This article presents the

general characteristics of the MOCONT Location Module, its structure, and the structure of data fusion,

besides some results obtained experimentally.

1 INTRODUCTION

The problem of localising the containers in the yard

of container terminals had been partially solved for

terminals equipped with huge machines, such as

Rubber Tyre (RTG) and Rail Mounted Gantry

Cranes (RNG) or even Straddle Carriers, but is still a

problem for those terminals equipped with Reach

Stackers or Front Loaders. Moreover, the automatic

identification of containers on board the handling

machine is still a problem remaining to be solved.

The European projects MOCONT (“Monitoring

the yard in container terminals”, IST-1999-10057)

and MOCONT-II (“Monitoring the yard in container

terminals – Trials”, IST-2001-34186), aimed at

providing a system to track the containers in the yard

in Real Time. The projects aimed at developing a

system that automatically identified containers and

localised them when moved by small machines,

called Reach Stackers.

This paper presents the Location Module of the

MOCONT system, developed in said European

projects, and more particularly, the development of

the inertial navigation system or Dead Reckoning

(DR) system. The Location Module is based on the

integration of a Differential Global Navigation

Satellite System (DGNSS) and an inertial DR

system. Data are integrated by means of a Kalman

Filter. This system reckons the position of the

vehicle at all times, improving estimates supplied by

the DGNSS. The exact position of the container is

determined by means of a transformation of

coordinates from the vehicle body since the length

and angle of inclination of the boom are known. The

system is particularly useful when the GNSS does

not supply quality data or is interrupted, due to few

satellites being accessible, multipath phenomena or

due to working inside container canyons or in dark

areas. The DR inertial system is able to continue

estimating vehicle position with precision for short

time intervals, with bounded errors, until GNSS

signals with sufficient quality are available.

In the literature, different structures of inertial

systems appear for land vehicles. The most common

case is the use of a sensor for rotation speed of yaw

angle and odometric information obtained from

vehicle wheels (Aono, 1999; Ramjattan, 1995).

Other redundant sensors are often used to help, each

being different depending on the type of vehicle and

application in question (Aono, 1999; Zhang, 1999).

In some cases, especially in vehicles for agricultural

purposes, a digital compass or a geomagnetic

direction sensor are used instead of a yaw angle

speed gyroscope (Benson, 1998; Zhang, 1999). In

order to avoid errors, which may be introduced by

odometric sensors on wheels due to slipping,

Sukkarieh (1999) proposes an Inertial Measurement

Unit (IMU), comprising three accelerometers, three

227

Landaluze J., del Río V., Nicolás C., Ezkerra J. and Martínez A. (2004).

MOCONT LOCATION MODULE: A CONTAINER LOCATION SYSTEM BASED ON DR/DGNSS INTEGRATION.

In Proceedings of the First International Conference on Informatics in Control, Automation and Robotics, pages 227-234

DOI: 10.5220/0001132802270234

 SciTePress

gyroscopes and two pendular gyroscopes, by which

vehicle acceleration and yaw and tilt rotation speeds

are obtained. Rogers (1999) presents an inertial

system consisting of a low cost fibre optic rate

gyroscope and of a radar ground speed sensor.

Due to the specifications of the MOCONT

project, according to which it was not possible to

change the structure of the Reach Stacker machines

and nor was it possible to install encoders in the

wheels to obtain odometric information, initially a

structure based on a triad of accelerometers and a

triad of solid state gyroscopes was selected.

Although this set up may be valid for on-road

vehicles, once field data were obtained for a Reach

Stacker in normal work tasks, it was concluded that

this set up was not valid for such a machine; due to

the low speeds and considerable degree of vibration

to which this machine is subject during operations,

the accelerometers did not provide information on

machine movements on the yard, and only the yaw

rate ant tilt rate gyroscope signals could be used

(Landaluze, 2003). Therefore, a sensor structure

similar to that proposed in (Rogers, 1999) was

chosen. Finally, since a subcentimetric receiver was

replaced by a submetric receiver, and taking into

account the “non-collocation” between the Reach

Stacker chassis sensors and the GPS antenna, fitted

on the highest point of the Reach Stacker boom,

sensoring was completed with a digital compass.

This paper firstly presents an overview of the

MOCONT project, followed by an explanation of

the structure of the Location Module in the

MOCONT system. Then follows a description of the

DR inertia system and the Kalman Filter

implemented. Lastly, some experimental results are

shown, as well as a statistical evaluation of the

results obtained by the Location Module of the

MOCONT system in the course of the MOCONT-II

project.

2 OVERVIEW OF THE MOCONT

SYSTEM

The MOCONT project was presented as a new

landmark in the application of telematics in

intermodal transport, especially in the control of

container terminals. The main objective of the

project was to develop a system to identify the

position of containers in the yard in Real Time.

Although said follow-up problem had already been

partially solved in the case of large loading and

unloading cranes (Rubber Tyre RTD, Rail Mounted

Gantry Crane RMG, Straddle Carriers), this was and

is a problem in terminals with Reach Stacker

machines. It was in these machines, therefore, where

the project comes to bear (Figure 1).

The Reach Stacker is an off-road machine used

to handle containers in the terminal yard. It is

characterised by a small body with an extensible

boom (the arm) mounted over of the operator cabin.

It is equipped with a spreader, namely the handling

device used to pick and keep containers by the

machine itself. The Reach Stacker can stack up to

the fourth height (up to the fifth height in case of

empty containers).

Taking into account the main objective of the

project, the MOCONT system should perform the

following operations:

• Identify the container, reading the identification

code when picked up by each Reach Stacker.

• Follow each movement of the container in the

terminal yard while being handled by the Reach

Stacker, recording (i.e., the position of the

container in the yard – row, column, height)

where the container is picked up or released.

• Inform on the position of the container and its

identification without the intervention of human

operators.

YARD

Ship Gate

Buffer

Rail Gate

Buffer

Truck Gate

Figure 1: Overall scheme of a container terminal and Reach Stacher container handling machine.

ICINCO 2004 - ROBOTICS AND AUTOMATION

228

• In Real Time, update the position of any

container handled by a Reach Stacker in the

Terminal Operating System (TOS).

In order to obtain the objectives proposed, the

MOCONT system incorporates three different

modules (Figure 2): the Location Module, the Visual

Identification Module and the Synchronisation and

Communication Module. The Location Module

determines the position of the container in the

terminal yard; the Visual Identification Module

reads the container’s identification code; the

Synchronisation and Communication Module

acquires the identification and position data on the

container and informs the TOS.

3 STRUCTURE OF THE

LOCATION MODULE

The Location Module consists of two subsystems:

the DR subsystem and the DGNSS subsystem. This

last one has two different parts: the GPS receiver

and the GNSS Processing Module.

In the final MOCONT Location Module

Trimble’s Ag132 GPS receiver is used at the heart of

the GNSS and, therefore, of the location system, for

the positioning of the Reach Stacker. The Ag132

GPS receiver combines high-performance GNSS

reception with radio-beacon DGNSS capability in a

single durable waterproof housing, ideal for use in

the yard environment. The receiver uses differential

GNSS to provide sub-metre accuracy.

Differential GNSS requires two or more

receivers. One receiver, called the reference or base

station, is located at a known point to determine the

GNSS measurement errors. This could be housed on

the roof of the main administration buildings, to

allow easy access and constant monitoring. An

unlimited number of AgGPS receivers, sometimes

called rovers, collect GNSS data at unknown

locations onboard each Reach Stacker. Over a radio

band, the reference station broadcasts correction

values, which are applied to the Ag132 GPS receiver

positions. Errors common at both the reference and

rover receivers and then removed from the solution.

The performance of the Ag132 GPS receiver is

improved by direct GNSS augmentation with height

aiding. Height aiding improves the solution by

enhancing satellite visibility, and reducing the

positioning challenge from a three-dimensional to a

two dimensional problem. Using a DTM of the port

and the current location of the Reach Stacker, an

interpolation algorithm provides an accurate

measure of the current ground height. With

knowledge of the Reach Stacker geometry, the boom

extension and boom inclination, the height of the

GNSS antenna on board the vehicle, and indeed the

height of the container carried by the Reach Stacker,

can be continually computed.

In addition, the Location Module provides

complimentary DR augmentation for periods when

GNSS positioning with height aiding is not possible.

The DR subsystem consists of a Processing Unit and

some DR sensors, by means of which the Reach

Stacker position is continuously estimated. The

GNSS Processing Module continually provides the

DR subsystem with the current position from the

Ag132 GPS receiver (in projected UTM

coordinates) and some indication of the quality of

that position fix (by means of a covariance matrix of

the computed parameters). In return the DR

subsystem continually updates the GNSS Processing

Module with the best estimate of the current

position.

The GNSS Processing Module will then pass the

position information to the driver and the rest of the

MOCONT system.

The GNSS Processing Module, which interfaces

with the Ag132 GPS receiver, the DR subsystem,

Terminal Operating System (TOS)

DR Inertial

avigation

DGNSS Visual Identification

Synchronisation and

Communication

Middleware

LM VIM

MOCONT

SCM

Figure 2: MOCONT functional architecture.

GNSS

Processing

Unit

Port DTM

Ag132 GPS

Receiver

Processing

Unit

Dis

MOCONT

sensors

Boom

sensors

Figure 3: Scheme and flow data of the Location Module.

MOCONT LOCATION MODULE: A CONTAINER LOCATION SYSTEM BASED ON DR/DGNSS INTEGRATION

229

the boom sensors and the Synchronisation and

Communication Module, has been developed using

the robust and compact AgGPS 170 Field Computer.

The AgGPS 170 is designed to withstand the

environmental extremes that are typical of the

container port environment.

A scheme of the Location Module, with its

components and the flow of data, is shown in Figure

3 and Figure 4. As can be observed, the

communication between the two subsystems of the

Location Module is by means of a CAN bus. The

communication between the GNSS subsystem and

the Synchronisation and Communication Module is

also by another CAN bus (Figure 4). The values

measured by the boom sensors are supplied by the

MOCONT middle-ware through this bus.

4 DR SUBSYSTEM DESIGN

4.1 Description of the DR subsystem

Although initially different sensorings were tested

for the DR subsystem, a sensor structure similar to

that proposed by Rogers (1999) was finally selected.

Likewise, replacing the MS750 subcentimetric GPS

receiver in the final structure of the Location

Module with an Ag132 submetric receiver, made it

necessary to complete sensoring in the DR

subsystem with a digital compass. As a result of

these changes, the final structure used for the DR

subsystem was as follows:

• The DR Processing Module, which consisted of

a sandwich of three PC/104 boards: a CPU

based on a 233 MHz Pentium Processor, an I/O

data acquisition board and a 2-channel CAN

communication board.

• The DR sensor set. This included the following

sensors: a solid state gyroscope to measure the

speed of rotation around the yaw axis; a radar

technology Ground Speed Sensor, which

provided the forward/backward speed of the

vehicle; a stop/direction sensor, in order to

detect if the vehicle is stopped or not, as well as

the movement direction of the vehicle, forward

or backward; a digital compass, to measure the

vehicle heading angle.

Figure 4 shows the main elements of the

Location Module and the structure of the DR

subsystem. A laptop computer was used to monitor

and configure the DR subsystem, as well as to

collect raw data. Most of the elements of the DR

subsystem were included in two boxes, the Location

Module box and the Heading box, as they appear in

Figure 5.

4.2 DR/DGNSS integration

Although different data fusion algorithms were

tested, a kinematical Kalman Filter was finally

chosen due to its simplicity and the good results

obtained. For the navigation equations, it is assumed

that the vehicle is moving on a tangent-plane, as it

was a point, so the positioning involves locating the

vehicle in cardinal directions: N-S-E-W. Figure 6

shows the local level geographic navigation and the

Gyro

Radar

Sensor

Stop/direct

Sensor

Laptop

Computer

PC/104

Proc. Module

AgGPS 170

GNSS

Processing Module

Ag132

GPS

receiver

24 V

CAN

RS-232

DR subsystem

GNSS subsystem

MOCONT

Digital

Compass

Figure 4: Elements of the Location Module.

Figure 5: Heading box and Location Module box.

Figure 6:Local level geographic and body frames.

ICINCO 2004 - ROBOTICS AND AUTOMATION

230

body reference frames. It is assumed that the only

transformation between these two frames is via the

heading angle

The DR navigation system implements a

distance travelled (integrated velocity) and a travel

direction. The distance travelled is referenced to the

body frame, which is then transformed into a local

level geographic navigation frame through the angle

. This implementation assumes that other vehicle

attitudes, i.e., roll and pitch, are sufficiently small as

to be ignored.

The body referenced velocity is represented as a

nominal velocity v, defined such that the X-

component is along the primary direction of travel,

plus velocity errors as a result of speed sensor scale

factors

The computed body referenced velocity is

represented as

vvv

⋅+= (1)

where v is the measured or estimated vehicle

velocity.

Assuming that the body to navigation frame

transformation and body-referenced velocity are

approximately constant over a small time interval,

the sampling time, it can be written:

ψεψψ

sinsinsin

coscoscos

⋅⋅+⋅=⋅=

vvve

vvvn

(2)

where

and e

are the velocities in the local frame

and v is the measured vehicle velocity.

The heading angle rate could be expressed as

follows:

ψψψψψ

bVV +

⋅

(3)

where

: gyroscope gain

: gyroscope scale factor error

b : gyroscope bias

V : measured gyroscope voltage

The speed sensor scale factor error

, the

gyroscope scale factor error

and the gyroscope

bias

are modelled as random-walk processes.

From equations (1), (2) and (3) the continuous-time

state-space realisation for the DR/DGPS is deduced:

vxhy

wuxfx

)(

),(

(4)

where

[

]

′

Vvu ,

[

]

′

ψψ

εψε

benx

and

w and v are random variables. The augmented

state equations for the DR subsystem can be stated

in direct form or in terms of residual errors, and

therefore the structure of the Extended Kalman Filter

can be deduced, the measurement vectors being:

DGPS

Receiver

Radio

Modem

Radio

Modem

Gyroscope

Speed sensor

Stop/dir sensor

GNSS

Processing Module

n, e, R

UTC,

,h,

PDOP, #sat,

GPS quality

DGPS

Receiver

Base Station

Rover Unit

DR Software Im

lementation

Compass

Extended Kalman Filter

Boom

sensors

Figure 7: DR/DGNSS integration scheme.

MOCONT LOCATION MODULE: A CONTAINER LOCATION SYSTEM BASED ON DR/DGNSS INTEGRATION

231

⎥

⎦

⎤

⎢

⎣

⎡

)( with a frequency of 200 Hz

⎥

⎦

⎤

⎢

⎣

⎡

kz )( with a frequency of 2 Hz

Figure 7 shows the DR/DGNSS integration

scheme. As outputs of the prediction steps of the

Extended Kalman Filter, a vehicle position estimate

is obtained with a frequency of 200 Hz, although the

correction of state estimate is applied with a

frequency of 2 Hz.

5 EXPERIMENTAL RESULTS

Figure 8 shows how all the MOCONT Location

Module elements are fitted in a Reach Stacker. In

the course of the MOCONT project, a system

prototype was tested at the Terminal Darsena

Toscana (TDT) in Livorno (Italy). In the MOCONT-

II project, however, the system was implemented on

8 prototypes tested intensively over a six-month

period at the TDT and at the VTE terminal in Genoa

(Italy), taking a considerable number of data for

statistical analysis.

The main objective of the Location Module is to

locate precisely the container in the terminal yard.

The yard is the surface of the terminal dedicated to

the container storage. It is subdivided into modules,

each one composed of corridors (carriageways used

to move containers within a module and between

different modules) and groups of slots. A group of

slot is referred to as a lane. Lanes are numbered

using capital letters, starting from A. One slot is

uniquely identified within a lane by its yard

coordinates: row, column and height (Figure 1).

Therefore, the container position in the

corresponding slot should be accurately estimated.

During the project MOCONT-II more than 5600

container position messages were collected. Results

presented in the Final Public Report (MOCONT-II,

2004) led to the conclusion that the performance of

the MOCONT Location Module was 99.7% of

correct localisation resulting from the wide set of

trials carried out.

The advantages of the DR subsystem are

highlighted during the work inside container

canyons, and generally, in the work near high stacks

of containers which, on the other hand, are the most

important moments for the correct position

identification as this is when the Reach Stacker is

picking up or releasing a container in a given slot.

Figure 9 shows typical results of a Reach Stacker

handling a container. It corresponds to data collected

in Livorno, where the cases of container canyons

were quite common. The figure shows the GNSS

estimate for the position of the vehicle chassis and

the estimate conducted by the DR subsystem, which

Figure 8: Installation of all elements of the MOCONT Location Module on a Reach Stacker.

ICINCO 2004 - ROBOTICS AND AUTOMATION

232

is transmitted to the GNSS on a frequency of 2 Hz.

North and east relative position components, with

respect to the starting point O, appear. The vehicle

moves forward from the origin O to point A (instant

28 s). It then moves backward to point B (instant 30

s) from where it moves forward once again to point

E (instant 85 s), after having passed through points

C (instant 40 s) and D. It is at point E until instant 91

s, when it then moves backward to point F (instant

104 s). Covariance is not very good (always greater

than 1), but also, between points C and D, it is

approximately 3. As shown in the graph of Figure 9,

when the covariance of the GNSS estimate is

relatively low (about 1), the DR estimate continues

to be in line with that of the GNSS. When the value

of the covariance is large, particularly in the D-E-F

stretch, the DR estimate is based mainly on the

information provided by its own sensors

In Figure 10 a detail of results obtained in Genoa

is shown. In Genoa true container canyons rarely

appeared, but sometimes the influence of high stacks

of containers was evident, as in the case shown in

the figure. The movement of the Reach Stacker

starts at the point O, point considered as the origin of

coordinates. The vehicles moves forward to point A

(instant 12 s) and it is stopped at that point until

instant 74 s, when it then moves backward to point B

(instant 78 s). The Reach Stacker is at point B until

instant 106 s. Then it goes forward to point C

(instant 110 s) and after 17 s at that point the vehicle

moves backward to point D (instant 130 s). After 11

s it continues going backward to point E (instant 153

s), changes movement direction and moves forward

to point F (instant 188 s) and then it continues its

travel. As it can be deduced from the figure, the

Reach Stacker performed operations with containers

at points A and C. A container was picked up from a

slot at point A and then it was released in other slot

at point C.

The GNSS data covariance was very bad from

instant 6 s until instant 78 s and from instant 106 s

until instant 110 s (east covariance value higher than

3.5 and north covariance value higher than 9).

Therefore, from point A to point C the GNSS

estimates have poor quality, as it can be observed in

Figure 10, but DR estimates show very well the

movement carried out.

6 CONCLUSIONS

Analysis of a large number of experimental data

obtained in the course of the MOCONT-II project

has proven the success of the MOCONT Location

Module in tasks involving tracking the position of

containers in terminal yards while these are being

handled by the Reach Stackers, recording the slot

(row, column, tier), where the container is picked up

or released. Integration of the Differential Global

Figure 9: Example of experimental results at the TDT terminal in Livorno.

MOCONT LOCATION MODULE: A CONTAINER LOCATION SYSTEM BASED ON DR/DGNSS INTEGRATION

233

Navigation Satellite System (DGNSS) with a Dead

Reckoning (DR) inertial system has been

demonstrated to be effective in estimating the

position of the vehicle and of the container.

Positioning is effective even when working in

container canyons, on the condition that it be for a

limited movement time of some 40 s. Apart from the

general characteristics of the MOCONT system, the

structure of the MOCONT Location Module has

been presented as well as the data fusion diagram.

Likewise, typical experimental data and the final

evaluation of effectiveness are also presented.

ACKNOWLEDGEMENT

The material in this paper has been partially funded

by the European Union under the scope of the

Information Society Technologies programme

(research projects MOCONT: IST-1999-10057 and

MOCONT-II: IST-2001-34186).

REFERENCES

Aono, T., K. Fujii and S. Hatsumoto. 1999. Positioning of

a Vehicle con Undulating Ground Using GPS and

Internal Sensors. T. SICE, vol. 35, no. 8, pp.

1004/1001.

Benson, E.R., T.S. Stombaugh, N. Noguchi, J.D. Will and

J.F. Reid. 1998. An evaluation of a geomagnetic

direction sensor for vehicle guidance in precision

agriculture applications. ASAE paper 983203.

Landaluze, J., V. del Río, A. Milo, J. M. Ezkerra and A.

Martínez. 2003. Desarrollo de un sistema de

localización DR/DGPS para vehículos “off-road”. 5ª

Semana Geomática. Barcelona, 11-14 febrero.

MOCONT-II. 2004. Final Public Report.

www.mocont.org.

Ramjattan, A.N. and P.A. Cross. 1995. A Kalman Filter

Model for an Integrated Land Vehicle Navigation

System. Amy 1995, vol. 48 (no.2), pp. 293-302.

Rogers, R.M. 1999. Integrated DR/DGPS using Low Cost

Gyro and Speed Sensor. ION National Technical

Meeting.

Sukkarieh, S., E.M. Nebot and H.F. Durrant-Whyte. 1999.

A High Integrity IMU/GPS Navigation Loop for

Autonomous Land Vehicle Applications. IEEE

Transactions on Robotics and Automation, vol. 15, no.

Zhang, Q, J.F. Reid and N. Noguchi. 1999. Agricultural

Vehicle Navigation Using Multiple Guidance Sensors.

FSR'99 International Conference on Field and Service

Robotics.

Figure 10: Example of experimental results at the VTE terminal in Genoa.

ICINCO 2004 - ROBOTICS AND AUTOMATION

234