12 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO.

1, MARCH 2002

Design, Simulation, and Evaluation of

Automated Container Terminals
Chin-I. Liu, Hossein Jula, and Petros A. Ioannou, Fellow, IEEE

Abstract—Due to the boom in world trade, port authorities are minimizes the ship time at port by choosing an appropriate
looking into ways of making existing facilities more efficient. One berth order. Reference [1] proposed solutions that range from
way to improve efficiency, increase capacity, and meet future de- new infrastructure inventions to improve and technology-assist
mand is to use advanced technologies and automation in order to
speed up terminal operations. In this paper, we design, analyze, operating procedures. Reference [15] addressed a concept of
and evaluate four different automated container terminal (ACT) integrated centers for the transshipment, storage, collection and
concepts. These concepts include automated container terminals distribution of goods. References [16] and [17] demonstrated
based on the use of automated guidance vehicles (AGVs), a linear using simulations that the throughput of the terminal could
motor conveyance system (LMCS), an overhead grid rail system be doubled if automated guidance vehicles (AGVs) are used.
(GR), and a high-rise automated storage and retrieval structure
(AS/RS). We use future demand scenarios to design the character- A wide range of researches for improving terminal operation
istics of each terminal in terms of configuration, equipment and efficiency has been investigated, but most of them concentrate
operations. A microscopic simulation model is developed and used on the logistic and/or operation strategies. Few of them issue
to simulate each terminal system for the same operational scenario the new design concept for the terminal to meet the future
and evaluate its performance. A cost model is used to evaluate the demand.
cost associated with each terminal concept. Our results indicate
that automation could improve the performance of conventional High-density, automated container terminals are potential
terminals substantially and at a much lower cost. Among the four candidates for improving the performance of container ter-
concepts considered the one based on automated guidance vehicles minals and meeting the challenges of the future in marine
is found to be the most effective in terms of performance and cost. transportation. Recent advances in electronics, sensors, infor-
Index Terms—Automated container terminal (ACT), control mation technologies and automation make the development
logic, cost model, simulation. of fully automated terminals technically feasible. This is
emphasized by the fact that the Port of Rotterdam is operating
a fully automated terminal using AGVs and automated yard
I. INTRODUCTION
cranes to handle containers whereas the Port of Singapore

W ORLDWIDE container trade is growing at a 9.5% an-

nual rate, and the U.S. rate is around 6%. It is antici-
pated that the growth in containerized trade continues as more
and Port of Hamburg [10] are experimenting with similar
ideas. Sea-Land at the Port of Hong Kong implemented a
grid rail (GR) system referred to as the GRAIL designed by
and more cargo are transferred from break-bulk to containers Sea-Land/August Design, Inc. This is a high-density manually
[1]. By 2010, it is expected that 90 percent of all liner freight operated terminal where the containers are served by shuttles
will be shipped in containers [2]. Every major port is expected moving on an overhead grid rail system.
to double and possibly triple its cargo by 2020. To handle this In this paper, we consider the design, simulation and evalua-
amount of freight and reduce the cost per “twenty-foot equiv- tion of four ACTs. Based on future projections made by several
alent unit” (TEU) slot, shipping companies are forced to order ports, regarding container volume and the use of larger ships to
faster, larger, and deeper ships. New massive container ships on be served at terminals as fast as possible, we came up with de-
one hand, and scarcity of the yard land on the other put an enor- sign characteristics an ACT needs to have in order to meet the
mous pressure on port authorities to find and deploy effective projected demand. A general layout of the ACT was developed
container handling systems in order to increase the throughput where the interfaces of the storage yard with the ship, inland
of the current container terminals. trucks, and trains as well as the desired storage capacity of the
The application of information technologies, optimization yard are specified in order to meet the projected demand. The
techniques and improvement of management are considered layout is such that different concepts regarding the storage yard
solutions that do not require too much investment of physical and the way containers are transferred between the storage yard
facilities [11]–[13]. Reference [14] proposed an algorithm that and the ship/truck/train buffers can be considered without major
changes to the configuration of the ACT.
Manuscript received April 2001; revised March 1, 2002. The Guest Editor for A model is developed that is used to simulate all the oper-
this paper was S. Washino. ations of the ACT down to the finest detail of the character-
C.-I. Liu was with the Electrical Engineering Department, University of
Southern California, Los Angeles, CA 90089 USA. He is now with Aplus Flash istics of each piece of equipment. The model is exercised for
Technology Inc., San Jose, CA 95131 USA (e-mail: [email protected]). each ACT system based on the same operational scenario, i.e.,
H. Jula and P. A. Ioannou are with the Electrical Engineering Department, based on the same incoming and outgoing traffic of containers
University of Southern California, Los Angeles, CA 90089 USA (e-mail:
[email protected]; [email protected]). at the interfaces. Performance criteria that include throughput in
Publisher Item Identifier S 1524-9050(02)04367-3. moves/h/quay crane, throughput per acre, ship turn-around time,
1524-9050/02$17.00 © 2002 IEEE
LIU et al.: DESIGN, SIMULATION, AND EVALUATION OF CONTAINER TERMINALS 13

arriving or departing by ships, trucks and trains in order to de-

sign the various components of the terminal and choose the ap-
propriate equipment.

A. Design Considerations
The post Panamax ships have capacities of 6000 TEUs, while
the largest ships today are 17 containers wide and capable of
over 8000 TEUs. It is important to note that ships with 20 con-
Fig. 1. General layout of automated container terminals. tainers wide could be accommodated by major ports to make
them viable in the near future [5]. A current service-window
expectation for mega-ships (over 6000 TEUs) is 48 h [1]. Ac-
and idle rate of equipment, etc., are used to evaluate each system
cording to the plan for the Port of Rotterdam, the North West
and make comparisons. A cost model is developed and used to
terminal will be able to accommodate container ships of 8000
calculate the average cost for moving a container through the
TEUs. It is expected that ten ships will arrive every week (85%
ACT. The performance and cost criteria are used to compare
loaded) to this terminal. If the maximum in port time is restricted
the pros and cons of each ACT system.
to 24 h, two berths for these ships with a capacity of 250 moves/h
The paper is organized as follows. Section II presents the gen-
will be required. This can be accomplished using five cranes per
eral layout of the proposed ACT, and calculates the number of
berth, each with a capacity of 50 moves/h [6]. Using similar pro-
equipment and desired characteristics necessary to meet a pro-
jections as for the Port of Rotterdam we come up with the fol-
jected volume of container traffic. Section III presents the cost
lowing design consideration for the proposed ACT systems.
and performance criteria and simulation and cost models that
Design Consideration 1: The ACT will serve ships capable
are used to evaluate different ACT systems. Sections IV–VII
of carrying 8000 TEUs. The ships will arrive every 24 h 85%
present the design, analysis and simulation of the proposed ACT
loaded and should be served in less than 24 h. In our design we
systems. In Section VIII, we compare and evaluate the proposed
assume a desired ship turnaround time of about 16 h.
ACT systems.
From the Port of Long Beach, U.S.A., approximately 15%
of the container traffic is carried directly via rail with no truck
II. ACT AND DESIGN CONSIDERATIONS movement involved. From the Port of Los Angeles, U.S.A., 55%
of containers are intermodal, and are destined for inland regions
The general layout of the ACT systems considered in this via rail. However, the port has estimated that approximately one
paper is shown in Fig. 1. The Fig. 1 shows the interfaces of the half of that number is first moved by truck to the rail yards. The
gate, train and quay crane buffers with the storage yard. In the Port of Los Angeles estimates that by 2020, up to 40% of in-
case of the AGV-ACT (see also Section IV), the storage yard is a termodal containers will be moved via on-dock rail, while 60%
collection of stacks separated by roads where the containers are will continue to be moved via trucks [7]. We use the projection
stacked and served by yard cranes. AGVs are used to transfer of the Port of Los Angeles to come up with the following design
containers within the terminal and the storage yard. In the case consideration for the proposed ACT systems.
of the LMCS-ACT the storage yard is the same as in the case Design Consideration 2: About 60% of the containers will
of the AGV-ACT system. The only difference is that shuttles arrive at the ACT by trucks and about 40% will arrive by rail.
driven on a linear motor conveyance system are used for the Reference [3] presents various container arrival patterns and
transport of containers. For the GR-ACT and AS/RS-ACT, the indicate the proportion of containers that arrive ( to )
container storage yard in Fig. 1 is replaced with a number of days before the cutoff time. Some ports advertise cutoff times
GR units and AS/RS modules. AGVs in these cases are used to for each ship, after which cargo for that ship is no longer re-
transfer the containers between the GR (AS/RS) buffers and the ceived in an effort to meet ship departure schedules. For ex-
gate, train and quay crane buffers. ample, for some ships, containers start trickling in 6 days before
The gate buffer is designed to interface between the manual the cutoff time with a maximum arrival rate the second day be-
operations (inland side) and the automated ones (internal fore the cutoff time. That is 5%, 5%, 10%, 20%, 40%, and 20%
terminal side). It provides a physical separation between the of container arrive during the six, five, four, three, two, and one
manual and automated operations for safety reasons and also day, respectively, before the cutoff time. According to data from
for efficiency. It helps reduce the turnaround time of trucks by the Port of Rotterdam, at the North West terminal the “time in
providing a temporary storage area for the export containers stack” (stay time) for import containers is limited to three days,
and the import containers waiting to be picked by trucks. The and for the export containers is limited to two days [6]. In our
train buffer is the area next to the train where loading and design we decided to adopt the data from the Port of Rotterdam
unloading between the AGVs and the train takes place. and use reference [6] to come with a design consideration for
The design of the ACT and its characteristics such as storage the arrival pattern of containers relative to the arrival time of the
capacity, number of gate lanes, number of berths, number of ship.
quay cranes, etc., are based on the expected maximum volume Design Consideration 3: The export container arrival pat-
of containers to be processed through the terminal as well as on tern relative to the ship they are bound to is 0.2, 0.5, and 0.3
the characteristics of the equipment. In the following subsec- meaning that 20% of containers arrive during the second day
tion, we consider the maximum expected volume of containers before the ship arrives, 50% arrive during the first day before
14 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO. 1, MARCH 2002

the ship arrives, and 30% arrive the same day and early enough is taken to be about 45% higher than the one dictated by the de-
to be loaded while the ship is at the berth. sign considerations 1–5, i.e., about 22 000 TEUs.
There is a tendency to keep the import containers in the
storage yard longer than the export ones. In general imports are C. Number of Berths and Quay Cranes
retrieved quickly during the first week after the ship arrival, The number of berths and quay cranes to meet the design con-
and then at a much slower rate. Castilho [8] claims that each siderations 1–5 depends on the speed of the quay cranes. The
ship carries different categories of containers that are retrieved maximum physical capacity of a quay crane is assumed to be 50
at different rates. Refrigerated cargo is often picked up imme- moves/h[6]. We assume that quay cranes can reach their max-
diately after it is discharged from the ship. It is also important imum capacity when they are operating in a single mode (i.e.,
to retrieve intermodal containers from the terminal quickly in either loading or unloading), while the average of 42 moves/h
order to keep a train schedule, while some domestic containers is assumed for double mode (i.e., combined loading and un-
bound for inland warehouses may be left at the terminal for loading). A 15% variance to the maximum capacity of the quay
a longer time to take advantage of the storage space available cranes is considered in our study due to the uncertainties in-
at the terminal and relieves space concerns at the destination volved in the quay crane operations. The number of quay cranes
warehouse. With these constraints and current practices in required to serve the ship with 3400 40-ft containers is given by
mind and the trend of using IT and improved scheduling and the relationship
dispatching techniques in the future, we adopted the Port of
Rotterdam numbers [6] and came up with the following design
consideration.
Design Consideration 4: The import containers are retrieved where denotes the ship turnaround time and denotes
during three days, with retrieval rates 0.5, 0.3, and 0.2 meaning the number of quay cranes. In design consideration one, we
50% of the containers are taken away by trucks and trains during assumed a desired ship turnaround time of about 16 h, which
the day the ship was served, 30% the second day and 20% during means that five quay cranes are required to meet the expected
the third day. Out of the 50% of the containers that are taken loading/unloading demand. Since five quay cranes in a single
away the same day, 30% are taken away directly without any berth can meet the demand, the number of berths can be kept as
intermediate storage and 20% are temporarily stored in the yard one.
before taken away.
In many of today’s ports, trucks operate in cycles of less than D. Number of Lanes at the Gate
24 h. There is a trend however to increase the time to close to The gate must be designed in such a manner as to provide the
24 h in order to meet the demand and avoid traffic delays in required number of lanes needed at peak. Since both truck ar-
the inland transportation system. This could be proven crucial rival and service time at the gate are random, we model the gate
in areas such as the Los Angeles Metropolitan area where high- operations as an queuing system, where , ,
ways and surface streets during peak areas are highly congested. and denote the mean arrival rate and mean service rate of
In our design we assume the following. the trucks and the number of lanes at the gate, respectively. The
Design Consideration 5: The trucks/trains of the ACT will mean service time of a truck at the inbound gates is assumed
operate in cycles of 24 h. three minutes and at the outbound gates, two minutes. The min-
The design considerations 1–5 are used in the following sub- imum number of lanes can be determined from the inequality:
sections for designing the characteristics of the ACT system. . It is assumed that 2040 export containers arrive by
The characteristics of the equipment that is specific to each ACT trucks per day. Some of these trucks pick up import containers,
concept will be developed when the particular ACT concept is and the rest leave the terminal without any load. In addition,
addressed in subsequent sections. empty trucks arrive at the gate to pick up import containers. We
have assumed that the number of empty trucks that arrive at the
B. Storage Capacity gate to pick up containers is equal to the number of trucks that
Given the design considerations 1–5 the storage capacity of arrive loaded and leave empty. We have assumed that 40% of
the terminal should be large enough to accommodate all the con- the incoming loaded trucks leave empty which corresponds to
tainers that are required to be stored. From consideration three 816 trucks. Therefore, the total number of trucks that are ex-
the average number of export containers that have to be stored pected to arrive at the gate for loading and/or unloading per day
in the terminal is about 6120 TEUs per day. From considera- is trucks/day.
tion four the average number of import containers to be stored By assuming a 24-h operation we find that the truck arrival
is about 9520 TEUs per day. This gives a total of 15 640 TEUs rate is equal to h h min. Then
required storage capacity per day. Therefore, a storage capacity for , we have , which implies
greater than 15 640 TEUs will meet the demand and operational that a minimum of six lanes is required in the inbound-gate in
requirements of the ACT as characterized by the design consid- order to meet the demand. The mean service time at the out-
erations 1–5. It is desirable however to have a storage capacity bound-gates is assumed to be two minutes which gives
higher than the 15 640 TEUs in order to meet emergencies such per min. The arrival rate at the outbound gate is equal to
as military deployment situations and others, have the flexibility h h min which is the same as the ar-
of putting an additional berth or even serving larger ships in the rival rate at the inbound gates. Since, ,
future. Given these considerations the desired storage capacity the minimum number of lanes in the outbound-gate required to
LIU et al.: DESIGN, SIMULATION, AND EVALUATION OF CONTAINER TERMINALS 15

meet the demand is equal to four. The number of six lanes, four TABLE I
lanes for the inbound and outbound gate, respectively, are the ARRIVAL RATES OF CONTAINERS
minimum possible. As the above inequalities are tight, the use
of six lanes, four at the gate will lead to a high utilization of the
gate during the assumed scenario. Small deviations from the as-
sumed arrival and departure rates may cause saturation at the
gates that lead to congestion on both sides of the gates. In order
to avoid such situations we increase the number lanes for the
inbound-gate to nine and for the outbound-gate to six.
TABLE II
E. Number of Yard Cranes at the Buffers NUMBER OF EXPORT CONTAINERS, BOUND FOR ONE SHIP,
We assume that the yard cranes used at the gate buffer have ARRIVED BY TRUCKS AND TRAINS
the following characteristics.
The yard crane’s speed is 5 mi/h. It takes 15 s to line up with
the stack, and an average time of 65 s to unload and load an
AGV. These characteristics give an average speed of about 36
moves/h/crane calculated by assuming
s/move where an average of 20 s are used for the lateral mo- TABLE III
CUMULATIVE NUMBERS OF EXPORT CONTAINERS,
tion of the crane along the stack. It is also assumed that these ARRIVING BY TRUCKS AND TRAINS EVERYDAY
cranes are gantry cranes of the same type used in the yard. They
are able to go over stacks of containers (up to four containers
high) and load and unload vehicles from both sides of the stack.
The number of containers handled by the yard cranes at the
buffer/day is calculated as follows:
number of containers (40 ft) that arrive by trucks 2040.
number of containers that arrive to the buffer from the yard
2040.
number of containers to be loaded on trucks that arrive empty TABLE IV
816. Therefore the maximum total number of containers to NUMBER OF IMPORT CONTAINERS, UNLOADED FROM ONE SHIP
be processed by the yard cranes at the buffer per day is AND RETRIEVED BY TRUCKS AND TRAINS

containers or containers/h.
This implies that the number of yard cranes needed to meet this
demand is equal to , i.e., six yard cranes will meet
the demand at the gate buffer. The use of six yard cranes gives
an expected average throughput at the buffer of
moves/h/yard crane.
TABLE V
The number of yard cranes to serve the train buffer is calcu- CUMULATIVE NUMBERS OF IMPORT CONTAINERS THAT ARE
lated similarly. The number of containers to be processed at the RETRIEVED BY TRUCKS AND TRAINS EVERYDAY
train buffer is 1360/day or containers/h. For
an assumed crane speed of 36 moves/h we have that
cranes are needed. Choosing two cranes for the train buffer,
the expected maximum demand is guaranteed to be met. In such
case, the expected average throughput at the buffer is
moves/h/crane. For the yard cranes, we assumed a variance
of 10% of the average speed in order to account for the random-
ness in the operation.

F. Operational Scenario
average cost for a container to go through the terminal is used
The operational scenario is based on the projected demand as the criterion for cost comparisons and analysis.
(design considerations) and will be used to evaluate different
ACT systems. It is summarized in Tables I–V. A. Performance Criteria
The performance criteria that are used in this study to evaluate
III. PERFORMANCE/COST CRITERIA AND MODELS and compare different ACT systems are summarized as follows:
The goal of every terminal is to perform efficiently and main- Throughput: Number of moves/h/quay crane.
tain competitiveness by providing low cost and high quality ser- Throughput Per Acre: Throughput/acre.
vices to customers. In this section, we present the performance Ship Turnaround Time: Time it takes for a ship to get
and cost criteria that are used to evaluate the ACT systems. The loaded/unloaded in hours.
16 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO. 1, MARCH 2002

Fig. 2. AGV-ACT layout.

Truck Turnaround Time: Average time it takes for a truck to IV. ACT USING AGVs
enter the gate, get served, and exit the gate, minus the actual Fig. 2, shows the basic configuration of the proposed
processing time at the gate. AGV-ACT system. In order to meet the desired storage
Gate Utilization: Percent of time the gate is serving the in- capacity of about 22 000 TEUs [6] the size and layout of
coming and outgoing container traffic. the storage is chosen according as follows: The storage yard
Container Dwell Time: Average time a container spends in consists of 36 stacks of containers and is divided into two
the container terminal before taken away from the terminal. sections. The import storage area where the import containers
Idle Rate of Equipment: Percent of time the equipment is idle. are stored and the export storage for export containers. Each
stack has 288 containers when containers are stacked four
high. It leads to the maximum capacity of the storage yard be
B. Cost Model 10 368 containers, i.e., 20 736 TEUs. In addition to the storage
yard, containers can also be stored at the gate buffer whose
The average cost per container (ACC) being processed
maximum storage capacity is 1728 TEUs giving a total storage
through a terminal is among the most important cost measures
capacity for the terminal of 22 464 TEUs, which is close to the
considered by port authority [9]. It provides a basis for eco- desired capacity, and the terminal dimensions are calculated to
nomic evaluation of container terminal operations. We adopted be ft (70.29 acres).
this measure to evaluate and compare the cost associated with Two types of roads are used in the proposed container ter-
each proposed ACT system. Costs associated with container minal: transit roads, and working roads. The transit roads are
handling and storage operations within a terminal can be denoted by dashed lines and the working roads by solid lines.
classified into three categories. No loading or unloading takes place along the transit roads as
— Cost of Locations: Cost of locations where activities these roads are used by AGVs to get to different points in the
(operations) take place, e.g., storage area, berth, etc. terminal. Loading and unloading take place along the working
— Cost of Equipment: Cost of yard equipment, e.g., yard roads. The vertical four-lane transit roads allow direct access
cranes, quay cranes, AGVs, etc. between the gate buffer and the berth in order to deliver con-
— Labor Costs. tainers between them without intermediate storage in the yard.
A similar access is provided in the rail side.
The ACC is equal to the sum of the total annual cost for ac- The terminal operates as follows: A truck arrives at the gate, it
tivities, equipment and labor divided by the total annual number checks in and moves along the gate buffer where it gets unloaded
of containers that are processed by the terminal [10]. The cost by a yard crane. The truck is either empty or it gets loaded again
model is simulated on an excel spreadsheet and is not included at the buffer before exiting gates. The yard crane at the gate
due to space limitations. The details of the cost model can be buffer loads the container directly to an AGV or if an AGV is
found in [10]. not available it stores the container at the buffer temporarily. An
LIU et al.: DESIGN, SIMULATION, AND EVALUATION OF CONTAINER TERMINALS 17

Fig. 3. Different tasks assigned to AGVs.

Fig. 4. Possible directions of AGVs reached at, and passing through an

export container loaded to an AGV at the gate buffer is either intersection.
transferred directly to a quay crane to be loaded on the ship, or
it is transferred to a particular stack to be unloaded by a yard
crane and stored in the yard. Similarly, an AGV loaded with an located between two adjacent nodes. To resolve this type
import container by a quay crane transfers the container to the of conflict, we use a “modified first come first pass” pro-
yard for storage or to the gate or the train buffer. tocol [10] as described in the next paragraph. Although
The main characteristics of the AGV-ACT system are the the protocol is complicated, it can resolve the conflict in
same as those of the general ACT described in Section II. What an efficient way by allowing many vehicles to pass the in-
is specific to the AGV-ACT system are the number of yard tersection without collision.
cranes needed in the storage yard and the number of AGVs to Fig. 4 shows all possible moves an AGV can make
perform the various tasks in order to meet the expected container when approaching and leaving an intersection. The
volume as described in Section II. Before we choose the number incoming directions toward the intersection are labeled
of equipment for the various tasks another design aspect of the east, west, south, and north based on the direction of
yard are the rules and logic that controls the motion of the AGVs arriving at the intersection. The outgoing directions are
in the yard [10]. The AGVs have to follow certain traffic rules labeled right, left, and straight, based on the direction
and protocols in order to avoid collision, possible deadlocks and of the turn the AGV would make when leaving the
congestion in the yard and complete tasks in an efficient way. intersection. For instance, (east, right) means that an
AGV is approaching the intersection from the east and
A. AGV Control Logic and Traffic Rules will make a right turn.
Table VI shows the directions of the possible conflicts
The transfer of containers between different transportation based on the timing of AGVs’ arrival at an intersection.
modes and storage area to be carried out by the AGVs in the For instance, if an (east, straight) AGV arrives first, then
AGV-ACT system can be divided into three tasks as shown in the (west, left) and (south, straight) AGVs have to stop
Fig. 3. until the first AGV finishes its maneuver, i.e., clears the
Task 1: Under this task the following subtasks are to be per- intersection. If two or more AGVs arrive at the intersec-
formed: 1) transfer of containers between the quay crane and tion at exactly the same time then the right of the way is
gate buffers; 2) transfer of containers between the quay crane given randomly.
buffers and the storage area; and 3) transfer of containers be- 2) Traveling along the same path with different speeds. An-
tween the quay crane and train buffers. other possible situation where collisions may occur is
Task 2: Under this task containers are transferred between the when AGVs are traveling along the same path with dif-
gate buffer and the storage area. ferent speeds. This situation is possible as loaded AGVs
Task 3: Under this task the containers are transferred between are assumed to have a lower speed than the ones that carry
the train buffer and the storage area. no load. To prevent this kind of collision, we use low
The terminal could be viewed as a network of intersections speed zone(s) in the portion(s) of the transit lanes where
with nodes where loading and unloading takes place. In our two or more AGVs with different traveling speeds may
design the AGVs are allowed to travel on the right lane of a exist [10]. When a particular AGV enters the low speed
two-lane road in their moving direction. Therefore, once the zone, it simply reduces its speed down to that of the loaded
pick-up and drop-off points are assigned to a particular AGV, AGV. For the container yard under consideration, the low
the path is uniquely determined by using the intermediate nodes. speed zone is the portion of the horizontal transit lane in
The control logic algorithm must be able to resolve any possible Fig. 2, which is located adjacent to the berth area.
conflict between AGVs. A conflict between two or more AGVs 3) An AGV stops ahead in the moving direction. The inter-
may occur during the following situations. vehicle spacing between the AGVs traveling in the same
1) Arriving at an intersection from different path segments direction in the same lane is chosen to be 45 ft so that if
at the same time. A segment is defined as a part of a road a particular AGV stops in order to perform a task or due
18 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO. 1, MARCH 2002

TABLE VI
FIRST ARRIVING AGV AND THE APPROACHING AGVS THAT NEED TO STOP TO AVOID COLLISION AT INTERSECTION

Fig. 5. Control logic of AGVs for task 1.

to an emergency the following AGVs have enough time B. Characteristics of Equipment

to stop without colliding with each other. The characteristics of equipment used by the AGV-ACT
The control logic that dictates the motion of the AGVs in system are considered to be the same as those described in Sec-
order to perform tasks 1–3 without collision, conflicts and dead- tion II for the general ACT layout. The additional equipment
locks for the proposed AGV-ACT systems is described by the specific to the AGV-ACT system is that associated with the
flowcharts shown in Figs. 5 and 6. storage yard and is discussed below.
LIU et al.: DESIGN, SIMULATION, AND EVALUATION OF CONTAINER TERMINALS 19

Fig. 6. Control logic of AGVs for tasks 2 and 3.

Yard Cranes for Import, Export Storage Yard: The yard AGVs serving the gate buffer and four AGVs serving the train
crane’s speed is assumed to be 5 mi/h. It takes 15 s to line up buffer.
with the stack, and an average time of 45 s to unload or load an As shown in Fig. 7(a), 48 AGVs are sufficient to meet the
AGV. We assume that one yard crane is used for each stack that maximum expected capacity of the quay cranes, which is 42
is a total of 36 yard cranes are used in the yard. The assumption moves /h/quay crane. Fig. 7(b) and (c) show the throughputs of
of one crane/stack is made mainly to simplify the control logic the cranes at the gate and train buffers. The throughput increases
of AGVs and cranes. as the number of AGVs increases. The number of AGVs for
Speed of AGVs: We assumed that an empty AGV travels with each task is calculated by choosing the combination with the
a speed of 10 mi/h while a loaded AGV travels with the speed minimum total number of AGVs that meet the expected max-
of 5 mi/h. These speeds are compatible with current AGVs used imum demand for tasks 1, 2, and 3. Considering that the max-
for the same application at the Port of Rotterdam. imum expected average throughput of the cranes at the gate
and train buffers is 34 and 28.3 moves/h per crane, it follows
Number of AGVs: The minimum number of AGVs that are
from Fig. 7 that the combination (48, 26, 6)—i.e., 48 AGVs for
required to meet the demand of the AGV-ACT system is de-
task 1, 26 AGVs for task 2, and 6 AGVs for task 3, a total of
termined by exercising the simulation model of the terminal 80 AGVs—will meet the demand for the AGV-ACT system.
for different combinations of AGVs. The objective is to have
a sufficient number of AGVs to feed the quay cranes fast C. Performance and Cost Analysis
enough so that the cranes operate close to their maximum ca- The characteristics of the AGV-ACT system are used as
pacity. This in turn will guarantee that the ship turnaround inputs to the simulation model together with the arrival/de-
time is minimized. We assume that the system is loaded, parture patterns of containers brought in and taken out by
i.e., there are always containers ready to be processed by the ships/trucks/trains as shown in Tables I–V. We assume that
AGVs at each buffer. While this scenario may not be true all the patterns of container arrivals and departures to/from the
the time, the system should have sufficient number of AGVs terminal by ship, trucks and train are repeated every 24 h
to deal with such possible extreme situation. The results of the so that a 24-h simulation was sufficient to make projections
simulations are presented in Fig. 7. about annual productivity. The results of a one-day (24-hour)
In Fig. 7, the number of AGVs for tasks 1, 2, and 3 satisfy the simulation are shown in Table VII.
ratio 6 : 3 : 1. For example, the simulation run that has 24 AGVs The ship turnaround time obtained from the simulations is
serving the quay crane buffer is the same simulation run for 12 16.81 h, which is close to desired one of 16 h. We should note
20 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO. 1, MARCH 2002

Fig. 7. (a) Throughput of quay crane. (b) Throughput of buffer crane. (c) Throughput of train crane versus the number of AGVs used.

TABLE VII
AGV-ACT: PERFORMANCE RESULTS FOR ONE-DAY (24-h) SIMULATION

that for a maximum speed of 42 moves/h/crane the best ship The idle rate of the yard cranes was found to be high. This is
turnaround time possible is 16.2 h. The difference between the due to two reasons. First, reshuffling has not been considered in
simulated and the best possible ship turnaround time is mainly our simulation. The second reason is the number of yard cranes
due to the variance introduced in simulations for the character- has not been optimized. Instead, one yard crane was assumed
istics of the quay cranes and other equipment. for each stack in order to simplify the operations and the control
It should be noted that the idle rate of the cranes is calcu- logic of the AGVs. A smaller number of yard cranes could be
lated over a period of 24 h. Since the ship was at the berth used to achieve the same throughput in case of the release of the
for only 16.81 h, it means that the quay cranes were idled for one yard crane per stack restriction. In addition, the simulation
h, which is 30.0% of the time that is close to results obtained together with the characteristics of the terminal
the 31.73% obtained from simulations indicating that while the are used to calculate the average cost of moving a container
ship was at the birth the quay cranes were operating very close through the terminal, i.e., the ACC value, by exercising the cost
to maximum capacity. Similarly, after the ship is serviced, the model for the AGV-ACT system. A result of ACC equal to $77.3
AGVs responsible for the task of serving the ship will be idle dollars is obtained [10].
until the next ship arrives about seven hours later. This accounts
for most of the 36.3% idle rate for the AGVs. The throughput V. AUTOMATED CONTAINER TERMINAL USING A
of the terminal is close to the maximum possible indicating that LINEAR MOTOR CONVEYANCE SYSTEM
the AGVs met the service demand imposed by the quay cranes’ Linear motor conveyance systems (LMCS) are among the
speeds. technologies recently been considered for cargo handling. A
LIU et al.: DESIGN, SIMULATION, AND EVALUATION OF CONTAINER TERMINALS 21

TABLE VIII
LMCS-ACT: PERFORMANCE RESULTS FOR ONE-DAY SIMULATION

prototype of a linear motor conveyance system has been con- in the cost, the ACC for this system is about $147. The differ-
structed and successfully tested in Eurokai Container Terminal, ence in cost comes mainly from the cost of installing a LMCS in
Hamburg [10]. LMCS have several attractive characteristics: the terminal, since the infrastructure cost of LMCS is high [10].
The motors are very reliable, and last a long time. A system such
as this could be ideally suited for port and terminal operations. VI. ACT USING A GRID RAIL (GR) SYSTEM

A. Terminal Layout The concept of loading and unloading containers in the yard
using overhead rail and shuttles is another attractive way of
The LMCS yard layout is identical to that of the AGV-ACT utilizing yard space more efficiently. It uses linear induction
system of Fig. 2 except that the paths are pre-built guide ways. motors, located on overhead shuttles that move along a mono-
For instance, a two-lane road in the AGV-ACT system becomes rail above the terminal. The containers are stacked beneath the
a two-guide way tracks that allow shuttles to travel in opposite monorail and can be accessed and brought to the ship as needed.
directions. The concept of the overhead grid rail (GR) system was used to
The AGVs are replaced with shuttles that are moving on the design, simulate and evaluate a GR-ACT system [2].
linear motors conveyance system. The shuttles can be consid-
ered as AGVs moving on a fixed path. Consequently the control A. GR-ACT: Terminal Layout
logic of the shuttles is similar to that of AGVs described in the
previous subsection and is not repeated. The GR-ACT system shown in Fig. 8 is similar to that of the
AGV-ACT system with the only difference that the storage yard
B. Characteristics of Equipment is replaced with 8 GR units. The use of several GR units instead
of a large one is done for robustness and reliability purposes
The characteristics of equipment used for the LMCS-ACT as well as for simplifying the operations as explained in [2].
system to meet the demand are the same as those of the general Eight units is chosen so that the storage capacity of the GR-ACT
ACT described in Section II. The characteristics and the number system is the same as that of the AGV-ACT and LMCS-ACT
of yard cranes are the same as in the AGV-ACT system. The systems. Due to the high density of the GR units, however, less
speed of empty shuttles and loaded shuttles are assumed to be land is needed to obtain the same storage capacity. As a result
the same as those in AGVS. We assumed that at each corner the total size of the terminal is ft (63.36 acres)
of the guide way, it takes five seconds for the shuttle to change versus 70.29 acres for the AGV-ACT and LMCS-ACT systems
its direction of movement. Despite this change the number of for the same storage capacity of about 22 000 TEUs.
shuttles needed to meet the demand was calculated to be the The 8 GR units communicate with the other parts of the yard
same as the number of AGVs used in the AGV-ACT system. through the GR gate/train (G/T) buffers: 1a, 2a, , 8a and the
GR quay buffers: 1b, 2b, , 8b. There are vertical transit roads
C. Performance and Cost Analysis between each two units. These transit roads are used for transfer-
A simulation model for the LMCS-ACT system is developed ring containers—using AGVs—to/from the gate buffer directly
and used to simulate the terminal based on the operation sce- to the berth area. The containers that have to stay in the yard
nario given in Section II. The results are shown in Table VIII. are stored in the GR units. The units number 1, 2 and 7, 8 are
Since the terminal yard layout, control logic of vehicles, speed used for storing import containers to be taken away by trucks
of the vehicles, and the characteristics of the yard equipment are and trains. The units 3, 4 and 5, 6 are used to store export con-
exactly the same for both AGV-ACT and LMCS-ACT systems, tainers brought in by trucks and trains. Note that in each unit
the performance of the two terminals is almost identical. The only one operation can take place at each time. For example the
difference is that AGVs are moving freely in the yard, while shuttles within GR unit 1 can serve either the buffer 1a or 1b but
LMCS shuttles are traveling on fixed guide paths. The differ- not both at the same time. The interaction of the GR unit buffers
ences between the AGV-ACT and the LMCS-ACT systems are with AGVs is as follows.
22 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO. 1, MARCH 2002

Fig. 8. The GR automated container terminal.

One AGV in one cycle goes from gate or train buffer with C. Characteristics of Equipment
an export container, unloads the container at the G/T buffer (ei-
ther 3a or 4a) and travels empty to the G/T buffer (either 1a or According to [2], the characteristics of the equipment associ-
2a) where it is loaded by an import container and travels back ated with the GR units are as follows.
loaded to the gate or train buffer. The AGVs at the rest four units Speed of Loading and Unloading the GR Buffers: It is as-
are operating as follows: When the ship is present, an AGV in sumed that it takes 30 s with a 10% variance to load or unload
one cycle goes from the quay buffer (either 5b or 6b) with an ex- a container to/from an AGV.
port container, unloads the container at the quay crane, loads an Number of Shuttles: The number of overhead shuttles in each
import container from the quay crane and travels to the GR quay GR unit is 15.
buffer (either 7b or 8b) where it unloads the container to the quay Speed of AGVs: The speed of AGVs serving the GR buffers
buffer and travels empty back to the GR buffers 5b or 6b. When and the quay cranes, gate and train buffers is 5 mi/h (loaded or
the ship is not present then the units 5, 6, 7, and 8 operate similar empty).
to the units 1, 2, 3, 4, i.e., one AGV in one cycle goes from the Number of AGVs: Simulations were used to calculate the min-
gate or train buffer with an export container, unloads the con- imum number of AGVs that are needed in order to meet the de-
tainer at the G/T buffer (either 5a or 6a) and travels empty to the mand in the GR-ACT system. In Fig. 9 the number of AGVs
G/T buffer (either 7a or 8a) where it is loaded by an import con- for tasks 1, 2, and 3 satisfy the ratio 6 : 3 : 1. The figure shows
tainer and travels back loaded to the gate or train buffer. that the combination (42, 21, 6)—i.e., 42 AGVs for task 1, 21
for task 2 and 6 for task 3, a total of 69 AGVs—can meet the
required demand for the GR-ACT system.
B. Control Logic of AGVs for the GR-ACT System
The tasks to be performed by the AGVs in the GR-ACT D. Performance and Cost Analysis
system are the same as tasks 1–3 given in Section III for the
AGV-ACT system. The only difference is that the GR units The characteristics of the GR-ACT system together with
are replacing the storage yard (see Fig. 3). The control logic those for each GR unit developed in [2] are fed into the
of the AGVs for the GR-ACT system is similar to that of the simulation model for the GR-ACT system and simulated for
AGV-ACT system. The difference is that the AGVs in the the operational scenario. By choosing an optimum number
GR-ACT system do not have to travel long distance inside the of shuttles and using a new dispatching algorithm to assign
storage area, (exception is the case of JIT loading/unloading containers to shuttles within the unit [2]. The results of the
operations) since they only have to serve the buffers of the simulation are shown in Table IX.
GR units. Because of that, the design of logic is simplified The simulation results indicate that the GR-ACT system per-
by assuming the same speed for loaded/unloaded AGVs. In forms efficiently by having the quay cranes operate close to
particular all traffic roads are designed to be low speed zones. maximum capacity and keeping the ship turnaround time close
LIU et al.: DESIGN, SIMULATION, AND EVALUATION OF CONTAINER TERMINALS 23

Fig. 9. (a) Throughput of quay crane. (b) Throughput of buffer crane. (c) Throughput of train crane versus number of AGVs used.

TABLE IX
GR-ACT: PERFORMANCE RESULTS FOR ONE-DAY (24-HOUR) SIMULATION

to the desired one. Similarly, the yard cranes at the train and gate in each AS/RS module served by a single SRM we have six
buffer worked close to maximum capacity. The idle rate of the rack structures that are built to store containers. The SRM is
quay cranes is over a 24-h period. This means that 31.38% of designed to move from one set of two racks to another within
the time the quay cranes were idle because the ship was not at the module. Each module has two buffers, one on each side.
the berth. The same goes for the AGVs dealing with task 1. The Each buffer has two slots, one for outgoing containers to be
ACC obtained for this system is $90.10 [10]. picked up by AGVs and one for incoming containers brought in
by AGVs. These buffers are referred to as pick-up and delivery
VII. ACT USING AUTOMATED STORAGE/RETRIEVAL (P/D) buffers.
SYSTEMS (AS/RS)
A. Terminal Layout
AS/RS with high-density storage capabilities could play
an important role in the future container terminal activities. In this concept, we replace the import and export container
It can be build on a small piece of land and add capacity storage area in the AGV-ACT system by AS/RS modules. As
by increasing the number of floors. The promise in the high shown in Fig. 10, the number of AS/RS modules is chosen so
productivity of the AS/RS lies in its capability to have access that the storage capacity is close to 22 000 TEUs. Assuming
to any container within the storage structure randomly, without that each rack can store 120 ( cells) containers and each
having to reshuffle containers. AS/RS module consists of six racks, the storage capacity re-
An AS/RS module has four major components: storage and quirement of 10 368 FEU’s can be achieved with 15 AS/RS
retrieval machine (SRM), rack structure, horizontal material modules. The total storage capacity of the AS/RS-ACT system
handling system, and planning and controls. The SRM simul- is equal to TEUs which together with
taneously moves horizontally and vertically to reach a certain the 1728 TEUs that could be stored at the gate buffer it gives a
location in the rack structure. The original design of the AS/RS total possible storage capacity of 23 328 TEUs and the dimen-
module consisted of only two racks served by an SRM [8]. It sion of AS/RS-ACT system ft (54.45 acres).
was found that one SRM for two racks was more than needed to The lanes adjacent to the gate buffer and P/D buffers and the
achieve a certain input/output throughput. In an effort to meet roads adjacent to the train/AGV interface are considered to be
demand and at the same time keep the cost low we modified the working roads, while all the other roads are transit roads. The
original design so that one SRM can serve six racks. Therefore, two transit roads located on both sides of the AS/RS structure
24 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO. 1, MARCH 2002

Fig. 10. Automated terminal yard layout using AS/RS.

Fig. 11. (a) Throughput of quay crane. (b) Throughput of buffer crane. (c) Throughput of train crane versus number of AGVs.

allow the direct transfer of containers that are not required to system. Similarly the tasks performed by the AGVs are the same
be stored in the yard. The containers that need to be stored (re- as indicated in Fig. 3.
trieved) in (from) the AS/RS structure are transferred by AGVs
from (to) quay crane, gate and train buffers. One AGV in one
C. Characteristics of Equipment
cycle carries an export container from the gate buffer to an
AS/RS module P/D buffer where it unloads the container and According to [4], the characteristics specific to the
gets loaded with an import container that it transfers back to the AS/RS-ACT system are the following.
gate buffer. Similarly, one AGV in one cycle goes from the berth Speed of loading/unloading at the P/D buffers: In [4] the op-
area to a specific P/D buffer (on the ship side) with an import erations within the AS/RS module were optimized so that at the
container, delivers it to the P/D buffer, gets loaded with an ex- P/D buffers an AGV can be served (load it and unload it) within
port container, which it transfers back to the berth area. 45 s with 10% variance.
Speed of AGVs: The speed of AGVs is 5 mi/h (loaded or
B. Control Logic for AGVs empty).
The control logic that dictates the motion of the AGVs within Number of AGVs: The AS/RS-ACT system was simulated
the AS/RS-ACT system is similar to the case of the GR-ACT with different combinations of AGVs performing tasks 1–3 in
LIU et al.: DESIGN, SIMULATION, AND EVALUATION OF CONTAINER TERMINALS 25

TABLE X
AS/RS-ACT: PERFORMANCE RESULTS FOR ONE-DAY

TABLE XI
PERFORMANCE AND COST RESULTS FOR DIFFERENT CONCEPTS

order to calculate the minimum number of AGVs that is nec- identical for all measures with the exception of the throughput
essary to keep the quay cranes operating close to maximum per acre. The highest throughput per acre was obtained for the
capacity. Fig. 11 shows that the combination (36, 14, 5)—i.e., AS/RS-ACT system since it requires less land to be imple-
36 AGVs for task 1, 14 for task 2, and 5 for task 3, a total of mented for the same storage capacity. Next comes the GR-ACT
55 AGVs—can meet the required demand for the AS/RS-ACT system that also requires less land for the same storage ca-
system. pacity. All the ACT systems operated close to the maximum
possible capacity of the quay cranes that was assumed to be 42
D. Performance Analysis moves/h/crane for combined loading/unloading. This is much
The characteristics of the AS/RS-ACT system are fed into higher than the average of about 28 moves/h measured in many
the simulation model, which was then exercised for the opera- of today’s conventional terminals.
tional scenario presented in Section II. The results of the simula- The significant difference between the various systems is the
tion are shown in Table X. The performance of the AS/RS-ACT average cost per container. The LMCS-ACT was found to be
system is comparable with that obtained with the other concepts. the most expensive due to the high infrastructure cost associ-
The throughput per acre, however, is higher due to the less land ated with the LMCS. The second most expensive system is the
required by the system. The ACC obtained for this system is AS/RS-ACT, due to the infrastructure cost of the AS/RS struc-
$102.24 [10]. ture. The AGV-ACT system was found to be the most cost ef-
fective followed by the GR-ACT. As the cost of land increases,
however, our model shows that after a certain land cost the
VIII. SUMMARY AND CONCLUSION AS/RS-ACT becomes more attractive.
The simulation results are summarized in Table XI. Since Our results demonstrate that automation could dramati-
the number of equipment and vehicles in each ACT system is cally increase throughput and reduce cost. For example the
chosen so that the ACT system can meet the same demand it is AGV-ACT system could increase capacity from the average
not surprising that the performance for each system is almost current values of 28 to 40 moves/h/quay crane and reduce the
26 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 3, NO. 1, MARCH 2002

ACC value from the range of $140–$200 in most of today’s [17] C. I. Liu, H. Jula, K. Vukadinovic, and P. A. Ioannou, “Comparing dif-
terminals to $77. ferent technologies for containers movement in marine container termi-
nals,” in Proc. 3rd IEEE Int. Conf. ITS, 2000, pp. 488–493.

ACKNOWLEDGMENT
The authors would like to thank Prof. K. Vukadinovic, Chin-I. Liu received the B.S. degree from
H. Pourmohammadi, Dr. A. Asef-Vaziri, P. Wright of Hanjin, Chung-Yuan Christian University, Chung-Li,
Taiwan, R.O.C., in 1989 and the M.S. degree from
P. Ford of Sea-Land Mearsk, Port of Long Beach, and Cleveland State University, Cleveland, OH, in 1994,
Capt. T. Lombard of American Presidents Lines (APL) for both in mechanical engineering. He received the
numerous discussions regarding terminal operations, terminal M.S. and Ph.D. degrees in electrical engineering
technologies and cost analysis. They would also like to thank from the University of Southern California, Los
Angeles, in 2001.
S. Wheatley of CCDoTT, B. Aird of MARAD and K. Seaman Previously, he was a Research Assistant at the Uni-
of USTRANSCOM for their inputs and constructive com- versity of Southern California, Los Angeles. He is
ments. Also, they would like to acknowledge the kind support currently with Aplus Flash Technology, Inc., as a Senior Design Engineer re-
sponsible for the design of the next generation of flash memory. His interests
of August Design, Inc. In particular, they would like to thank include VLSI systems, ASIC design, automated ports, and optimization.
E. Dougherty, the President of August Design, Inc., for sup-
plying them with data and useful information, commenting on
their work and making himself available for discussions and
Hossein Jula received the B.S. degree in electrical
meetings. engineering from Sharif University of Technology,
Tehran, Iran, and the M.S. degree in electrical engi-
neering from Tehran University, Tehran, Iran, in 1991
REFERENCES and 1994, respectively.
[1] N. K. Ryan, “The future of maritime facility designs and operations,” in Since 1996, he has been a Research Assistant at the
Proc. Simulation Conf., 1998, pp. 1223–1227. University of Southern California, Los Angeles, in
[2] P. A. Ioannou, E. B. Kosmatopoulos, K. Vukadinovic, C. I. Liu, H. Pour- the field of electrical engineering-systems with em-
mohammadi, and E. Dougherty, “Real time testing and verification of phasis on control and automation. His research inter-
loading and unloading algorithms using Grid Rail (GR),” Center for Ad- ests include automated highway systems, automated
vanced Transportation Technologies, Univ. of Southern California, Los ports, and optimization.
Angeles, Tech. Rep., Oct. 2000.
[3] M. Taleb-Ibrahimi, “Modeling and analysis of container handling in
ports,” Ph.D. dissertation, Univ. of California, Berkeley, 1989.
[4] A. Asef-Vaziri and B. Khoshnevis, “Automated storage and retrieval in Petros A. Ioannou (S’80–M’83–SM’89–F’94)
maritime container terminals,” Progress Mater. Handling, pp. 165–178, received the B.Sc. degree with First Class Honors
2000. from University College, London, U.K, in 1978 and
[5] L. W. Nye, “Planning intermodal capability at west coast container ter- the M.S. and Ph.D. degrees from the University of
minals,” in Proc. Ports, vol. 2, 1998, pp. 907–916. Illinois, Urbana, in 1980 and 1982, respectively.
[6] J. J. M. Evers and H. Boonstra, “Coaster express: An option for From 1979 to 1982, he was a Research Assistant
large-scale coastal container feedering,” in Proc. 3rd European Re- at the Coordinated Science Laboratory at the Uni-
search Roundtable Conf. Shortsea Shipping, F. A. J. Waals, Ed. Delft, versity of Illinois. In 1982, he joined the Department
The Netherlands: Delft University Press, 1996. of Electrical Engineering-Systems, University of
[7] Meyer, Mohaddes Associates, Inc., “Gateway cities trucking study,” Southern California, Los Angeles. In fall 1988, he
Gateway Cities Council of Governments Southeast Los Angeles was a Visiting Professor at the University of New-
County, Los Angeles, CA, 1996. castle, Newcastle, Australia, and in summer 1992, the Technical University of
[8] B. DeCastilho, “High-container terminals: Technical and economic Crete. He served as the Dean of the School of Pure and Applied Science at the
analysis of a new direct-transfer system,” Ph.D. dissertation, Univ. University of Cyprus, in 1995. Currently, he is a Professor in the Department
California, Berkeley, 1992. of Electrical Engineering-Systems, University of Southern California and the
[9] T. Toth, “Analysis of a simulated container port,” M. S. thesis, Univ. Director of the Center of Advanced Transportation Technologies. He was the
Delaware, Newark, 1999. author and coauthor of five books and over 150 research papers in the area
[10] P. A. Ioannou, H. Jula, C. I. Liu, K. Vukadinovic, and H. Pourmoham- of controls, neural networks, nonlinear dynamical systems and intelligent
madi, “Advanced material handling: Automated guided vehicles in transportation systems. His research interests are in the areas of adaptive
agile ports,” Center for Advanced Transportation Technologies, Univ. control, neural networks, nonlinear systems, vehicle dynamics and control,
Southern California, Los Angeles, Tech. Rep., Oct. 2000. intelligent transportation systems, and marine transportation.
[11] T. B. Wan, E. L. C. Wah, and L. C. Meng, “The use of IT by the Dr. Ioannou held a Commonwealth Scholarship from the Association
Port of Singapore Authority,” World Development, vol. 20, no. 12, pp. of Commonwealth Universities, London, U.K., from 1975 to 1978. He was
1785–1795, 1992. awarded several prizes, including the Goldsmid Prize and the A. P. Head
[12] K. V. Ramani, “An interactive simulation model for the logistics plan- Prize from University College. In 1984, he was a recipient of the Outstanding
ning of container operations in seaports,” Simulation, vol. 66, no. 5, pp. Transactions Paper Award for his paper, “An Asymptotic Error Analysis
291–300, 1996. of Identifiers and Adaptive Observers in the Presence of Parasitics,” which
[13] Y. Merkuryev et al., “A modeling and simulation methodology for man- appeared in the IEEE TRANSACTIONS ON AUTOMATIC CONTROL, in August
aging the Riga Harbor container terminal,” Simulation, vol. 71, no. 2, 1982. He is also the recipient of a 1985 Presidential Young Investigator Award
pp. 84–95, 1998. for his research in Adaptive Control. He has been an Associate Editor for the
[14] I. Akio, N. Ken’Ichiro, and W. T. Chan, “Efficient planning of berth IEEE TRANSACTIONS ON AUTOMATIC CONTROL and the International Journal
allocation for container terminals in Asia,” J. Advanced Transport., vol. of Control and Automatica. Currently, he is an Associate Editor of the IEEE
31, no. 1, pp. 75–94, 1997. TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS and an Associate
[15] J. W. Konings, “Integrated centers for the transshipment, storage, collec- Editor at Large of the IEEE TRANSACTIONS ON AUTOMATIC CONTROL. He
tion and distribution,” Transport Policy, vol. 3, no. 12, pp. 3–11, 1996. is a member of the Control System Society’s Council Committee on IEEE
[16] C. I. Liu, H. Jula, and P. A. Ioannou, “Automated guided vehicle system Intelligent Transportation Systems and Vice-Chairman of the International
for two container yard layouts,” Transport Res. Part-C, submitted for Federation on Automatic Control (IFAC) Technical Committee on Transporta-
publication. tion Systems.