62 ◾ Vehicle Scheduling in Port Automation
research has been done and by whom. e second column shows decisions and
solutions for the problem. From this table, it can be clearly seen that most of the
container terminals considered their vehicle problems in the research.
3.7 Survey on Simulation, Implementation,
Solution Methods, and Evaluation
In the previous sections, we formulated the problems. ere are four important
phases to provide practical software for the decisions: (1) simulation and setting
the parameters, (2) selecting the design architecture, (3) the solution methods, and
(4)monitoring the performance. In this section, we review the rst phase and suggest
two frameworks for implementation in the second phase. After that, the solutions for
Table 3.6 (Continued) Container Terminals around the World and Their
Decisions
Ports
Decisions and Solution
Method Authors (Year) Ref. No
Hong Kong
Container
Terminal No. 9
(New)
• RTGC deployment in the
yard (Vogel’s solution)
• Storage space assignment
and vehicle routing (linear
programming)
• Storage space allocation
(integer programming)
• Crane/RTGC deployment in
the yard (MILP and
Lagrangian relaxation).
• Whole system (simulation)
Murty et al. (2005)
Zhang etal. (2001)
Zhang etal. (2002)
Liu, Jula, and Ioannou
(2002)
Real-Size
Terminal
Virtual Terminal
• Vehicle scheduling
(heuristic algorithm/
Lagrangian relaxation)
• Scheduling of QCs and
vehicles (distributed-agent
system)
Zhang etal. (2002)
Thurston and Hu (2002)
Carrascosa etal. (2001)
Real Port (not
mentioned)
• Whole system (multi-agent
system, not implemented)
Rebollo etal. (2000)
Port of Izmir,
Turkey
• Mathematical programming
(two dynamic strategies for
stacking containers)
Güvena and Türsel (2014)
Formulations of the Problems and Solutions ◾ 63
the problems are summarized and then some indices for evaluation and monitoring
of the result in each decision are presented.
3.7.1 Simulation and Setting the Parameters
e rst step is to set parameters in the formulations. For this phase, we propose to
provide a program to animate or simulate some operations in the terminal. e pro-
gram will be very useful to evaluate dierent values for the parameters and to generate
some input data for the next steps. In the problem specication, some operations or
decisions should be synchronized to each other, if two or more decisions are likely to
be studied together. For example, scheduling and routing of vehicles (the problem
in Section 3.4) can be combined with the storage space assignment (the problem in
Section 3.2). In the complex system, a few parameters should be considered in the
integrated model to synchronize the decisions. Tsang (1998) suggested some methods
to represent time and space. Bontempi et al. (1997), Seifert et al. (1998), Gambardella,
Rizzoli and Zaalon (1998), Hartmann (2002), as well as urston and Hu (2002),
applied some scenarios for simulation of terminal systems with several restrictions.
Additionally, Kim etal. (2000) introduced a simulation-based test-bed to test various
control rules. ey suggested a control system consists of ship operation manager,
system controllers for AGVs, automated yard crane, and QC. ree control strategies,
synchronization, postponement, and re-sequencing, were introduced in the work as
promising alternatives for controlling the trac of vehicles.
Rizzoli etal. (1999) presented a simulation model for terminal resource alloca-
tion policies and ship loading/unloading policies that are obtained by means of
operations research techniques. During the research, the simulation model for a
case study of La Spezia container terminal was designed, implemented, and vali-
dated. e results showed that the application of computer-generated management
policies could improve the performance of the terminal, making possible the allo-
cation of fewer resources, and to a better usage of the yard cranes.
Kim etal. (2000) presented a simulation-based test-bed to audit various control
rules and investigated it for Pusan Port. e research suggested a control system,
consisting of ship operation manager, system controllers for AGV, automated yard
crane, and QC. ree control strategies including synchronization, postponement,
and re-sequencing, were introduced in the paper as promising alternatives for con-
trolling tracs of vehicles in the port.
Duinkerken and Ottjes (2000) implemented a simulation model for an ACT
and applied their model to Delta Sealand container terminal of European container
terminal (ECT) Rotterdam. eir objectives were to determine the sensitivity con-
cerning a number of parameters like the number of AGVs, maximum AGV speed,
crane capacity, and stack capacity. e authors concluded that the most critical per-
formance indicators were the average number of moves per hour per QC, QC utiliza-
tion (percentage of time that the QCs are not waiting for AGVs) and the average trip
duration ratio (the ratio between the actual duration of a trip divided by the technical
64 ◾ Vehicle Scheduling in Port Automation
trip time-distance/speed) andaveraged over all connections between the stack and
quay. Some experimental results were presented in the work.
Carrascosa etal. (2001) designed an architecture of a distributed multi-agent
system for a real port. In the architecture, the research considered ve agent classes:
1. Ship agents: ey control the ships loading and unloading sequence schedul-
ing process.
2. Stevedore agents: ey manage the loading and unloading of all the ships
docking in the port.
3. Service agents: ey distribute the containers in the port.
4. Transtainer agents: ey optimize the use of the vehicles in the port.
5. Gate agents: ey handle the inbound and outbound containers to the land.
To solve the conguration problem, the goal of the service agents is to maximize
the stacking density in the yard, based on criteria such as time, stack allocation
conicts, and low stacking density. A preliminary version of the system is currently
being developed and implemented, which models the functions of the port.
Ioannou etal. (2001) presented and simulated an ACT (Ioannou etal. 2000,
2001, 2002). e ACT system is based on AGVs, grid rail (GR), and automated
storage/retrieval system (AS/RS). e main objective of the research was automa-
tion in improving terminal capacity and eciency in the context of the agile port
concept, in general, and to use of GR, in particular. Based on future projections
made by several ports, regarding container volume and the use of larger ships to be
served at terminals as fast as possible, the research came up with design character-
istics needs to meet the projected demand. During the research, a general layout
of the ACT was designed with considering the interfaces of the storage yard with
the ship, ITs and trains, as well as the desired storage capacity of the yard to meet
the projected demand. e layout was such that dierent concepts regarding the
storage yard and the container movements between the storage yard and the ship/
truck/train buers could be considered without major changes to the congura-
tion of the ACT. During the research, a cost model was developed and then the
simulation was performed to compare several competitive concepts that include
the GR, AS/RS, and AGVs. e authors assessed the performance of the model
by throughput (moves per hour per QC), throughput per acre, annual throughput
per acre in terms of the number of processed twenty foot equivalent units (TEUs)
per acre per year, ship turnaround time, truck turnaround time, gate utilization,
container dwell time, and the idle rate of equipment. e results indicated that the
proposed GR-ACT system is an attractive solution for places where land is limited
and expensive and high terminal productivity is required. Based on cost data found
in the open literature and a base scenario considered for a conventional terminal,
the average cost per container is about 60% less for the proposed GR-ACT system.
During the multiple autonomous robots for transport and handlingapplica-
tions (MARTHA) project, urston and Hu (2002) implementedamulti- and
Formulations of the Problems and Solutions ◾ 65
distributed-agent simulator for a virtual container terminal. e layout of the terminal
(space, roads, and junctions) was modeled in grids, comprising cells. e researchers
proposed four dierent types of agents in the system: (1) QC agents, (2) SC agents,
(3) trac agents (TAs), and (4) area manager agent. Each QC is controlled by a
QC agent, each SC is controlled by an SC agent, and each cell of the yard highway
that contains more than one entry point, such as a crossing, is governed by a traf-
c agent. e area manager agents represent physical resources of the system and
oversee the initial assignment of container jobs for any SC agents in the area it is in
charge of. e current standard for ne-grained route-scheduling was developed.
In this approach, which is known as the plan merging paradigm (PMP), each SC is
allowed to reserve resources. Each SC may reserve up to n cells ahead of its route.
In the project, the researchers adopt a strategy akin to the PMP. e research found
that if n is too small, an SC cannot reach its top speed, since it can only travel as
fast as it can safely break in time before a crossing, which it does not own (hence
not necessarily clear). Conversely, if n is too large, then the SC prohibits others from
using the resource unnecessarily. However, even if n is set to an optimum setting
for the SC’s capabilities, there is no notion of priority and thus an SC with a heav-
ily constrained deadline might have to wait for an SC with ample time to reach its
destination.
Liu, Jula, and Ioannou (2002) studied four ACTs, Port of Rotterdam, Port of
Hamburg, Port of Hong Kong, and Port of Singapore, and then evaluated their
operations by a simulation. e simulation used future demand scenarios to design
the characteristics of the terminals in terms of conguration, equipment, and
operations. e authors developed a microscopic simulation model and used it to
investigate several dierent terminal systems for the same operational scenario and
evaluate its performance. During the research, the authors evaluated ship turn-
around time, throughput of terminals, gate utilization, idled time of yard crane,
and buer cranes, dwelling times of containers and average cost of a container dur-
ing the simulation time. e research found AGVs to be the most eective in terms
of performance and cost. Additionally, the results indicated that automation could
substantially improve the performance of terminals.
Hartmann (2002) developed an approach for generating realistic scenario data
of port container terminals as input for simulation models, and for test of optimiza-
tion algorithms. In the scenario, the research considered data concerning arrivals of
ships, trains, and trucks within a time horizon and information about containers
being delivered or picked up. In the developed software, the user can control various
typical parameters. On the basis of statistics from a container terminal in the Port of
Hamburg, the simulation helped to improve use of the block capacities in the yard.
Saanen, Van Meel, and Verbraeck (2003) presented a simulation model to
account for cost values of dierent types of equipment to be installed at a termi-
nal (Saanen 2000; Saanen, Van Meel, and Verbraeck 2003). e simulation was
performed in a case study for the layout of terminals in Hamburg and Rotterdam.
eauthors compared productivity of equipment to handle container jobs when
66 ◾ Vehicle Scheduling in Port Automation
SCs could be used instead of AGVs. One of the major results was that at a certain
point adding further equipment could no longer increase the productivity.
e simulation also showed that if too many vehicles are blocking each other, the
productivity would have decreasing.
Valkengoed (2004) did a simulation to study how passing cranes inuence yard
operations in a container terminal (Saanen, Van Meel, and Verbraeck 2003). eauthors
assumed that in a high-density storage yard, each block is served by two RTGCs. Using
two cranes has the advantage that both sides of a block can be served at the same time
with a higher performance. But because they share the rails, they hinder each other
and each crane can only serve one particular side of the yard. During the research, two
dierent congurations (the yard operations with and without passing cranes) were
compared. e results found very little dierence in performance for the two dierent
congurations. e research also obtained results for an entire terminal where passing
cranes were implemented, with AGVs transporting the containers between the quayside
and yardside. Compared to the no passing cranes the research found the passing cranes,
when the gain in exibility is used, were able to give a higher quayside performance.
In order to ameliorate the productivity in a conned space of container termi-
nals, it is important to optimize the assignment of the existing distributed resources,
such as the cranes, storages, vehicles, and routs inside terminals. Zhang, Collart, and
Khaled (2013) study how to run container terminals in the maximum productivity
with minimum cost. In this work, given a P-time Petri net model of a small-size or
middle-size port with repetitive and periodic operation process, the author propose
a method to adjust the initial setting of system’s parameters to keep itself run with
maximum productivity and minimum cost. Moreover, the necessities for changing
the parameters of the resources are studied and a simple mathematical model to
evaluate the cost of change is also proposed at the end of this work.
Bohács, Kulcsár, and Gáspár (2013) provided an overview of container terminal
processes and related optimization models. In the overview, a survey of some simula-
tion-based model for intermodal systems is presented. Upon the drawn conclusions
from the surveyed works, a novel simulation model is described. e model enables
applicability of adaptive and intelligent methods and enables optimization of the cer-
tain modules and the whole model as well. Within the model, the authors propose
implementation of behavior-based control with several modules. e modules contin-
uously observe behavioral states of the other modules and change their own behavioral
state if necessary. ese methods are widely used at other areas of intelligent comput-
ing and each module may work in various operational modes. e proposed simula-
tion model was implemented in Simul8 logistic simulation software environment.
3.7.2 Selecting an Architecture
e second phase is to design architecture for the system. Two distinct systems
architecture including centralized system and distributed system have been suggested
by urston and Hu (2002); the latter was implemented by agents. For the rst
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.