IMPROVING THE PRE-NOTIFICATION PROTOCOL

OF THE CONTAINERS PICK-UP PROCEDURE

An Agent-based Approach

Meditya Wasesa

1,3

, Pim Nijdam

, Ismail Habib Muhammad

and Eric van Heck

Almende BV, Westerstraat 50, Rotterdam, The Netherlands

DEAL Services BV, Coolhaven 236, Rotterdam, The Netherlands

Rotterdam School of Management, Burgemeester Oudlaan 50, Rotterdam, The Netherlands

Keywords: Agent based simulation, Container Terminal.

Abstract: As the global container traffic flow is consequently increasing, the maritime container terminals (CTs) are

seeking new ways to improve their service. Taking the container import pick-up context as a test case, we

show that operational service can be improved by re-evaluating the information protocol. In this study, we

analyze and propose measures to improve the existing pre-notification approval protocol, the procedure that

bridges the CT and the incoming drayage trucks (DTs) in the container pick-up process. We put the

emphasis on the importance of considering the containers’ exact location in the pre-notification approval

request. The proposal is assessed by conducting agent based simulation experimentations. The results reveal

the opportunities to improve the CT’s performance in terms of the reshuffling frequency, the truck turn-

around time, and the average queue length at the CT’s gate-in, in favour of the proposal.

1 INTRODUCTION

The increasing container traffic has enhanced

business opportunities for all businesses that support

the supply chain, especially for the container

terminals (CT) community. Nowadays, the CTs are

expected to serve more transshipment requests

within a shorter service period. While most CT

operation problems are mitigated by focusing solely

on the physical flow analysis, this study’s main goal

is to show that operational performance can also be

increased by improving the information flow. As a

test case, we have analyzed the existing pre-

notification approval protocol for the container pick-

up process. We put the emphasis on the importance

of considering the containers’ exact location in

approving the pre-notification request sent by the

freight forwarders. By conducting agent based

simulation studies, we demonstrate that operational

performance (i.e. reshuffling frequency, truck turn-

around time, and truck queue length) can indeed be

improved by improving the information protocol.

To present the study, the paper is organized as

follows. In the next section, we discuss the related

literatures. Then, we explain the existing container

pick-up procedures. This is followed by problem

identification and the solutions proposal sections.

Then, we describe the agent based simulation setup,

present the simulation results and discussion, and list

some conclusions.

2 RELATED LITERATURE

The improvement of CT operation has received a lot

of attention from the research community. For many

years, this field has been dominated by operations

research (OR) studies (Stahlbock & Voss, 2008).

Although not as extensive as done by the mentioned

school of thought, the studies that apply Agent

Based Models (ABM) in tackling numerous CT

operations have also emerged (Davidsson et. al.,

2005). In the earliest phase, the ABM studies were

mainly focusing on the conceptual level. For

example, the work of Rebollo, et al. (2000) presents

an ABM architecture concept for solving automatic

transshipment problems for the operations of all four

of the CT’s sub systems: the marine side interface,

the transfer system, the container storage system,

and the landside interface. Another conceptual study

190

Wasesa M., Nijdam P., Habib Muhammad I. and van Heck E..

IMPROVING THE PRE-NOTIFICATION PROTOCOL OF THE CONTAINERS PICK-UP PROCEDURE - An Agent-based Approach.

DOI: 10.5220/0003742401900196

In Proceedings of the 4th International Conference on Agents and Artiﬁcial Intelligence (ICAART-2012), pages 190-196

ISBN: 978-989-8425-96-6

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

is presented by Thurston and Hu (2002). They put

forward an ABM architecture concept for handling

three different operations: the containers’ retrieval

from the stacking yard, the containers’ transportation

between the yard and the quay area, and the

containers’ transfer from the quay side to the vessel.

More recently, more validated CT-ABM studies

have emerged. These do not only propose ABM

architecture concepts for specific operation

application; validated studies normally also present

simulation works to validate their proposals

(Davidsson et. al., 2005). One good example is the

work of Henesey (2006). In his study, he presents a

thorough overview of the CT operations, proposes a

concept of ABM architecture, and elaborates on

several simulation studies focusing on numerous

operational issues such as: vessel arrival sequencing,

vessel berthing, inter-terminal transportation, etc.

Another recent work is carried out by Winikoff et.

al., (2011). They present an ABM open source

emulation platform (ContMAS) that is intended as a

decision support system for assessing different CT

policies. Moreover, in their study, they present a

negotiation-based module for container moves

allocation to the CT’s internal transporters and they

also propose a genetic-algorithm based solution for

allocating the containers in the yard.

We noted that most studies focus on the CT’s

marine interface operation (Henesey, 2006). The

landside interface operation is largely unexplored.

We have noted (only) one ABM study that analyzed

the interface operations between the CT and the land

carriers (Vidal & Huynh, 2010). Focusing on the

containers’ pick-up operation context, they have

analysed the impact of applying different yard crane

strategies to the drayage trucks’ service time. Instead

of solely focusing on the crane’s operational strategy

as was done in the study by Vidal and Huynh

(2010), we have extended their research by

concentrating more on the information protocol

improvement issue.

3 PROBLEM DESCRIPTION

The CT conducts two services: the marine interface

and the landside interface operations. The marine

interface operations are all activities that correlate

with the vessel service. The landside interface

operations deal with the service of land connecting

carriers (i.e. trains and trucks). In this study, we

focus on the CT’s landside interface operations.

More specifically, we analyze the container pick-up

process carried out by the drayage trucks (DTs).

The container pick-up process consists of two

different procedures: the pre-arrival and the on-

arrival procedures. In the following sections, we will

discuss each procedure in detail and we will discuss

the opportunity to improve operational performance

of the pick-up process by synchronizing both

procedures.

3.1 Container Pickup - Pre-arrival

Procedure

Prior to the container pick-up by the DT, the freight

forwarder has to finalize the pre-arrival formalities.

The pre-arrival procedure, which is actually a

request for pick-up confirmation, consists of several

information exchange formalities. The formalities

are regulated by the United Nations Committee for

Electronic Data Interchange for Administration,

Commerce, and Transport (UN/EDIFACT)’s

regulation. Figure 1 represents the pre-arrival

procedure.

Figure 1: Container pick-up pre-arrival procedure.

As portrayed, firstly, the freight forwarder sends

a container pick-up permission request to the CT in

the pre-notification message format, COPINO. A

COPINO message contains numerous details that

have to be completed by the freight forwarder. In

short, the COPINO confirms three important issues:

the identity of the container, the identity of the DT

that will pick up the container, and the proposed

pick-up date. Note that an incomplete COPINO form

may lead to request rejection.

Once the COPINO is received by the CT, several

checks will be carried out, including the evaluation

of the information details completeness, the presence

of the container in question at the CT, the customs

clearance procedures and some operational issues.

Once all the checks have been finalized, the CT

sends a reply message to the freight forwarder. The

reply is written in an APERAK (Application Error

and Acknowledgement) format. APERAK indicates

the approval/rejection status of the COPINO.

Upon receipt of the APERAK, if the COPINO

request has been approved, the freight forwarder can

IMPROVING THE PRE-NOTIFICATION PROTOCOL OF THE CONTAINERS PICK-UP PROCEDURE - An

Agent-based Approach

191

send an order to the DT for pick-up execution.

However, if the freight forwarder receives an

APERAK indicating COPINO rejection, a pick-up

order will not be sent. The freight forwarder might

review the reason for the rejection indicated in the

APERAK, carry out corrective actions, and re-

submit the COPINO for approval.

3.2 Container Pickup - On-arrival

Procedure

In this section, we discuss the pick-up on-arrival

operation. This procedure is only executed after

completion of the pre-arrival procedure. From the

CT’s perspective, the on-arrival procedure starts

when the DTs reach the gate-in area. As shown in

Figure 2a, especially at peak hours, the DTs

normally have to wait in a queue before receiving

service at the gate.

Whenever it is a DT’s turn, the gate-in officer

will check the DT’s documentation, ensuring that

the DT’s pick-up service request has already been

pre-registered in the CT’s EDI system (via the

COPINO-APERAK mechanism). After finalization

of the documentation check, the officer will give

permission for service, and will show the DT where

the container in question is located in the yard.

Figure 2: (a) Container pick-up operation overview

(b) A stacker crane servicing a truck.

Upon the DT’s arrival at the predefined location,

the DT will wait for the quay crane to come and

deliver the container. As shown in Figure 2b, once

the crane approaches the DT and stops at the same

row, the crane will then pick up the container from

the stack and place it on the truck’s trailer. After

delivery of the container, the truck will go to the

gate-out for the final administrative formalities

before departing from the CT.

3.3 Problem Indication

Figure 2b shows the ideal situation of a truck pick-

up service; it portrays a quay crane serving a

container that sits on top of other containers and is

right next to the DT. However, the service condition

is not always ideal. Many of the quay cranes have to

serve requests for containers that sit beneath other

containers, which results in extra work. This type of

service will occupy the quay cranes for a longer

period than the one requesting for the topmost

containers. Moreover, the request will also result in

negative consequences for the other DTs as they will

need to wait much longer in the queue. As a result,

the total DT’s turn-around time (i.e. the sum of the

waiting time and the service time) also increases.

3.4 Solution Proposal

To mitigate the inefficiencies, we propose a measure

to solve this issue. Our main concern lies with the

existing pre-arrival procedure (section 3.1). As

explained before, the approval of the pre-notification

(COPINO) request is determined by the evaluation

of two main aspects: the completeness of

information details and the presence of the container

in question. None of those aspects incorporate the

container’s exact location in the yard as an approval

criterion. Thus, we put forward a proposal to

incorporate the container’s exact location in the pre-

notification approval process. We propose that the

CT’s pick-up performance can be significantly

improved if the pre-notification approval gives

higher priority to the service requests for the

topmost containers.

Figure 3: The insertion of the synchronization module to

the existing COPINO approval system.

Responding to the stated supposition, we are

proposing to embed a synchronization (container’s

location checking) module in the existing pre-

notification approval system (see Figure 3). This

ICAART 2012 - International Conference on Agents and Artificial Intelligence

192

module will give priority to any pick-up request for

containers in the top layer. Since the introduction of

such a module may initiate several follow up

customizations in the existing EDI (electronic data

interchange) system that may cost a considerable

budget allocation, a preliminary feasibility study is

required.

4 SIMULATION SETUP

For the feasibility study, we have conducted ABM

simulations to assess the effectiveness of the

proposed synchronization module to the CT’s pick-

up key performance indicators/KPIs (i.e. reshuffling

frequency, truck turn-around time, and gate-in queue

length). We define the reshuffling frequency as the

fraction (percentage) of DT’s request services that

require reshuffle operations. The truck turn-around

time is defined as the total time spent by the DT in

the CT (i.e. the sum of the waiting time and the

service time). The gate in's queue length is defined

as the number of trucks waiting in front of the

entrance gate.

For executing the study, we have extended the

agent-based container terminal simulation platform

developed by Vidal and Huynh (2010). To align the

existing platform with our needs, we introduced

several customizations. First, in the referred

platform, 1000 containers are generated and

distributed randomly across four yard blocks at the

start of the simulation. In reality, the containers

come off the vessels continuously. Thus, we

continuously generate containers arrival during the

simulation run at 2 hour intervals (7200 ticks) to

maintain the containers’ buffer level at 1000.

Another customization concerns the crane utility

function definition. In the previous study, they

analyze the operation of the cranes that work

opportunistically based on three different utility

functions: distance based (OB-1), time based (OB-

2), and the combination of the two. Realizing the

importance of the reshuffling factor, we extended

the distance based utility by adding the extra

reshuffling consideration factor, so that the crane

will give priority to the truck that requires total

minimum crane occupation. We call this strategy the

processing-time-based utility function, OB-3 (see

equation 1).







=−







.



(

,

)

+



.



+



.









.







(

,

)









(1)

where:







= 









= ℎ.



(,)

= ℎℎℎ.





= ℎ.





= ℎ.





= ℎ.





= ℎ.





= .





= ℎℎℎ?





= ?





= ℎℎ?





= ℎ?

The final important customization is the

implementation of the pre-notification

synchronization concept. While in Vidal and Huynh

(2010), the incoming trucks request random

containers regardless of their position, in our study

we introduce the synchronization policy (i.e.

assigning the requests of the incoming trucks to the

topmost containers). The share of trucks that follow

the synchronization policy is controlled by the sync

platform utilization variable. Sync platform

utilization of 100% means that all incoming trucks

obey the proposed policy, thus only those requesting

the topmost containers will be served. On the other

hand, sync platform utilization of 0% reflects the

existing situation, where the pre-notification

approval is given regardless of the container’s

position in the yard.

Recall that our main interest is to analyze the

impact of applying the pre-notification

synchronization concept to the CT’s performance. In

response, we test the performance of the system at

three levels of sync platform utilization: 0%, 50%,

and 100%. To test the system’s performance under

different levels of occupation, we set different levels

for the CT’s workload by simulating 4 truck arrival

rate alternatives. In addition, we are also interested

in comparing the performance of different crane’s

working modes (i.e. OB-1, OB-2, and OB-3). In this

study, each crane agent will work opportunistically.

In total, we run 4*3*3 experiment set-ups (see

Table 1). Each experiment run lasts for 201,600

ticks (7 working days). We treat the first 57,600

ticks (2 days) as the warm-up period. In addition, we

employ 4 cranes, so that each crane will be

responsible for one container block area (see Figure

2a). Note that other than the customizations stated in

this paper, all simulation set-up details are similar to

the benchmark paper.

IMPROVING THE PRE-NOTIFICATION PROTOCOL OF THE CONTAINERS PICK-UP PROCEDURE - An

Agent-based Approach

193

Table 1: Simulation scenario.

Control

Variable

Setup Unit

Simulation

Length

201,600 Ticks

Warm-up

Period

28,800 Ticks

Sync

Platform

Utilization

0, 50, 100 %

Truck

Arrival

0.25, 0.50, 0.75, 1 Trucks/ Minute

Crane’s

Utility

Function

Distance Based (OB-1),

Time Based (OB-2),

Processing-Time Based

(OB-3)

5 SIMULATION RESULTS

Figure 4 indicates the impact of applying the pre-

notification synchronization module to the

reshuffling probability at different levels of

utilization and different experimentation conditions.

Without the application of the module (i.e. platform

utilization of 0%), the reshuffling probability per

container service is more than 40%. The ideal

situation with a sync platform utilization of 100%

will decrease the reshuffling probability to less than

1%. The mentioned situation is idealistic of course;

however, as the results indicate, by applying the

proposed module at only 50% use, we will have an

opportunity to decrease the reshuffling probability

between 23% and 30%, regardless of the CT’s

occupation level (i.e. truck arrival rate) and the

crane’s utility function.

Figure 4: The impact of synchronization module

utilization on the reshuffling frequency.

The application of the synchronization module

will also improve the CT’s performance in terms of

the truck’s turn-around time and the queue length in

front of the gate in area. As shown in Figure 5 and

Figure 6, the level of improvement increases as the

truck arrival rate goes up. This means that the

proposed platform offers bigger improvements when

the CT’s workload is high. As indicated also, the

improvement magnitude grows even higher when

the CT applies the OB-2 operational policy to the

yard cranes. Recall that the OB-2 is indeed the

existing crane strategy that is based on a first-in-

first-out policy. At an extremely busy day (i.e. truck

arrival = 1 truck per minute), if we apply the

location synchronization module to the existing

system (OB-2), we can improve the truck’s turn-

around time from 30 minutes to less than 12.5

minutes. The improvement in the truck turn-around

time is consequently accompanied with the queue

length reduction also (see Figure 6).

Figure 5: The impact of synchronization module

utilization on the truck’s turn-around time.

In comparison with the time based crane

operation (OB-2), the distance based (OB-1) and the

processing time based (OB-3), are more robust in

coping with the CT’s occupational intensity

fluctuation. For both utility functions, as the truck

arrival rate is increasing, the truck’s turn-around

time and the queue length is increasing moderately.

Although both non-time based utility functions are

already robust to the increase of CT’s workload, the

synchronization module application still brings

considerable improvements at any level of

occupancy. In general at each experimental

condition setup, higher module utilization indicates

better CT’s performance.

Platform Utilization

(

)

Trucks Turn Arround Time (Minutes)

TA = 0.25 tr/min

0 50 100

TA = 0.50 tr/min

050100

TA = 0.75 tr/min

0 50 100

TA = 1.00 tr/min

0 50 100

OB-1 OB-2 OB-3

ICAART 2012 - International Conference on Agents and Artificial Intelligence

194

Figure 6: The impact of synchronization module

utilization on the trucks’ queue length.

6 DISCUSSION

Although the application of the synchronization

module application is delivering better CT’s

performance at any experimental setup, the

implementation of the processing time based crane

strategy, OB-3, does not deliver considerable

improvement in comparison with the distance based

strategy, OB-1. This is contrary to our previous

conjecture. We expect that by applying a crane’s

strategy which considers the holistic processing

time, we will gain a better performance than

applying the crane’s strategy that omits the

importance of the reshuffling activity.

Triggered by this question, we re-evaluated the

simulation result data log and did an assessment to

the time fraction that is spent by the yard cranes for

conducting each activity (i.e. idle, travelling,

reshuffling, and servicing).

As we analyse the activities composition (Figure

7), we remark that the time spent in conducting the

reshuffling activities is surprisingly small. At high

occupation level (e.g. 1 truck per minute), the cranes

are allocating not more than 7.5 % of their time for

operationalizing reshuffling activities. The number

shrinks even lower when the truck arrival rate

plummets and the sync module utilization increases.

In addition, if we set the module utilization near to

100%, in principal there will be no (less) pick up

requests that require reshuffling activity. In this

situation, the introduction of the reshuffling factor

applied in the processing-time based strategy will

have diminishing impact in comparison with the

distance based crane strategy.

Figure 7: The overview of the yard cranes’ activity.

7 CONCLUSIONS

In this study we analyze and propose the application

of a synchronization module to improve the existing

pre-notification approval protocol. We do

preliminary study to assess the impact of applying a

synchronization module to the CT’s performance by

conducting agent based simulation. The results

reveal opportunities to improve numerous CT’s

performance by considering the container’s physical

location in the CT’s pre-notification protocol. In this

study we also introduce and assess the new

processing time based crane operation strategy that

behaves equally well with the distance based crane

strategy.

REFERENCES

Davidsson, P., et al., An Analysis of Agent based

Approaches to Transport Logistics. Transportation

Research Part C: Emerging Technologies, 2005.

13(4): p. 255-271.

Henesey, L. E., Multi Agent Systems for Container

Terminal Management, in Department of Systems and

Software Engineering. 2006, Blekinge Institute of

Technology: Blekinge. p. 1-275.

Rebollo, M., et al., A Multi Agent System for the

Automation of a Port Container Terminal.

Proceedings of Autonomous Agents: Workshop on

Agents in Industry, 2000: p. 1-6.

Stahlbock, R. and S. Voss, Operations Research at

Container Terminals: a Literature Update. OR

Spectrum, 2008. 30(1): p. 1-52.

IMPROVING THE PRE-NOTIFICATION PROTOCOL OF THE CONTAINERS PICK-UP PROCEDURE - An

Agent-based Approach

195

Thurston, T. and H. Hu. Distributed Agent Architecture

for Port Automation. in Proceedings of the 26th

Annual International Computer Software and

Applications Conference (COMPSAC '02). 2002.

Oxford, United Kingdom: IEEE Computer Society.

Vidal, J. M. and N. Huynh, Building Agent Based Model

of Seaport Container Terminals. Proceedings of 6th

Workshop on Agents in Traffic and Transportation,

2010: p. 1-10.

Winikoff, M., et al., Agent-Based Container Terminal

Optimisation. Proceedings of the Tenth International

Conference on Autonomous Agents and Multiagent

Systems, 2011: p. 1-9.

Wooldridge, M., Responding to Real-World Complexity:

Introduction to Multi-Agent Technology. A Magenta

Technology Whitepaper, 2007: p. 1-15.

ICAART 2012 - International Conference on Agents and Artificial Intelligence

196