3.1. Research Approach
A quantitative research approach was employed to reach the final conclusion, based on three steps, which are briefly described as follows:
Development of scenarios based on the literature review for a possible solution to solve the loading/unloading problems based on different stacking strategies between the quay and yard side.
Development of agent-based modeling to analyze the different scenarios earlier developed.
Utilizing the Overall Equipment Effectiveness (OEE) index for the two cases to measure how well the operation is run compared to its ideal and full potential.
In order to develop the two cases, the container loading/unloading process based on the block-stacking strategy is considered. The two scenarios were developed based on evidence from the literature where three stacking strategies of block, bay, and slot assignment were discussed [
31]. Considering the existing stacking strategy used in a container terminal of the Port of Antwerp, the block stacking strategy was selected to investigate the determined scenarios. A schematic overview of the developed scenarios has been presented in
Figure 2 below.
Loading and unloading processes are shown by the loop in
Figure 2. In order to unload a container onto a stacking area, the container stacking system (CSS) determines whether there is a dedicated space for a container or not. Then, the unloading process begins, considering knowledge about the container stacking area’s status. However, in the loading process, containers’ locations in the stacking area should first be identified, then vessel, barge, train, and truck operators should get information about the loading process.
The first scenario shows a situation of block assignment rules. In the base case, it means no information about how the containers should be handled further. The second scenario examines a situation of block assignment rules in which where and how containers proceed further is known.
In this section, the Agent-Based Model (ABM) is explained, along with the developed Overall Equipment Effectiveness (OEE) index. Performance measurement is essential for all container terminals, especially in container terminal handling systems. A competitive advantage allows a container terminal to offer and sell services more attractive than those of its national and international competitors [
36]. In this sense, OEE, as one of the commonly used performance indicators, was adopted for determining the container terminal’s equipment utilization. Although this index initially appears linked to maintenance, it applies to broader activities to identify losses due to sustainability and establishes a complete understanding of the production process in terms of availability, performance, quality, and sustainability [
37]. The different developed scenarios are explained in
Section 3.4.
3.2. Agent-Based Model
Models and simulations are tools used to simplify existing complicated systems and allow the optimization procedure to be implemented before real-world model initialization [
38]. The ABM was first developed by Uri Wilensky [
39] to evaluate different scenarios using NetLogo. The ABM method has some advantages compared to traditional modeling approaches. Firstly, the ABM application has no agent selection limit, which basically provides the ability to evaluate each carrier across a set of variables. Secondly, ABM allows determining of the performance and the respective interactions using explicit modeling of behaviors and the interactions of each agent. Thirdly, in the ABM method, both agents and systems are able to memorize their actions in a dynamic modeling system [
40]. The ABM is particularly well suited for complex systems over a time period. This makes it possible to find out the micro-level and macro-level patterns that emerge from agents’ interactions [
39]. Different models and techniques have been conducted to examine container-stacking strategies. A multi-agent approach was applied in a study conducted by Rekik and Elkosantini (2019) to minimize limitations related to online stacking strategies, distributed control, and efficiency [
31]. The results of this study led to a system of container stacking with the ability to handle dangerous containers and decentralized control in an uncertain and disturbing environment. The inbound container volume, unloading, and stacking problems were evaluated using a two-stage search algorithm. Based on the formulation, an integer programming model is formed to decrease rehandling of containers and optimally distribute loading orders based on the stacking strategy [
41].
Although several models have investigated the container unloading/loading problems with respect to stacking strategies, the main reason for the container handling problem with respect to the further mode of transport is not well documented. Furthermore, previous studies did not consider the main cause of the container handling problem with respect to the stacking strategies in their model.
Therefore, in the present study, we aimed to integrate the main cause of container handling, considering different stacking strategies, with the effect of different scenarios of further information about transportation mode, inflow, and outflow of the container.
In order to examine this, agent-based modeling was considered a powerful tool to model these approaches and get conclusions from the interactions between system agents. Hence, the ABM can provide a unique model for investigating the impact of further transportation mode on the container handling process in a container terminal. Therefore, the model examined the effect of the container handling process based on different scenarios. The ABM can also exhibit complex behavior patterns in a container terminal and provide valuable information about the dynamics of the real-world system that it emulates. To manage this highly interconnected network, the intelligence of agents and the average knowledge of agents [
42] were taken into account. Therefore, any agents were coded to have a certain level of awareness using message communication ability when they initiate a command in the model. Moreover, in ABM, an observer can influence the dependence of knowledge spread (highly knowledgeable agents) within networks and the way agents select other agents for knowledge acquisition by conditioning the Knowledge Management Strategy [
43]. This way, we as observers can give commands to any agents to select communication with other agents after filtering for knowledge acquisition.
Accordingly, the structure of the knowledge dynamics network is determinative in improving innovation, and therefore a competitive advantage of an organization, meaning that agents with a higher knowledge level are effectively more knowledgeable than agents with a lower knowledge level [
44]. This assumption is reflected in scenarios defining the impact of having information about further transportation modes on agents and terminal performance.
The ABM consisted of three major components, which are known as agents, interactions, and environment. The agent can be classified as an independent entity with specific characteristics, while each agent can behave autonomously and has the ability to sense and communicate. It should be noted that agents may have complete or incomplete information about their surroundings, and they may have the ability to impact other environments. Accordingly, the ABM model can determine instructions to a hundred or even more agents in an environment where agents can interact with each other and impact each other based on the characteristics of the defined environment.
The main goals of the present study are to provide an insight into the container handling problem concerning stacking strategy and to examine different scenarios which could give a broader view of this process. Therefore, the comprehensive technical details of the container handling process are not taken into account in the current model. The developed model can be a foundation for future and more complicated models, including more sophisticated algorithms.
The main outline of the agent-based model is presented in
Figure 3. It represents a simplified interaction among different agents (terminal, sea vessels, barges, trains, and trucks).
Figure 3 presents the relationship between the agents (terminal, vessel, barge, train, and truck). The container inflow is the rate that new containers enter the system, which depends on the number of container vessels, barges, trains, and trucks, and the number of containers that are already in the system. The number of the stacking-area sliders sets the number of containers in a block-stacking, which in this case was set at 300 TEU. However, this parameter is calibrated in line with the different cases. Moreover, the container outflow is dependent on various factors, such as the number and efficiency of SCs, YCs, QCs, the number of containers already on the system, and terminal efficiency. To do a sensitivity analysis and to be in line with different cases, all parameters were calibrated.
As said before, a model only represents a part of the real situation. Hence, this model only deals with the container handling process for the various inflow transportation modes, namely vessel, barge, train, and truck. However, other effective factors with minor impacts on the results of the ABM, such as detailed technical measures, container stacking rules (weight limit, dangerous goods), and how QCs, YC, and SCs are allocated, were not taken into account in the model.
Based on the ABM, container inflow was determined by the number of containers set for different modes of transport controlled by the number-of-trains (TEU) slider, number-of-trucks (TEU) slider, number-of-barges (TEU) slider, and number-of-vessels (TEU) slider. In our model, a total number of 300 TEU containers was split among Block A, Block B, Block C, and Block D. The number-of-stacking-blocks sliders controlled the number of containers in each stacking block.
In the current study, the data for scenario analysis was generated based on the ABM’s output. The information was derived from a terminal operator in the Port of Antwerp. Consequently, the container terminal activity and related information with respect to the number of vessels, barges, trains, and trucks as the inflow container and the number of SCs, QCs, YCs as the outflow container were retrieved from this stakeholder. The container terminal in Port of Antwerp now contains a total of 41 quay cranes across 9 berths, with maximum depth at Chart Datum (m) 17,200 straddle carriers, and a quay length of 3700 m [
45].
Thus, we tried to model scenarios that could be a reflection of real-world conditions and contain valid assumptions in line with the operational practices of the container loading/unloading process inside a terminal in the Port of Antwerp. Moreover, the validity of the process was shown by a meaningful interaction among the agents corresponding to the real-world loading/unloading process.
With respect to the SCs’, QCs’, and YCs’ configuration, the assumptions were based on the publicly available information from the Port of Antwerp website. Configuration of SCs was as follows: 3- or 4-high stacking (one over two containers/one over three containers); lifting speed (full load), 18 m/min; lifting speed (empty), 26 m/min; driving speed (full load), 30 km/h; driving speed (empty), 30 km/h [
46]. QCs were set at 41, with outreach up to 25 containers wide [
47]. SCs at the terminal are responsible for both unloading/loading and stacking. Therefore, in our model, we presented different shapes named YCs but with the same characteristics as SCs. Deep-sea-going vessels are assumed to be an average of 370 m long and 55 m in width, with a capacity of 17,000 TEU, while barges’ average length was 110 m and average width was 11.4 m, with a capacity of 200 TEUs [
48].
Figure 4 shows the interface of the container terminal on NetLogo’s agent-based modeling software version 6.2 for different stacking areas and vehicles set for each of the situations.
Considering the early developed scenarios and based on the existing equipment and berthing capacity in a container terminal of the Port of Antwerp, we assumed that three deep-sea-going vessels and one barge could berth on the quay for the (un)loading process simultaneously. Moreover, trains and trucks in the landside are also part of the (un)loading process. Every block was allocated to a QCs group as well as SCs. For instance, the Block A container loading/unloading process is only done by a group of SCs-A and QCs-A with the same technical specifications. Terminal efficiency will vary considering different performance indexes of terminal equipment for the two possible scenarios.