Designing Indonesian Liner Shipping Network

As the largest archipelago nation in the world, Indonesia’s logistics system has not shown excellence according to the parameters of logistics performance index and based on logistics costs percentages from overall GDP. This is due to the imbalances of trading on the western and eastern regions in Indonesia, which impacts the transportation systems costs to and from the eastern regions. Therefore, it is imperative to improve the competitiveness of Indonesian maritime logistics through maritime logistics network design. This research will focus on three levels of decision making in logistics network design, which include type of ships in the strategic level, shipping routes in the tactical level, and container allocation in the operational level with implementing butterfly routes in Indonesia’s logistics networking problems. Furthermore, this research will analyze the impact of Pendulum Nusantara and Sea Toll routes against the company profits and percentages of containers shipped. This research will also foresee how demand uncertainties and multi-period planning should affect decision making in designing the Indonesian Liner Shipping Network.


Introduction
In 2011, the proportion of Indonesian logistics cost is equivalent to 24.66% of the Gross Domestic Product (GDP).Meanwhile, 11% of the logistics cost are logistic transportation costs (Arvis et al. [1], Guericke and Tierney [2], Meeuws et al. [3]).This shows the high cost of logistics in Indonesia.A survey conducted by the World Bank in 2014 stated that Indonesia was ranked 5th on the Logistic Performance Index (LPI) in ASEAN and ranked 53rd from 160 countries around the world.The survey was conducted to assess the performance of each logistics sector in each country by observing the six indicators: process efficiency, infrastructure quality, competitive shipping costs, competence and quality of logistics services, ability to track and trace goods, and travel time (Guericke and Tierney [2], van der Baan [4]).
Indonesia's economic growth in the second quarter of 2015 increased by 4.67%.Spatially, the structure of economic growth in Indonesia is dominated by a group of provinces in Java and Sumatra.Java provides the largest contribution to the Gross Domestic Product (GDP), which is equal to 58.35%, followed by Sumatra that contributes 22.31%, followed by 8.22% from Borneo (Central Bureau of Statistics, 2015).Year to year, Indonesia's economic growth tends to be dominated by Indonesia's western regions.Therefore, the trend of economic activity, the center of production, development, and infrastructure development that mostly occurs in western Indonesia causes a distribution imbalance for the flow of goods between western and eastern Indonesia.
1* Faculty of Engineering, Industrial Engineering Department, Universitas Indonesia, Kampus Baru Universitas Indonesia, Depok 16424.Indonesia.Email: armand.omar@ui.ac.id * Corresponding author These findings show increased costs of logistics and maritime transportation in Indonesia, resulting in commodity price differrences throughout various regions in Indonesia.With these conditions at hand, opportunities for development and improvement in the field of maritime logistics Indonesia becomes abundant (Agarwal and Ergun [5], Kap and Gunther [6], Cullinane and Khanna [7]).
The imbalance of the flow of goods, increased costs of logistics and transportation is a challenge for stakeholders in the logistics sector, especially the shipping company in the decision process.In designing the network maritime logistics, there are three (3) levels horizon of decision-making: strategic level, tactical level and operational level, where each level will solve the problem of the number and type of vessels, scheduling shipping routes, and the allocation of cargo.We use three types of routes in this research, which are back and forth, butterfly, and pendulum nusantara routes.Bank and forth route is liner shipping routes between 2 ports where ship depart from origin port to destination port and go back again to origin port.Butterfly route is liner shipping route that consist more than 2 ports.Butterfly routes ensure one port will be passed twice (Ronen [8], Meijer [9]).This paper reflects the on going research conducted in our laboratory/research group regarding Maritime Logistics topics (Wibowo et al. [10], Moeis and Goputra [11]), particularly Indonesian Liner Network Design (Rahmawan et al. [12], Komarudin et al. [13,14]).

Methods
There are three stages in this research that consists of data gathering, model developing, and shipping routes determination.Consequently, this research also inquires profit and operational costs if market demand is to be fulfilled.This research is limited for the main ports of Indonesia, which include Belawan, Tanjung Priok, Tanjung Perak, Makassar, Banjarmasin, and Sorong (van Rijn [15], Smith and Taskin [16]).

Mathematical Model
This research adopts a mathematical model that was developed by Mulder and Dekker [17].The objective function of this model is maximizing profit and revenue, while minimizing handling costs and transshipping costs.This model is then refined by Meijer [9] by adding fuel costs and fixed costs.Furthermore, this research refines the Meijer's model with adding more variables which are stochastic demand, multi period planning, butterfly routes, and 13 ports.Below is the base model that developed by Mulder and Dekker, refined by Meijer.

Constraints
Below are constrains in this linear programming formulation.
Ensure that the total cargo shipped from one port to another does not exceed the demand of that port combination.
Make sure that the total load of a ship between each two consecutive ports does not exceed the capacity of the ship. ℎ 1 ,ℎ 2 , ≤     (ℎ 1 , ℎ 2 , ) ∈  Ensure that the flow to a Transshipment port with destination port ℎ 2 ∈  has to equal the flow from that Transshipment port to port ℎ 2 .
Define the amount of flow between two consecutive ports.
Define the total flow between each two ports in the same cycle.

Multi Period Model
This model is our refinement/development from base model with adding set of periods.This model can incorporate demand forecasting as necessary input to make better decision in designing Indonesia liner shipping network.

Stochastic Demand Model
In this model, we added a module that reflects demand uncertainty between 2 ports that commonly happen in the real shipping operations where demand cannot precisely be determined (Grammenos [18]).

Model Validation
We validate this model by comparing it with Meijer's model using the same input.The output comparison between Meijer model and this research model is summarized in Tabel 1.

Back and Forth + Butterfly Shipping Routes Optimization
There are four shipping route combinations in this part.The first route combines butterfly routes with back and forth routes and the second routes incorporates the first routes with Pendulum Nusantara.
The optimization result of route combination between butterfly routes and back and forth routes using base model as mathematical model can be seen in Table 2 and Figure 1. Figure 1, shows back and forth routes, is the majority in Indonesia liner shipping routes.Demand from and to Sorong must be come from Makassar.It will cause delay for transporting containers from and to Sorong due to many transhipments.

Back and Forth, Butterfly Shipping Routes + Pendulum Nusantara Route Optimization
This research also foresees how the Pendulum Nusantara route will affect weekly profits and demand fulfillment percentages.Hence, the route combination in Figure 2 combines the route combinations in Figure 1 with the Pendulum Nusantara route.The result is showed in Table 2 and Table 3.In optimal condition, Pendulum nusantara will reduce the weekly profit because Pendulum Nusantara is a long route that contributes siginificant number of costs.However, it will increase the percentage of container shipped (97.20% to 99.68%).Based on the optimization model results of both scenario 1 and scenario 2, the obtained results describe the financial conditions and operations of the company as a substantive consideration in decision making for Indonesian maritime logistics network design.Figure 4 shows the illustration of liner shipping 13 ports without obligation to fulfill all the demand.As the result, there are 5 ports (Palembang, Palembang, Pontianak, Ambon and Bitung) that not served.Figure 5 obligates all the demand to be fulfilled, it will increase the percentage of container shipped but will reduce the profit gained.Figure 3 is a financial performance comparison of shipping companies in scenario 1 and scenario 2 (Rahmawan et al. [12]).From Figure 3, we can conclude that fulfilling all demand will increase the total revenue, however it will increase the cost as well, especially in fuel cost.As the result, the weekly profit will be slightly decreased.The result from this optimization shows fulfill all the demand will give deduction in profit by 3.6% compared with scenario 1.Therefore, not many shipping companies wishing to serve the entire container ports in Indonesia.

Optimization using Multi Period Planning
This research aims to obtain maximum profit for liner shipping companies by including multi period planning.Thus, the initial step is analyzing profit generated in each scenario and each condition.The scenarios are: (1) Fixed number of routes and ships, (2) Fixed number of routes, flexible number of ships, (3)Flexible number of routes and ships.The conditions are: (1) Demand will follow the forecast, (2) Demand will follow the forecast until year 5, then fixed until year 10, (3) Fixed demand.Based on the results of the model, we can see that the 3 rd scenario generates the highest profit, compared to the 1 st and 2 nd scenario, for each condition.Comparison of profit results is shown in the Figure 6 (Komarudin et al. [13]).3 rd scenario will generate highest profit because this scenario allows the company to adjust the routes and number of ship accordingly based on demand at that year compare to 1 st and 2 nd screnario.
Next step is to conduct an analysis of fleet allocation for each scenario on each condition.Table 5 shows the comparison of fleet size at the beginning of period.
At the beginning of the period, the 1 st scenario has the highest amount of fleet size and the 2 nd scenario has the least.The 1 st scenario has the highest amount of fleet size because the optimal fleet size must be determined at the beginning of period and will have the same amount on each period.Whereas the 2 nd scenario has the least amount because the fleet size is adjusted in accordance to demand on each period.Because the 2 nd scenario allows the fleet size to be fixed or increased as the time goes by, based on demand and costs.Comparison of fleet size at the end of period is shown in Table 6.

Optimization using Demand Uncertainties
The optimization result using back and forth routes with demand uncertainties (Komarudin et al. [14]), can be seen in Table 7 and Figure 7.
It is logical to say that the routes being used in stochastic model is better because it could tackle various conditions of container demand than the predetermined routes model.It gives a strong proof that using stochastic model from the beginning would be more beneficial for liner shipping company.
The different routes being used in predetermined routes and stochastic model is shown in Figure 7.
The difference lies on the route between Tanjung Perak and Banjarmasin.Stochastic model shows that it is better to use bigger ship on that route (Panamax 2400) compare to predetermined routes use Panamax 1750.This result further confirmed that it is more beneficial for liner shipping companies to use stochastic model from the beginning as it generates better routes.

Conclusion
There are several conclusions from this research: (1) multi period planning is important to determine optimal number of ships that needed in every years in-line with container demand forecasting, (2) butterfly and Pendulum Nusantara can be a good alternative to reduce the transportation cost and increase the percentage of containers shipped percentage as well, (3) one of the challenge in liner shipping is volatility.Stochastic model try to mimic volatility and come up with robust solution to integrate that, and (4) serving all of containers demand will give less profit than not serving all of the demand.It can become a national plan to decide logistic subsidy.

Figure 1 .
Figure 1.Back and forth+butterfly routes illustration

Figure 5 .
Figure 5. Illustration of containers flow with demand uncertainties Make sure that the total load of a ship between each two consecutive ports does not exceed the capacity of the ship. ℎ 1 ,ℎ 2 ,, ≤    , (ℎ 1 , ℎ 2 , ) ∈ ,  ∈  Ensure that the flow to a Transshipment port with destination port ℎ 2 ∈  has to equal the flow from that Transshipment port to port ℎ 2 .
Total cargo flow on ship route  ∈  between ports ℎ 1 ∈  andℎ 2 ∈  for scenario  ∈ .Make sure that the total load of a ship between each two consecutive ports does not exceed the capacity of the ship. ℎ 1 ,ℎ 2 ,, ≤     (ℎ 1 , ℎ 2 , ) ∈ ,  ∈  Ensure that the flow to a Transshipment port with destination port ℎ 2 ∈  has to equal the flow from that Transshipment port to port ℎ 2 .

Table 1 .
Output comparison between Meijer model and research model

Table 3 .
Weekly profit calculation for back and forth, butterfly + pendulum route

Table 5 .
Comparison of fleet size in the beginning of period

Table 6 .
Comparison of fleet size in the end of period

Table 7 .
Profit calculation with demand uncertainties between ports