Benchmark Problems for
J. Lee and B. Kim ¡°Industrial Ship Routing Problem with Split Delivery
and Two Types of Vessels,¡±
working paper, 2015
1.
Introduction
Lawrence (1972)
classified ocean transportation types into three groups according to operation
modes, namely, industrial, tramp, and liner. Industrial operation is a popular
ocean transportation mode used to carry bulk products. Cargo owners operate a
fleet of vessels in industrial shipping. All the given cargoes should be
served, and the objective is to minimize operating cost. In tramp shipping, the
objective is to maximize profit by selecting profitable cargoes for a given
fleet of ships. Tramp ships deliver available cargoes from point to point, much
like taxis. Tramp ships also carry spot cargoes, which are temporary customer
requests of transportation. The cargoes should be directly delivered from
pickup ports to delivery ports within specific times. Usually, a tramp ship
transports a cargo for a single customer at a time. Liners follow a given
itinerary and schedule similar to a bus line and usually transport cargoes for
many customers.
This paper
introduces an industrial SRP (ISRP) faced by a real-world steel manufacturing
company that delivers raw materials from supply ports to steelworks. The
subject company operates a fleet of company-chartered heterogeneous vessels.
Each cargo has a specific quantity, type, time window, and supply (pickup) and
demand (delivery) ports. The company planner generates the schedules of the
fleet to pick up and deliver the given cargoes. Some cargoes may not be served
by the fleet after a schedule made by the planner. Such cargoes are treated as
spot cargoes and are served by tramp ships. More tramp ships in the schedule
results in increased financial burden on the company because the delivery cost
of tramp ships is more expensive than that of a company-chartered fleet. In
addition, the cost of tramp ships is difficult to fix at the planning stage
because a delivery plan
is made 45 days before every quarter starts and the cost fluctuates frequently
following the global economy. Thus, the planner attempts to utilize the company-owned
fleet to serve cargoes as much as
possible (i.e., minimize the use of tramp ships). Cargoes can be split and
loaded by more than one ship if the time windows of the cargoes at pickup and
delivery ports are not violated. The cargoes can also be picked up from
multiple supply ports and delivered to multiple demand ports. In other words,
split delivery and combining cargoes are allowed.
ISRP can be
considered a variant of the pickup and delivery vehicle routing problem with
split delivery (PDSL). Compared with PDSL that considers VRP with split loads,
the proposed ISRP has additional considerations, namely, two types of vessels
are used to deliver cargoes and mixed loading is allowed. This paper presents a
mathematical model for the problem. For a solution procedure, an adaptive large
neighborhood search (ALNS) heuristic is developed. Computational results on
thirty benchmark problem instances show the effectiveness of the proposed
solution.
2.
Benchmark
Problems
To the best of our knowledge, no available
benchmark data for ISRP exist. Thus, 30 benchmark instances were developed
based on practical data used in the subject company. Three data sets were
created to reflect the practical data. Each data set, which contains 10
benchmark instances, was distinguished according to the number of vessels and cargoes.
The first set has five vessels and 10 cargoes; the second set has 10 vessels
and 50 cargoes; and the third set has 20 vessels and 100 cargoes. Table 1 shows
the summary of the data sets.
The cargoes and vessels of the benchmark problems were generated using
real-world data. The steel company provided a list of cargoes for this paper. A
total of 10 cargoes were randomly selected from the list for each problem
instance of data set 1, 50 for data set 2, and 100 for data set 3. All cargo
amounts were scaled down because of the confidentiality of the original data.
In a similar way, the vessels of the benchmark problem instances were
generated, and the vessel capacities were scaled down. Each vessel has ID, load_spd, emp_spd, capa, and unit_cost. ID denotes
the identification number of the vessel. load_spd
and emp_spd denote the speed of vessel when it has
cargoes on or not, respectively. capa
and unit cost denote the capacity of the vessel and the travel cost of the
vessel per an hour. The generated cargo weight ranged between 30 and 270 tons
for each cargo. About 100 cargoes are handled within a single quarter in
practice. The time windows of the cargoes and the position of ports were
generated based on real-world data. The distance between ports ranges from 1
mile to 20,167 miles in practical data.
Table 1. Benchmark data summary
|
Data set |
No. of vessels |
No. of cargoes |
|
1 |
5 |
10 |
|
2 |
10 |
50 |
|
3 |
20 |
100 |
The 30 benchmark data sets can be downloaded by
click here.
Required
data can be calculated or retrieved as follows:
Travel
time (days) from port i
to port j by vessel k (Tijk) = Euclidean
distance between the coordinates of the ports/(24¡¿the speed
of the vessel).
Travel cost
from port i
to port j by vessel k (Cijk) = (24¡¿Tijk¡¿unit_cost) + port
fee of j
The total amount of cargo at port i (Li) = Amt_load
Anchoring time at port i (Si) = anchoring_days
Maximum loading capacity of vessel k (Hk) = capa
Early and late time windows of cargo i at the pickup port(Ei, Fi) = (e_pick,
f_pick)
Early and late time windows of cargo i at the delivery port(Ei, Fi) = (e_del,
f_del)
Reference cost of tramp ships for the cargo i (Ri) = 10000.0