EWGLA 2013 Marielle Christiansen

Optimization of
Maritime Transportation
Marielle Christiansen
Department of Industrial Economics and Technology Management
Norwegian University of Science and Technology,
Trondheim, Norway
EURO Working Group on Locational Analysis (EWGLA),

Ankara, Turkey, 17-19 April 2013
PDF compression, OCR, web optimization using a watermarked evaluation copy of CVISION PDFCompressor
Relevant references
Other relevant references (review
articles):
M. Christiansen, K. Fagerholt, B. Nygreen and

D. Ronen (2013). Ship routing and scheduling
in the New Millennium. EJOR, 228, 467-483
M. Christiansen, K. Fagerholt, and D. Ronen

(2004). Ship routing and scheduling: Status
and perspectives. Transportation Science ,
38(1), 1-18
M. Christiansen, K. Fagerholt, B. Nygreen and

D. Ronen (2007). Maritime transportation.
Handbooks in Operations Research and
Management Science, Vol. 14: Transportation,
C. Barnhart and G. Laporte (eds.), 189-284
Outline
• Introduction to maritime transportation
• Maritime cargo routing
• Maritime inventory routing
• Trends relevant for ship routing and scheduling
• Links to location analysis
• Concluding remarks
Cargo ships
• Tankers
– Carry liquids in bulk
• Bulk carriers
– Carry dry bulk commodities
• Reefers
• Liquefied gas carriers
• Ro-ro ships
• Container ships
• General cargo vessels
• Etc…
Photo: Gary Blakeley
Cargo ships - ports
• Ships operate between ports
– Loading (pickup) and unloading
(delivery)
– Supply of fuel, fresh water, etc…
• Physical limitations impose compatibility
constraints between ships and ports
– Draft, length, etc…
• Ships pay port fees Port operations are not
• Transhipment of containers focused here
• Storages placed near ports
• Limited number of berths, cranes or
pipes
Crainic, Kim (2007). Intermodal transportation. Chapter in Handbooks in
Operations Research and Management Science, Vol. 14:
Transportation, C. Barnhart and G. Laporte (eds.),
Stahlbock, Voss (2008) Operations Research at container terminals: a

literature update. OR Spectrum.
Maritime transportation - key figures
• Major mode of international trade

– Over 80% of international trade in goods is carried by sea (UNCTAD 2009)
• Monopoly on transportation of large volumes between

continents (deep sea shipping)
• Major role in domestic trades (short
sea shipping) for countries with
– Long shoreline (e.g. Norway)
– Many rivers (e.g. USA)
– Many islands (e.g. Greece)
World seaborne trade
Increase last decade: 40 %
World fleet
Increase last decade: Oil tankers: 60 % , Dry bulk ships: 65%, Container ships: 164%
In 2010, oil tankers and dry bulk ships consituted 75 % of the world fleet.
Basic modes of operation (Lawrence ’72)
• Industrial shipping
– Shipper(cargo owner) controls the fleet of vessels (owned or on TC)
– Must ship the total demand while minimizing costs Mostly
– Decisions: Routing and scheduling bulk
– Vertically integrated companies shipping
• Tramp shipping
– Combination of contract and optional spot cargoes
– Ships follow the available cargoes, similar to a taxi service
– Decisions: Routing/scheduling and selection of spot cargoes
– Maximize profit
• Liner shipping
– Ships follow a published schedule, similar to a bus line
– Container, ro-ro and general cargo vessels
Importance of proper ship routing
• Ships involve major investments and high operating costs:
Example: A VLCC, 200.000 dwt:

• Building cost: 100 mill. USD
• Daily TC rate: 70 000 USD
• Daily fuel costs: 50 000 USD
• Port fee: 100 000 USD
• Value of one cargo: 100 mill. USD
(figures only roughly estimated)
• Good planning is crucial:

– Reducing costs
– Increasing revenue by carrying spot cargoes
– Giving important positive environmental effects
• However, relatively little OR-work has been done in ship

routing and scheduling
Growth in number of publications
• Ronen, EJOR (1983)

– 3 decades, 40 references
• Ronen, EJOR (1993)

– 1 decade, 40 references
• Christiansen, Fagerholt and Ronen, TS (2004)

• Christiansen, Fagerholt, Nygreen and Ronen, EJOR (2013)

Planning levels
• Strategic planning level (1-20 years)
– Market and trade selection (Contract evaluation)
– Fleet size and mix decisions
– Network design in liner shipping
– Port/terminal location, size and design
• Tactical planning level (1 week – 1 year)

– Ship scheduling for industrial and tramp shipping
– Port management (inventory management, berth scheduling,…)
– Fleet deployment in liner shipping
– Empty container distribution
• Operational planning (1 day – 1 week)

– Environmental routing (Weather routing and ocean currents)
– Speed selection for a sailing leg
– Ship loading (Container stowage, Crane scheduling)
Outline
Maritime cargo routing
• Routing and scheduling of bulk cargoes
• Tactical planning level with planning horizon from 1
week to 1 year
• Focus on tramp shipping
• A cargo routing problem in industrial shipping can be

obtained by including mandatory cargoes only
• Models and solution methods
• Extensions and variants
Tramp shipping: Planning process
• Mix of mandatory contract cargoes and optional spot

cargoes
• Several daily requests for spot cargoes
• Negotiate spot cargoes and schedule the fleet
– Continuous and interwoven process
• Objective: Maximize profit

• Core business for many shipping companies
(spot)
Traditional planning methods
• Using a voyage estimation software

– Calculates only one voyage for one vessel simultaneously
• Few optimization-based DSSs in use
• Building complete schedules is done manually
• Graphical problem representation
• Spreadsheet for visualization
• Improvement potential is significant
Tramp Cargo Routing
(spot)
Full shiploads
Planning horizon Multiple cargoes
Pickup
Delivery
Time window
• No natural depot • Cargoes with specified pickup

• Heterogeneous fleet (speed, and delivery ports
capacity, cost, etc…) • Both contracted and optional
spot cargoes
• Time windows
Basic tramp cargo routing problem
• Objective: Maximize profit (gross margin)
– Revenue from transported cargoes – variable sailing costs (fuel
costs, port costs, canal costs, etc.)
• Constraints
– All mandatory contract cargoes must be transported once
– Optional spot cargoes can be transported at most once
– Routing constraints
– Time constraints
– Ship-port compatibility constraints
– Ship capacity constraints
• Maritime Pickup and Delivery Problem with Time

Window (PDPTW)
Tramp shipping: A path flow model
max  Pi   Pi yi    Cvr xvr N C - Set of contracted cargoes, i

iN iN vV rRv
N O - Set of optional spot cargoes,i
C O
V - Set of ships, v
Rv - Set of routes for ship v, r
  Aivr xvr  1 i  N C
vV rRv
Pi - Revenue for transporting
cargo i
 i  N O
Aivr xvr  yi  0 Cvr - Cost of sailing route r
vV rRv
with ship v
x vr 1 v  V Aivr = 1 if route r for ship v
services cargo i
rRv
x vr  0,1 v  V , r  Rv xvr = 1 if ship v sails route r

yi = 1 if spot cargo i is
transported
Solution methods
• Existing research:
– Column generation based on path-flow model (a priori1) or delayed2))
– Multi-start local search heuristic3), Unified tabu search heuristic4)
– Variable neighborhood search heuristic5)
• Both 3) and 4)
are implemented
as solvers in the
DSS TurboRouter
1) Fagerholt and Christiansen (2000). A combined ship scheduling and allocation problem. JORS
2) Stålhane, Andersson, Christiansen, Cordeau, Desaulniers (2012). A branch-price-and-cut method for a ship routing
and scheduling problem with split loads. COR
3) Brønmo, Christiansen, Fagerholt and Nygreen (2007) A multi-start local search heuristic for ship routing and
scheduling – a computational study. COR
4) Korsvik, Fagerholt and Laporte (2010) A tabu search heuristic for ship routing and scheduling. JORS
5) Malliappi, Bennell, Potts (2011). A variable neighborhood search heuristic for tramp ship scheduling.
Computational Logistics
Planning flexibility
However, in real life the planner often has more flexibility

than modelled in the basic tramp cargo routing
problem...
... thus the focus has recently been on model extensions

that reflect common situations in tramp cargo routing
Maritime PDPTW with Split Loads (SL)
• Split loads - More than one ship can transport one cargo
• Splitting loads may
– increase port costs
– reduce sailing costs
– give business opportunities (adding more spot cargoes to the
solution)
• However, splitting loads interact with the routing decisions
and affect several ships
• More interesting in deep-sea shipping than short-sea
shipping. Why?
– Sailing costs dominate port costs in deep-sea shipping
Extension: Example PDPTWSL
35000
30000
50000 60000
35000
30000
(spot)
15000
50000 60000
35000 (too late)
30000
(spot)
15000
50000 60000
35000
30000
(spot)
15000
50000 60000
Models and solution methods
• A priori column generation1

– Cargo splitting is decided in master problem.
• Branch-and-price method2
– Columns for all routes with optimal quantities decided in the master
problem
– Columns including routes and delivery patterns (best)
• Large neighbourhood search heuristic3
• Testing shows huge effect on economic result by utilizing
split loads
1 Andersson, Christiansen and Fagerholt (2011) The maritime pickup and delivery problem with
time windows and split loads. INFOR
2 Stålhane, Andersson, Christiansen, Cordeau, Desaulniers (2012) Branch-price-and-cut for a
maritime pickup and delivery problem with split loads. COR.

3 Korsvik, Fagerholt and Laporte (2011) A large neighbourhood search heuristic for ship routing
and scheduling with split loads. COR
Model notation: Maritime PDPTWSL
• Path flow formulation *

• Variables:
– xvrw representing a combination of a geographical route r for a
ship v, and a delivery pattern w
– yi representing the pickup of an optional cargo i
• Parameters:
– Qi – quantity of cargo i
– Qivrw – quantity of cargo i picked up on route r for a ship v by
delivery pattern w
* Stålhane, Andersson, Christiansen, Cordeau, Desaulniers (2012) Branch-price-and-cut for a maritime

pickup and delivery problem with split loads. COR.
Path flow formulation: PDPTWSL
max  P   Py    C
i i i
vV rRv wW
x ,
vr vrw
iN C iN O
  A x
ivr vrw  1, i  N C ,
vV rRv wW Cargo assignment
 
vV rRv wW
Aivr xvrw  yi  0, i  N O ,
x
rRv wW
vrw 1 v  V , Convexity
  Qivrw xvrw  Qi ,
vV rRv wW
i  N C ,
Quantity assignment
 Q
vV rRv wW
x
ivrw vrw  Qi yi  0, i  N O ,
x
wW
vrw  {0,1}, v  V , r  Rv ,
yi  {0,1}, i  N O .
Exact Branch-and-Price-and-Cut
• Master problem
– Linear relaxation of the Path flow formulation
• Subproblem
– One for each ship
– ESPPRC + Multi-dimensional Knapsack problem
– Solved by Dynamic Programming
• Dominance is tricky because:

– Quantities of a partial path can only be calculated every time the ship is
empty.
– Otherwise, the value of a partial path is dependent on what happens
next, not only what has happened.
-> Makes the subproblem difficult to solve quickly
Complex maritime cargo routing
• Flexible cargo quantities

• Soft time windows
• Restricted opening hours in port
• Speed optimization
• Stowage considerations
• Bunkering decisions
• Dry-docking decisions
• Fleet size and mix
• Considering uncertainty
– Modelling of market opportunities (TC and spot cargo rates
fluctuate)
– Sailing time and time spent in ports
* - Some research results exist
Outline
Maritime inventory routing
• Often, inventories exist at the loading and unloading ports
of the sailing leg
• When the ship operator has the responsibility for both

transportation and inventories
 Maritime inventory routing problem (MIRP)
Large interest to
combine inventory
management and
routing from industry
(shipping companies or
vertically integrated
companies)
Basic MIRP:
Design routes and schedules that minimize the transportation cost
without interrupting production and consumption of the product
Ships with capacity and cost structure

Constrained - Routing Sufficient amount
inventory capacity available
-Arrival time
- Loading quantity
Number of
Number of calls
calls
Arrival time Arrival time
Loading Unloading
quantity quantity
Production of Consumption of
product product
- product
Routing
• Generate routes including information about time to

start service and (un)loading quantities
• Partial (un)loading, such that several (un)loading ports
can be visited in succession
Geographical route P1→D2→D1→P2→D1→P2→D2
Schedule T1 T2 T3 T4 T5 T6 T7
Quantity Q1 Q2 Q3 Q4 Q5 Q6 Q7
Basic MIRP: Inventory management
• Single product
• Inventory storages in all ports
• Production/consumption rates are constant in time
• Partial (un)loading allowed
Production port Consumption port
S im S im
S im S im
Time
m 1 m2 m 1 m2
Example: Maritime IRP
Example: Maritime IRP
Comparisons to other routing problems
• MIRP is different from maritime cargo routing due to

– The number of visits at a node is predetermined
– The quantity loaded/unloaded is normally fixed
– Predefined pickup and delivery pair
for maritime cargo routing
• MIRP is different from road-based IRP due to

– Often one central supplier (depot) with large inventory capacities and
many customers consuming a small amount
– Asymmetry between each type of inventory
– Small quantities are delivered compared to the maritime IRP
– Finite planning horizon
for road-based IRP
Model description
• Minimise transportation and port
costs
• Each port call is visited at most once
• Flow balance constraints
• Load management on board the ship
Subproblem constraints
• Ship capacity constraints in a path flow model
• Scheduling constraints
• Time windows
• Inventory balance constraints
• Inventory capacity constraints
Path flow model – MIRP
sim
• Master Problem S im
– Production / consumption rate: Ri Ri
– Inventory management at port visit (i,m), sim
S im
• Path r for each ship v m 1 m  2
– Ship routing X imjnvr vr
– Ship scheduling Timvr vr Qimvr
– Ship loading/unloading Qimvr vr v
i,m j,n
X imjnvr
Timvr
Path flow model for MIRP
min Cvr vr ,

vV rRv
si ( m 1)    Ri (Timvr  Ti ( m 1) vr )  I i Qi ( m 1) vr  vr  sim  0, i  N , m  M i , Inventory

vV rRv balance
S i  sim  S i , i  N , m  M i ,
Inventory
limits
S i  sim    RiTivQ Qimvr  I i Qimvr  vr  S i , i  N , m  M i ,
vV rRv

rRv
vr  1, v  V , Convexity
X
rRv
imjnvr vr  0,1, i  N , m  M i ,
j  N , n  M i , v V ,
vr  0, v  V , r  Rv .
Solution methods
• Very complex problem to solve due to high degree of
freedom
– The variable quantities, no predetermined number of port
calls, no pickup and delivery pairs
– Problem specific issues
• Existing research on solution methods are mainly

– Several branch-and-price approaches1,2)
– Heuristics (constructive and improvement heuristics)3)
1) Christiansen (1999). Decomposition of a combined inventory and time constrained ship routing problem.
Transportation Science.
2) Grønhaug, Christiansen, Desaulniers, Desrosiers (2010). A branch-and-bound method for a liquefied natural gas
inventory routing problem. Transportation Science.
3) Christiansen, Fagerholt, Flatberg, Haugen, Kloster, Lund (2010). Maritime inventory routing with multiple products: A
case study from the cement industry, European Journal of Operational Research
Complex MIRPs
• Multiple products
• Constant, time varying or variable consumption/production rate
• Inventory constraints in either production or consumption ports

– Contracts with given quantities and time windows in the other type
of ports
• Combined inventory routing and tramp shipping

– Inventory management in some ports
– Both contracted and optional spot cargoes
– Often beneficial to trade cargoes with other operators to better
utilize the fleet and to ensure the product balance at own ports
• Combining inventory routing with other planning aspects

– Consider larger parts of the supply chain (for instance include
production)
Outline
Trends (1:2)
• Mergers and collaborations

– To increase market power and gain flexibility in the services
– Larger fleet and more complex transportation pattern
• New generation of planners

– With computer experience and more academic background
• Shift from industrial to tramp shipping

– Outsourcing the transportation
– Industrial shipping companies are also in the spot market
– More integration with the market
Trends(2:2)
• Focus on supply chains (SC)

– Integration along the SC (cargo owner and shipping company)
– Vendor managed inventory systems
– Shipping integrated into a multimodal door-to-door SC
• An increasing LNG activity

– Challenging inventory routing problems
• Oversupply of tonnage
– Years with global financial crisis. Economic situation has
brightened, but multiple risks threaten for a stable world
economy.
– Reduced growth in world seaborne trade
– Slow steaming and use of less costly routes
Research Trends
• Fast growth of the containership fleet

– Similar growth in research on liner network design and related
topics
• MIR, LNG and offshore supply vessel routing have

attracted increasing attention
• Increasing price of bunker fuel
– Sailing speed (third power relationship between bunker fuel
consumption and speed)
– Environmental impact of ships and CO2 emissions
• More theoretical models and basic research that are less

grounded in real operations
Outline
Links to Location Analysis
• Maritime supply chains
– At least one sea transportation leg
– Which ports with storages should be used?
– Size of storages at ports
• Liner shipping design

– Location of hubs
– Hub and spoke assignments
• Floating units or installations at sea for storage

– Where to locate the hubs
– Type of hubs
Maritime supply chain design for
distribution of pellets
• Pellets are compressed dried wood used for heating
• Pellets supply chain:
Pellets distribution in Norway
• Many of the customers and the main producer of pellets are

located close to the coast
• Project:
– Estimate the profitability of a maritime distribution network for
the producer instead of the current land based network
• Main decisions:
– The number of ships needed in different seasons
– Which ports to use and where to establish storages
• Also consider routing and storage inventory management
K. Andersen, A. Sjamsutdinov (2011). Maritime supply chain design for distribution of pellets.
Master Thesis, NTNU, Trondheim, Norway.
Map and
simplified supply chain structure
Structure of the distribution network
• Model exploiting the

geography
• Branch-and-cut
solution approach
• The profitability of a
maritime distribution
network is more than
3 mill. USD per year
with an operating
profit margin of 8 %.
Container liner shipping design
Two categories of Container ship
The huge main line ships that
sail the world's major trade
routes between hub ports
(e.g. Singapore, Rotterdam)
The smaller "FEEDER"

container ships that will ferry
containers between hubs and
feeder ports (spokes)
A liner network structure and problem
- Hub ports
- Spoke ports
- Location of hub ports on a

circular hub route
- Choosing the spoke and hub
connections
- An upper bound on how many
hub ports a spoke port can be
connected to
S. Gelareh , D. Pisinger (2011). Fleet deployment, network design and hub location of liner
shipping companies. Transportation Research Part E: Logistics and Transportation Review
47(6), 947 - 964
A typical solution and solution method
Four vessel types
In addition to hub locations and

hub-spoke allocation,
the model deploys different
ship types on the routes.
The model is solved by

decomposition following
Bender’s principles to isolate
the integer decisions from the
flow decisions
S. Gelareh , D. Pisinger (2011). Fleet deployment, network design and hub location of liner
shipping companies. Transportation Research Part E: Logistics and Transportation Review
47(6), 947 - 964
Maintenance operations of offshore
wind farms
Wind
farm
• Doggerbank
case study:*
Distance to
shore: > 200
km (108 nmi)
Offshore
Onshore station • Wind farms
harbour need regularly
maintenance
and
inspections
* Ongoing research project at MARINTEK
Offshore wind farm problem
• Problem: Executing operation and maintenance (O&M) activities
during the operational phase of an offshore wind farm at minimum
cost
• Determine the number and type of vessels to support the O&M

activities:
– Mothership/daughtership vessel concepts
– Supply vessels - Crew transfer vessels
– Crane vessels - Helicopters
• Determine the infrastructure of the offshore transport system:

– Onshore harbour
– Offshore station (platform, mothership)
Dutch Harbour at sea
Mother vessel concepts
Ship or helicopter
Some of the model complexities
• Invest in different type of resources and infrastructure:
– Choose to invest in existing vessel concepts
– Invest in "new"/extended vessels concepts such as
• Flotel (offshore hotel accommodation)
• mothership/daughtership concepts
– Invest in offshore maintenance platforms
• Several types of maintenance activities (corrective,

preventive)
• Uncertainties
• One year plan horizon
• Today: Deterministic MIP model
Outline
Concluding remarks
• Focused on ship routing and scheduling

• Ship scheduling has high potential for improvements by
optimizing fleet utilization (a few % - 25 %)
• A few DSSs are described in the literature
• Rich real problems and external conditions changes rapidly
• Some future research challenges
– Liner network design (obstacles: uncertainties, problem size and complexity)
– Combining ship routing with port management issues (location analysis)
– Considering uncertainties (sailing and port times, demand, cost of ships, freight rates)
– Incorporating more real aspects
– Solve larger MIRP instances
Optimization of
Maritime Transportation
Marielle Christiansen
Department of Industrial Economics and Technology Management
Norwegian University of Science and Technology,
Trondheim, Norway
EURO Working Group on Locational Analysis (EWGLA),

Ankara, Turkey, 17-19 April 2013

EWGLA 2013 Marielle Christiansen

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

EWGLA 2013 Marielle Christiansen

Uploaded by

Copyright:

Available Formats

Optimization of

EURO Working Group on Locational Analysis (EWGLA),

M. Christiansen, K. Fagerholt, B. Nygreen and

M. Christiansen, K. Fagerholt, and D. Ronen

M. Christiansen, K. Fagerholt, B. Nygreen and

• Maritime cargo routing

• Maritime inventory routing

• Trends relevant for ship routing and scheduling

• Links to location analysis

Photo: Gary Blakeley

Stahlbock, Voss (2008) Operations Research at container terminals: a

• Major mode of international trade

• Monopoly on transportation of large volumes between

Increase last decade: 40 %

Example: A VLCC, 200.000 dwt:

• Good planning is crucial:

• However, relatively little OR-work has been done in ship

• Ronen, EJOR (1983)

• Ronen, EJOR (1993)

• Christiansen, Fagerholt and Ronen, TS (2004)

• Christiansen, Fagerholt, Nygreen and Ronen, EJOR (2013)

• Tactical planning level (1 week – 1 year)

• Operational planning (1 day – 1 week)

• Maritime cargo routing

• Maritime inventory routing

• Trends relevant for ship routing and scheduling

• Links to location analysis

• Focus on tramp shipping

• A cargo routing problem in industrial shipping can be

• Models and solution methods

• Extensions and variants

• Mix of mandatory contract cargoes and optional spot

• Objective: Maximize profit

• Using a voyage estimation software

• Few optimization-based DSSs in use

• Building complete schedules is done manually

• Graphical problem representation

• Spreadsheet for visualization

• Improvement potential is significant

Planning horizon Multiple cargoes

• No natural depot • Cargoes with specified pickup

• Maritime Pickup and Delivery Problem with Time

max  Pi   Pi yi    Cvr xvr N C - Set of contracted cargoes, i

x vr  0,1 v  V , r  Rv xvr = 1 if ship v sails route r

However, in real life the planner often has more flexibility

... thus the focus has recently been on model extensions

• A priori column generation1

maritime pickup and delivery problem with split loads. COR.

and scheduling with split loads. COR

• Path flow formulation *

* Stålhane, Andersson, Christiansen, Cordeau, Desaulniers (2012) Branch-price-and-cut for a maritime

• Dominance is tricky because:

• Flexible cargo quantities

• Maritime cargo routing

• Maritime inventory routing

• Trends relevant for ship routing and scheduling

• Links to location analysis

• When the ship operator has the responsibility for both

Ships with capacity and cost structure

• Generate routes including information about time to

Geographical route P1→D2→D1→P2→D1→P2→D2

• MIRP is different from maritime cargo routing due to

• MIRP is different from road-based IRP due to