Collaborative Logistics in the Forest Industry Mikael

Collaborative Logistics in the Forest Industry Mikael

CollaborativeLogistics
intheForestIndustry
MikaelRönnqvist
Université Laval,Québec,Canada
SSAFR 2015 – 16th Symposium for Systems Analysis in Forest
Resources, August 19-21, Uppsala, Sweden
Outline
• Collaboration ‐ basis
• Collaboration forms & industry cases
– Collaboration by horizontal integration
– Collaboration by standardization
• Challenges / open questions
• Concluding remarks
Presentation based on joint work together with many organisations, companies and colleagues.
COLLABORATIONBASICS
Definitionofcollaboration
“An intentional cooperative action between two or more entities that exchange or share resources, with the goal of making decisions or realizing activities that will generate shared advantages or losses. “
J.F. Audy, N. Lehoux, S. D’Amours, M. Rönnqvist, A Framework for an Efficient Implementation
of Logistics Collaborations, Int. Transactions of Operations Research, Vol 19, 633-657, 2012.
COLLABORATIONTHROUGH
VERTICALINTEGRATION
Example with one integrated company:
J. Troncoso, S. D'Amours, P. Flisberg, M. Rönnqvist, A. Weintraub, A mixed integer programming
model to evaluate integrating strategies in the forest value chain - A case study in the Chilean
forest industry, Canadian Journal of Forest Research, Vol 45, 937-949, 2015.
Example with vendor managed inventory (VMI):
D. Carlsson, P. Flisberg, M. Rönnqvist, Using robust optimization for distribution and inventory
planning for a large pulp producer, Computers & Operations Research, Vol. 44, 214-225, 2014.
COLLABORATIONTHROUGH
HORIZONTALINTEGRATION
Potentialincollaboration withtwo
companies sharing supply anddemand
Simple approach
Norra
Sala
Hallstavik
Heby
Östra
Västerås
Västra
Backhaulage tour
Norra
Hallstavik
Heby
Sala
Östra
Västerås
Västra
Casestudy
• 8 participating companies in southern Sweden
• 898 supply areas
• 101 industries
• 39 assortments, 12 assortment groups
M. Frisk, M. Göthe-Lundgren, K. Jörnsten and M. Rönnqvist, Cost allocation in collaborative
forest transportation, European Journal of Operational Research, Vol. 205, 448-458, 2010.
Volume/
month
Proportion
(%)
77,361
8,76
Company 2
301,660
34,16
Company 3
94,769
10,73
Company 4
44,509
5,04
Company 5
232,103
26,29
Company 6
89,318
10,12
Company 7
36,786
4,17
Company 8
6,446
0,73
Company
Company 1
Total
882,952
1
2
3
4
5
6
7
8
Optimizationmodel
•
•
•
•
•
# supply points: 4842
# demand points: 310
# constraints: 5,053
# variables (direct flows): 240,000
# variables (backhaul flows): > 400 million
– (240,000 + 125,00 through column generation)
opt1: direct flows (individual)
opt2: backhaul flows (individual)
Company
Company 1
Company 2
Company 3
Company 4
Company 5
Company 6
Company 7
Company 8
All
Total
Saving (%)
opt3: direct flows (collaboration)
opt4: backhaul flows (collaboration)
Real
3,894
15,757
4,828
2,103
10,704
opt 1
3,778
14,859
4,742
2,067
10,340
opt 2
3,640
14,684
4,703
2,043
10,153
5,084
1,934
0,333
4,959
1,884
0,333
4,826
1,877
0,332
44,637
0,00
42,963
3,75
42,257
5,33
opt 3
opt 4
39,253
39,253
12,06
38,315
38,315
14,16
Costallocationbasedonvolume
Company
Company 1
Individual
3,778
Volume
3,439
Company 2
Company 3
Company 4
Company 5
Company 6
Company 7
Company 8
Sum
14,859
4,742
2,067
10,340
4,959
1,884
0,333
42,963
13,411
4,213
1,979
10,318
3,971
1,635
0,287
39,253
%
9,0
9,7
11,2
4,3
0,2
19,9
13,2
14,0
Gametheoryapproach– Basics
coalition S : a subset of participants N (or players)
grand coalition S  N : all participants
c(S) : cost for the transportation for S
Efficient allocation :  y j  c(N)
jN
Individual rational : y j  c( j)
Core :
y
jS
j
 c(S),
SN
Models
Volume based allocation : y j  w j c( N )
w j is equal to participant j' s share of the total volume
Shapley value : y j 
y
S  N : jS
j
 S  1 N  S  c(S )  c(S   j)
N
VolumevsShapleyvalue
Company
Company 1
Volume
3,439
Company 2
%
9,0
Shapley
3,586
%
5,1
13,411
9,7
13,528
9,0
Company 3
4,213
11,2
4,102
13,5
Company 4
1,979
4,3
1,889
8,6
Company 5
10,318
0,2
9,747
5,7
Company 6
3,971
19,9
4,503
9,2
Company 7
1,635
13,2
1,587
15,8
Company 8
0,287
14,0
0,310
6,9
Sum
39,253
39,253
Stable
No
Yes
Model– EqualProfitMethod
c 
i   yi
y
 1  i . Stable allocation c 
i   yi .
c 
i
c 
i
yj
yi
Relative difference between tw o participan ts :

c 
i  c  j
Minimize difference i.e.
Relative saving :
min f
s.t.
yj
yi

f 
c 
i  c  j
y
j
 c ( S ), S  N
y
j
 c( N )
jS
jN
EPM
Company
Volume
Shapley
EPM
Company 1
9,0
5,1
6,7
Company 2
9,7
9,0
8,8
Company 3
11,2
13,5
8,8
Company 4
4,3
8,6
8,8
Company 5
0,2
5,7
8,8
Company 6
19,9
9,2
8,8
Company 7
13,2
15,8
8,8
Company 8
14,0
6,9
8,8
Morepracticalconsiderations
–
–
–
–
–
–
Three companies agreed to continue collaboration
No sharing of sensitive information
No financial transactions
Higher (minimum) responsibility for own mills
Cost sharing included directly in the OR model
Long term relations to motivate same profit proportion
– Design process of coalitions • Business model
• Leadership
• M. Guajardo and M. Rönnqvist, Operations Research models for coalition structure in collaborative
logistics, European Journal of Operational Research, Vol. 240,147–159, 2015.
• J.F. Audy, N. Lehoux, S. D’Amours, M. Rönnqvist, A Framework for an Efficient Implementation
of Logistics Collaborations, Int. Transactions of Operations Research, Vol 19, 633-657, 2012.
• J.-F. Audy, S. D’Amours and M. Rönnqvist, An empirical study on coalition formation and
cost/savings allocation, International Journal of Production Economics, Vol 136, 13–27, 2012.
M. Guajardo, K. Jörnsten, M. Rönnqvist, Constructive and blocking power in collaborative
transportation, OR Spectrum, published online
Costallocationmethodsin
transportation,VRP,TSPandinventory
• 55 scientific articles (most since 2010)
• 40 different cost allocation methods based on game theory
–
–
–
–
–
Ad hoc methods: 31
Shapley values: 23
Proportional : 18
Nucleoulus: 12
Dual: 8
• Savings reported: 5‐40% • Number of companies: 2‐50 (most 2‐5)
M. Guajardo, M. Rönnqvist, A review on cost allocation methods in collaborative transportation,
International Transactions in Operational Research, to appear
Forest biomass - COUNTRY-WIDE STUDY
All yearly forest fuel
transport operations in
Sweden:
58,000 harvest sites
647 heating plants and
terminals
200,000 transports
6.1 million ton of
biomass
Many companies
involved (28+1)
P. Flisberg, M. Frisk, M. Rönnqvist, M. Guajardo, Potential savings and cost allocations for forest
fuel transportation in Sweden: A country-wide study, Energy, Vol. 85, 353-365, 2015.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
RESULTS
~12% (6%) savings by collaboration.
RESULTS
~22% savings by adding all aspects
together, representing 140 million SEK
COLLABORATIONTHROUGH
STANDARDIZATION
CALIBRATEDROUTEFINDER
Europe (standard):
Max 18 m
Max 40 ton
Sweden:
Max 25 m
Max 64 ton
Film1
Driving time
Stress
Safety
Hilliness
Fairness
Curvature
Maintenace
Distance
Fuel consumption
shortest path fastest path
Distance (km)
84.3
85.8
105.6
total resistance (points)
2094
2407
1492
Time (min)
74.7
72.6
87.1
Curviness (points)
161
190
55
Hilliness (points)
74
36
3
Width 3‐3.5m (km)
8
19
0
Gravel road (km)
20
33
9
RC 0 (km)
0
0
0
RC 1
1
0
3
RC 2
0
2
4
RC 3
0
0
67
RC 4
9
19
14
RC 5
18
13
0
RC 6
44
24
6
RC 7
3
5
9
RC 8
3
14
0
RC 9
7
8
1
‐ 73 million m3
‐ 2 million loads
‐ 200,000 new landings annually
‐ 440,000 unique routes annually
‐ 800 mills/terminals
‐ 2,000 logging trucks ‐ 5,000 truck drivers
‐ Value 600 million euros
• SDC is the logistic hub in Swedish forestry
• Responsible for SNVDB – the forest National road database
• Daily update of national road database
• CRF ‐ in operation since 2009
• 4 full time staff at SDC for management
• 3 webservers (3,000 calls per day)
2000
500
Contract
Payment
Transportassignment
Transportdistanceandroute
Invoicedistanceandpayload
Invoicebase
Deviation reports
Swedish road network
542 000 km roads
• 102 000 km government
• 56 000 km local communities
• 384 000 km private
GIS network
• 4.43 million arcs • 2.20 million nodes
Road features (No. attributes)
•
•
•
•
•
•
•
•
•
•
•
Functional road class (10)
Maximum speed (13)
Road width (16)
Bearing class (4)
Terrain class (3)
Road owner (3)
Timber route (2)
Passing route (2)
Length (1)
Ferry line (1)
Surface type (2)
Road features (No. attributes)
•
•
•
•
•
•
•
•
•
•
•
Functional road class (10)
Maximum speed (13)
Road width (16)
Bearing class (4)
Terrain class (3)
Road owner (3)
Timber route (2)
Passing route (2)
Length (1)
Ferry line (1)
Surface type (2)
• Curvature (30)
• Hilliness (20)
Geometry coordinates (x,y,z)
Key‐routes
1500 best practice routes agreed between forest and transport companies Translation process
Road feature (No. attributes)
Objectives
• Fairness
• Distance
• Driving time
• Safety • Fuel consum.
• Work environ.
• Maintenance
• Stress
• Curvature
• Hilliness
107
weights ?
Key routes
• Functional road class (10)
• Maximum speed (13)
• Road width (16)
• Bearing class (4)
• Terrain class (3)
• Road owner (3)
• Timber route (2)
• Passing route (2)
• Length (1)
• Ferry line (1)
• Surface type (2)
• Curvature (30)
• Hilliness (20)
Solution method
Select initial weights and Generate initial minimum cost (MC) routes
Solve MIP or LP problem  new weights
1500 key routes and previous weights
[MIP] max z   ci yi    lij xij
iI
s.t.
iI jJ i
a
ik
wk   bijk wk  M ij (1  yi ), i, j
b
ijk
wk   bimk wk  M ijm (1  xij )  M ijm yi , i, j , m
k K
k K
k K
kK
yi   xij  1, i
L P 
jJ i
wk  wl ,
Generate new MC routes and constraints
Convergence criteria
stop
k s, l. t .
wk   wk , wk  ,
k
yi , xij  0,1 ,
i, j

(
w
 zi,
m in z 
i I

a

b ijk w
k K
k K
ik
k
 wl,
w
k

w k
z ij  0 ,
•
•
•
•
k
w

a
k K
k
ik
 zi,
w
k
 z i ) / li
 i
 i, j
 k ,l
, w
k
,
 k
 i, j
0.1 second for each MC route
3 minutes for each MIP
Average 10‐15 iterations
Average 3600 new MC routes.
Film2
Generated routes related to normalized key routes
1,3
1,2
1,1
1
0,9
shortest path
0,8
0,7
0,6
0,5
0
500
1000
sorted key routes
1500
Generated routes related to normalized key routes
1,3
1,2
1,1
1
0,9
shortest path
0,8
fastest path
0,7
0,6
0,5
0
500
1000
sorted key routes
1500
Generated routes related to normalized key routes
1,3
1,2
1,1
1
0,9
shortest path
0,8
fastest path
0,7
2009 setting
0,6
0,5
0
500
1000
sorted key routes
1500
Generated routes related to normalized key routes
1,3
1,2
1,1
1
0,9
shortest path
0,8
fastest path
0,7
2009 setting
0,6
2015 setting
0,5
0
500
1000
sorted key routes
1500
CRF – Development over time
80
70
60
50
40
30
20
10
0
2009
2010
2011
2012
2013
2014
2015
CRF– Development overtime
80
70
60
50
40
30
20
10
0
2009
2010
2011
2012
2013
2014
2015
Qualitative analysis of CRF routes
1500 Key‐routes Average dist. (km) Key‐routes
87.4
CRF
86.8
Shortest path
80.7
Fastest path
83.2
Quality (points)
10,198
9,983 (2.1% difference)
15,431 (‐51% loss)
12,930 (‐21% loss)
2 million trp CRF
Shortest path
Fastest path
Quality (points)
9,926
17,408 (‐75% loss)
13,616 (‐37% loss)
Average dist. (km) 88.3
81.3
84.3
CHALLENGES
Relatedopenproblemsfor
collaborationand/orcoordination
OPEN PROBLEM 5 (Operational harvesting)
How can we use pricing as a coordination mechanism for forest operations?
OPEN PROBLEM 6 (Transportation and routing)
How can we model and propose efficient sharing principles for practical
collaboration in transportation?
OPEN PROBLEM 10 (Tactical planning)
How can we develop methodology to coordinate and synchronize a set of
stakeholders with individual agendas?
OPEN PROBLEM 24 (Value chain management)
How can we design the coordination mechanisms needed to integrate and
synchronize the units within the forest value chain?
TAKEAWAYMESSAGE
Takeawaymessages
• Very large potential for savings in logistic collaboration
• Many game theoretic models developed for collaboration and analysis
• Many obstacle to overcome before practical use
– Trust, information sharing, decision support systems, standard, sharing mechanisms
– Be aware of peoples personal motivation to participate
• Educational tools to understand problem and solution concepts for managers, planners / students
• Interesting future challenges for OR analysts
CollaborativeLogistics
MikaelRönnqvist
Université Laval,Québec,Canada
[email protected]
Questions?
SSAFR 2015 – 16th Symposium for Systems Analysis in Forest Resour
August 19-21, Uppsala, Sweden