Transcript Thesis Proposal Meeting

1.206J/16.77J/ESD.215J
Airline Schedule Planning
Cynthia Barnhart
Spring 2003
1.206J/16.77J/ESD.215J Airline
Schedule Planning: Multi-commodity
Flows
Outline
•
•
•
•
•
Applications
Problem Definition
Formulations
Solutions
Results
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
2
Application I
• Package flow problem (express package
delivery operation)
– Shipments have specific origins and destinations
and must be routed over a transportation
network
– Each set of packages with a common origindestination pair is called a commodity
• Time windows (availability and delivery time)
associated with packages
– The objective might be to minimize total costs,
find a feasible flow, ...
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
3
Application II
• Passenger mix problem
– Given a fixed schedule of flights, a fixed
fleet assignment and a set of customer
demands for air travel service on this
fleeted schedule, the airline's objective is to
maximize revenues by accommodating as
many high fare passengers as possible
– For some flights, demand exceeds seat
supply and passengers must be spilled to
other itineraries of either the same or
another airline
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
4
Application III
• Message routing problem
– In a telecommunications or computer
network, requirements exist for
transmission lines and message requests, or
commodities.
– The problem is to route the messages from
their origins to their respective destinations
at minimum cost
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
5
MCF Networks
• Set of nodes
– Each node associated with the supply of or
demand for commodities
• Set of arcs
– Cost per unit commodity flow
– Capacity limiting the total flow of all
commodities and/ or the flow of
individual commodities
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
6
MCF Commodity Definitions
• A commodity may originate at a subset of
nodes in the network and be destined for
another subset of nodes
• A commodity may originate at a single node
and be destined for a subset of the nodes
• A commodity may originate at a single node
and be destined for a single node
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
7
MCF Objectives
• Flow the commodities through the networks
from their respective origins to their
respective destinations at minimum cost
– Expressed as distance, money, time, etc.
• Ahuja, Magnanti and Orlin (1993)-- survey of
multi-commodity flow models and solution
procedures
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
8
MCF Problem Formulations -Linear Programs
• Network flow problems
– Capacity constraints limit flow of individual
commodities
– Conservation of flow constraints ensure flow
balance for individual commodities
– Flow non-negativity constraints
• With side constraints
– Bundle constraints restrict total flow of ALL
commodities on an arc
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
9
MCF Constraint Matrix
Network flow
problem,
commodity k=1
Network flow
problem,
commodity k=2
Network flow
problem,
commodity k=3
Network flow
problem,
commodity k=4
Bundle constraints limiting total flow of all commodities to arc capacities
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
10
Alternative Formulations for O-D
Commodity Case
• Node-Arc Formulation
– Decision variables: flow of commodity k on each arc ij
• Path Formulation
– Decision variables: flow of commodity k on each path for
k
• “Tree” or “Sub-network” Formulation
– Define: super commodity: set of all (O-D) commodities
with the same origin o (or destination d)
– Decision variables: quantity of the super commodity k’
assigned to each “tree” or “sub-network” for k’
• Formulations are equivalent
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
11
Sample Network
2
a
1
c
3
b
Arcs
i j cost capy
1
1
2
2
3
7/22/2015
2
3
3
4
4
1
2
3
4
5
20
10
20
10
40
d
4
e
Commodities
# o d quant
1
2
3
4
Barnhart 1.206J/16.77J/ESD.215J
1
1
2
3
3
4
4
4
5
15
5
10
12
Notation
Parameters
•
•
•
•
A: set of all network arcs
K: set of all commodities
N: set of all network nodes
O(k) [D(k)]: origin [destination]
node for commodity k
• cijk : per unit cost of
commodity k on arc ij
• uij : total capacity on arc ij
(assume uijk is unlimited for
each k and each ij)
• dk : total quantity of
commodity k
7/22/2015
Decision Variables
• xijk : number of units
of commodity k
assigned to arc ij
Barnhart 1.206J/16.77J/ESD.215J
13
c
k
ij
Minimize
ij
Node-Arc Formulation
k
x ij
k
subject to
x
k
ij
  x ji  d k
k
j
if i  O ( k )
j
  d k if i  D ( k )
0
x
k
k
ij
 uij
k
ij
0
x
otherwise
(i , j )  A
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
a
b
c
d
e
: Nonnegativity constraints
k2
b
1
-1
c
d
1
-1
1
-1
e
k3
a
b
1
-1
1
c
d
1
-1
1
e
k4
a
b
1
-1
1
c
d
1
-1
1
e
a
b
1
-1
1
c
d
1
-1
1
-1
1
-1
-1
-1
1
-1
-1
-1
1
1
1
1
1
1
1
1
cb
1
xb
1
cc
1
xc
1
cd
1
xd
1
1
1
1
1
ce
1
xe
1
1
1
2
ca
2
xa
2
cb
2
xb
2
cc
2
xc
2
cd
2
xd
2
ce
2
xe
1
1
3
ca
3
xa
1
-1
1
1
1
1
e
1
-1
-1
ca
1
xa
7/22/2015
: Bundle constraints
( i, j )  A, k  K
k1
a
1
-1
: Conservation of Flow
3
cb
3
xb
3
cc
3
xc
Barnhart 1.206J/16.77J/ESD.215J
3
cd
3
xd
3
ce
3
xe
1
4
ca
4
xa
4
cb
4
xb
4
cc
4
xc
4
cd
4
xd
RHS
= d1
= 0
= -d1
= 0
= d2
= 0
= 0
= -d2
= 0
= d3
= 0
= -d3
= 0
= 0
= d4
= -d4





4
ce
4
xe
14
ua
ub
uc
ud
ue
Additional Notation
Parameters
Decision Variables
• Pk: set of all paths for • f : fraction of total
p
commodity k, for all k
quantity of
• cp : per unit cost of
commodity k
commodity k on path p
assigned to path p
= ij p cijk
• ijp : = 1 if path p
contains arc ij; and = 0
otherwise
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
15
O/D Based Path Formulation
 d
Minimize
k
subject to
 d
p P
k

p Pk
fp
k
f
p
ij p
 u ij
p P
k
C
f
p
p
k
(i, j )
 A
: Bundle constraints
k
fp
 0
 1
k
p
 K
 Pk , k
: Flow balance constraints
 K
: Non-neg. constraints
Path
k=1
k=2
k=3
k=4
RHS
Dual
a
d1
0
d2
d2
0
0
0
0
<= ua
a
b
0
d1
0
0
d2
0
0
0
<= ub
b
c
d1
0
d2
0
0
d3
0
0
<= uc
c
d
0
0
0
d2
0
0
d3
0
<= ud
d
e
0
0
d2
0
d2
d3
0
d4
<= ue
e
k=1
1
1
=1

=1

=1

=1

k=2
1
1
1
k=3
1
1
k=4
1
Cost.
C1 d1
C 2 d1
Variable
f1
f2
7/22/2015
C3 d 2
f3
C4 d 2
f4
C5 d 2
f5
C6 d 3
C7 d 3
C8 d 3
f6
f7
f8
Barnhart 1.206J/16.77J/ESD.215J




16
Additional Notation
Parameters
• S: set of source nodes
nN for all commodities
• Qs: the set of all subnetworks originating at s
• TCqs: total cost of subnetwork q originating at s
• pq : = 1 if path p is
contained in sub-network
q; and = 0 otherwise
7/22/2015
Decision Variables
• Rqs : fraction of
total quantity of the
super commodity
originating at s
assigned to subnetwork q
Barnhart 1.206J/16.77J/ESD.215J
17
Sub-network Formulation
 (   C
Minimize
subject to
 (   d
q Q
s
s
R s
q Q
q
k s p P
1
k
sS q Q s
k q p P k
 p q dk ) Rq
s
p
ijp ) Rqspq  uij ( i , j )  A
: Capacity Limits on Each Arc
k
s  S
: Mass Balance Requirements
s
R qs  0
q  Qs , s S
: Nonnegative Path Flow Variables
Sub- network
o=1
o=2
o=3
RHS
Dual
a
d1+ d2
d1+ d2
d1
d2
d2
0
0
0
0
<= ua
a
b
0
0
d2
d1
d1
d1+ d2
0
0
0
<= ub
b
c
d1
d1+ d2
d1
0
d2
0
d3
0
0
<= uc
c
d
d2
0
0
d2
0
0
0
d3
0
<= ud
d
e
0
d2
d2
0
d2
d2
d3
0
d4
<= ue
e
o=1
1
1
1
1
1
1
=1
 
=1
 
=1
 
o=2
1
1
o=3
1
1
Cost.
TC1
Variable
R1
7/22/2015
1
1
TC 2
1
R2
1
TC 4
1
R4
TC 3
R3
1
1
1
TC 5
1
R5
1
2
TC 6 TC1
1
R6
2
R1
2
TC 2
2
R2
Barnhart 1.206J/16.77J/ESD.215J



3
TC1
3
R1
18
Linear MCF Problem Solution
• Obvious Solution
– LP Solver
• Difficulty
– Problem Size: (|N|=|Nodes|, |C|=|Commodities|,
|A|=|Arcs|)
• Node-arc formulation:
– Constraints: |N|*|C| + |A|
– Variables: |A|*|C|
• Path formulation:
– Constraints: |A| + |C|
– Variables: |Paths for ALL commodities|
• Sub-network formulation:
– Constraints: |A|+|Origins|
– Variables: |Combinations of Paths by Origin|
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
19
General MCF Solution Strategy
• Try to Decompose a Hard Problem Into a Set of
Easy Problems
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
20
MCF Solution Procedures I
• Partitioning Methods
– Exploit Network Structure to Speed Up Simplex
Matrix Computations
• Resource-Directive Decomposition
– Repeat until Optimal:
• Allocate Arc Capacity Among Commodities
• Find Optimal Flows Given Allocation
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
21
MCF Solution Procedures II
• Price-Directive Decomposition
– Repeat until Optimal:
• Modify Flow Cost on Arc
• Ignore Bundle Constraints, Find Optimal Flows
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
22
Revisiting the Path Formulation
MINIMIZE  k K  pPk dk cp fp
subject to:
pPk  k K dk fpijp  uij  ijA
pP(k) fp = 1  kK
fp  0  pPk,  kK
22-Jul-15
1.224J/ESD.204J
23
By-products of the Simplex
Algorithm: Dual Variable Values
Duals
-ij: the dual variable associated with the bundle
constraint for arc ij ( is non-negative)
k : the dual variable associated with the commodity
constraints
Economic Interpretation
 ij : the value of an additional unit of capacity on arc
ij
 k/dk : the minimal cost to send an additional unit of
commodity k through the network
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
24
Modified Costs
Definition: Modified cost for arc ij and
commodity k = cijk+ij
Definition: Modified cost for path p and
commodity k = ijA (cijk + ij )ijp
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
25
Optimality Conditions for the Path Formulation
f*p and *ij , *k are optimal for all k and all ij iff:
Primal feasibility is satisfied
1. pPk  k K dk f*pijp  uij  ijA
2. pP(k) f*p = 1  kK
3. f*p  0  pPk,  kK
Complementary slackness is satisfied
1. *ij(pPk  k K dk f*pijp - uij ) = 0,  ijA
2. *k (p Pk f*p – 1) = 0,  kK
Dual feasibility is satisfied (reduced cost is non-negative
for a minimization problem)
1. (dk cp +  ij A dk ij ijp ) - k = dk (  ij A (cijk + ij)
ijp - k /dk )  0,  p Pk,  k K
22-Jul-15
1.224J/ESD.204J
26
Multi-commodity Flow
Optimality Conditions
•
•
The price for an additional unit of capacity is 0
unless capacity is fully utilized
1. *ij(pPk  k K dk f*pijp - uij ) = 0,  ijA
A path p for commodity k is utilized only if its
“modified cost” (that is, ijA (cijk + *ijijp)) is
minimal, for all paths pPk
1. Reduced Costs all non-negative:
c’p = dk (  ij A (cijk + *ij) ijp - *k /dk )  0,
 p Pk,  k K
2. f*p (ijA (cijk + *ij ) ijp - *k /dk ) = 0,
 p Pk,  k K
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
27
Column Generation- A Price
Directive Decomposition
Millions/Billions of Variables
Restricted Master
Problem (RMP)
Never Considered
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
28
RMP and Optimality Conditions
Consider f*p and *ij , *k optimal for RMP, then
Primal feasibility is satisfied
1. pPk  k K dk f*pijp  uij  ijA
2. pP(k) f*p = 1  kK
3. f*p  0  pPk,  kK
Complementary slackness is satisfied
1. *ij(pPk  k K dk f*pijp - uij ) = 0,  ijA
2. *k (p Pk f*p – 1) = 0,  kK
Dual feasibility is guaranteed (reduced cost is nonnegative) ONLY for a path p included in RMP
1. (dk cp +  ij A dk ij ijp ) - k = dk (  ij A (cijk +
ij) ijp - k /dk )  0,  p Pk,  k K
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
29
LP Solution: Column Generation
• Step 1: Solve Restricted Master Problem (RMP) with
subset of all variables (columns)
• Step 2: Solve Pricing Problem to determine if any
variables when added to the RMP can improve the
objective function value (that is, if any variables
have negative reduced cost)
• Step 3: If variables are identified in Step 2, add
them to the RMP and return to Step 1; otherwise
STOP
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
30
Pricing Problem
• Given ,the optimal (non-negative) duals for
the current restricted master problem,the
pricing problem, for each p Pk, k K is
min p Pk (dk (  ij A (cijk + ij) ijp - k /dk )
Or, equivalently:
min p Pk  ij A (cijk + ij) ijp
 A shortest path problem for commodity k (with
modified arc costs)
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
31
Example- Iteration 1
Path
k=1
k=2
k=3
k=4
RHS
Dual
a
5
0
15
15
0
0
0
0
<= 20
a= 0
b
0
5
0
0
15
0
0
0
<= 10
b= 0
c
5
0
15
0
0
5
0
0
<= 20
c= 0
d
0
0
0
15
0
0
5
0
<= 10
d= 0
e
0
0
15
0
15
5
0
10
<= 40
e= 0
k=1
1
1
=1
 = 10
=1
 = 135
=1
 = 20
=1
 = 50
k=2
1
1
1
k=3
1
1
k=4
1
Cost.
20
Variable
f1
7/22/2015
10
f2
=1
135
f3
=1
75
f4
105
40
f5
f6
20
f7
=1
Barnhart 1.206J/16.77J/ESD.215J




50
f8
=1
32
Example- Iteration 2
Path
k=1
k=2
k=3
k=4
RHS
Dual
a
5
0
15
15
0
0
0
0
<= 20
a= 0
b
0
5
0
0
15
0
0
0
<= 10
b= 2
c
5
0
15
0
0
5
0
0
<= 20
c= 0
d
0
0
0
15
0
0
5
0
<= 10
d= 4
e
0
0
15
0
15
5
0
10
<= 40
e= 0
k=1
1
1
=1
 = 20
=1
 = 135
=1
 = 40
=1
 = 50
k=2
1
1
1
k=3
1
1
k=4
1
Cost.
20
Variable
f1
7/22/2015
10
f2
=1
135
f3
=1/3
75
f4
=
1/3 f 5
105
=1/3
40
f6
20
f7
=1
Barnhart 1.206J/16.77J/ESD.215J




50
f8
=1
33
MCF Optimality Conditions
•
For each pPk, for each k, the reduced cost cp:
– cp (dk cp +  ij A dk ij ijp ) - k = ij (dkcijk + dkij)ijp k = ij (cijk + ij)ijp - k /dk  0
•
where , are the optimal duals for the current restricted
master problem
– cp 0,for each utilized path p implies
ij (dkcijk + dkij) ijp = k
or equivalently,
ij (cijk + ij) ijp = k/dk
– So if, minpP(k) cp = ij (cijk + ij) ijp* - k/dk  0,the
current solution to the restricted master problem is
optimal for the original problem
– If minpP(k) cp = ij (cijk + ij) ijp* - k/dk <0,add p* to
restricted master problem
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
34
• Data Set
Data Set
Nodes
807
Links
1,363
capacitated
292
uncapacitated
1,071
O/D
17,539
# Origin
136
• Constraint Matrix Size
Improvement
Node_Arc
Path
Sub-network
7/22/2015
row
column
new_row
14,155,336
23,905,657
-
18,902
-
17,832
1,499
-
428
Barnhart 1.206J/16.77J/ESD.215J
35
Computational Results
• Number of Nodes: 807
• Number of Links: 1,363
• Number of Commodities: 17,539
• Computational Result (IBM RS6000, Model
370)
– Path Model: 44 minutes
– Sub-network Model: < 1 minute
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
36
Conclusions I
• Choose your formulation carefully
– Trade-off memory requirements and solution time
– Sub-network formulation can be effective when
low level of congestion in the network
• Problem size often mandates use of combined
column and row generation
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
37
Conclusions II
• Solution time is affected dramatically by
– The complexity of the pricing problem
– Exploitation of problem structure, preprocessing, LP solver selection, etc.
7/22/2015
Barnhart 1.206J/16.77J/ESD.215J
38