A technique for solving stochastic programs with first stage integer variables

1
A technique for solving stochastic programs with
first stage integer variables

• Chandra A. Poojari
• Gautam Mitra

2
Outline

• Review of current solution approaches for Integer
  SP.
• Discussion of our approach.
• Analysis of Bounds.
• Preliminary computational results.

3
New Perspective of Investment Decisions

• The issues
• DCF is inadequate
• Three leading characteristics
• Investment Decisions (costs) irreversible
• Future returns are uncertain
• Another key aspect is timing
• Invest
• Disinvest
• Not investpostpone
• Are all strategic decisions

4
Decision making under uncertainty

• Buy flexibility
• ?
• Hedge against uncertainty
• ?
• Make robust decisions

5
General two-stage SP
Subject to
let
Subject to
6
Shape of the recourse function
7
Stochastic integer programs
Applications

• First Stage integer
• Supply chain planning, Mitra et al.
• Asset allocation, Mulvey and Rusczynski
• Goods distribution, Cheung and Powell
• Second Stage integer
• Scheduling decisions, Dempster et al.
• Electric utility planning, Bienstock and Shapiro
• Routing decisions, Spaccamela et al.
• Fixed-charge, Bitran et al.
• Change over costs, Caroe et al.

8
Current approaches

• Cutting plane - Wollmer
• Scenario decomposition- Rockafellar and Wets
• Integer L-shaped Laporte and Louveaux
• Convex hull for Simple integer recourse
  Haneveld, Stougie, Van der Vlerk

9
Current approaches

• Progressive hedging - Takriti et al.
• Groebner bases Schultz et al.
• Integer L-shaped Caroe and Tind
• Dual decomposition Caroe and Schultz
• Branch and bound based approach Ahmed et al.

10
Problem Statement
Subject to
11
Motivation for the algorithm

• A Strategic Supply Chain model

12
Assumptions

• Relatively complete recourse
• Discrete distribution

13
The Approach

• Lagrangean relaxation used to generate a
  candidate set of integer feasible solutions for
  each WS model.
• Construct a convex hull of this set of integer
  feasible solutions.
• Not necessarily a convex hull to the original
  2-stage ISP.
• We compute bounds to the 2-stage SP model and
  qualify the approximated model.

14
Algorithm
15
Algorithm
16
Algorithm
LR
fix
in
Sub-gradient optimisation
17
Algorithm
Stopping Criteria
1.
Pass ? Maximum Number of Iterations
2.
3.
Satisfaction of the relaxed constraints,
18
The Approximated model
APXP2SP
19
Bounds
Upper Bound
Lower Bound
Then we have
20
Qualifying the bound
?
The ratio
measures the compactness of the bound
21
Computational Results
Model description
SCP-1 Strategic supply chain, opening and
closing of sites, DCs,lines. SCP-2 Trade-off
between capacity acquisition and maintaining
inventory.
Model Statistics
22
(No Transcript)
23
Computational Results
Generation of integer solutions
24
Computational Results
Generation of integer solutions
Comparison of computational time
25
Outstanding Issues

• Step-size in subgradient optimisation.
• The constraints to be relaxed.
• Parallelisation of the algorithm.
• Study the performance of the algorithm to SP test
  set.

26
Thank you !