Graph Transformations for Vehicle Routing and Job Shop Scheduling Problems J.C.Beck, P.Prosser,...
-
Upload
stephanie-kirby -
Category
Documents
-
view
214 -
download
0
Transcript of Graph Transformations for Vehicle Routing and Job Shop Scheduling Problems J.C.Beck, P.Prosser,...
Graph Transformations for Graph Transformations for Vehicle Routing and Job Shop Vehicle Routing and Job Shop
SchedulingScheduling Problems Problems
J.C.Beck, P.Prosser, E.Selensky
[email protected], {pat,evgeny}@dcs.gla.ac.uk
ICGT 2002, E. Selensky 2
w1
w2
w12
wn
wi
wn-1
w1,n
wn-1,nw1,n-1
w2,n
w2,n-1
Find a cycle of min cost
Basic Problem
ICGT 2002, E. Selensky 3
Lexicographic ordering of nodes: A,B,C,D
Example
ICGT 2002, E. Selensky 4
Motivation
• Core problem in vehicle routing and shop scheduling
• Edge weights to node weights:– Large for VRP, small for JSP
• Can we use graph transformations to make VRP look like JSP and vice versa?
ICGT 2002, E. Selensky 5
Vehicle Routing
[2:25pm 2:40am]
[9:00am 9:15am]
[3:00pm 5:00am]
[3:00pm 5:00am]
[9:00am 5:00am]
[4:00pm 5:00am]
NP-hard!Go find vehicle tours with min travel
ICGT 2002, E. Selensky 6
Job Shop Scheduling
J1: (M1,t11) (M3,t13) (M2,t12)J2: (M3,t23) (M1,t21) (M2,t22)J3: (M2,t32) (M3,t33) (M1,t31)
3 machines: M1, M2, M3
3 jobs: J1, J2, J3
Go find a schedule with min Makespan NP-complete
TimeMakespan0
M1
M2
M3
ICGT 2002, E. Selensky 7
Hypothesis
Graph Transformation
VRP Solver
JSP Solver
Graph Transformation
VRP
JSP
Is it important?
ICGT 2002, E. Selensky 8
Cost-Preserving Transformations
• Assumptions:– Graphs: complete (true for VRP, JSP subsumed),
undirected (directed case subsumed);
– A solution is a cycle on the graph (for Hamiltonian paths everything is similar);
– Transformations should preserve cost and order of nodes in a cycle.
ICGT 2002, E. Selensky 9
Caveat
• This is not a comprehensive study of all possible transformations
• Rather, we propose some transformations and study them
ICGT 2002, E. Selensky 10
Types of TransformationsDirect: Reduce Edge Weights, Increase Node Weights
Inverse: Increase Edge Weights, Reduce Node Weights
ICGT 2002, E. Selensky 11
• lexicographic order of nodes
• choose a node whose cheapest incident edge is a maximum
• choose a node whose cheapest incident edge is a minimum
Order Dependent Transformations
MaxMin:
MinMin:
Lex:
ICGT 2002, E. Selensky 12
Example
Order Independent Transformation
ICGT 2002, E. Selensky 13
Inverse TransformationReminder: Increase Edge Weights, Reduce Node Weights
• Order-independent• GG’inv; GG’dodG’inv; GG’doiG’inv;
Express as if odd and if even 12 kwiiw kwi 2
ICGT 2002, E. Selensky 14
• Weight transfer from nodes to edges:– change in proportion of weight of cycle C:
– a similar measure for the whole graph:
where W and W’ are graph weights before and after transformation
Performance measures
iji
ijijc ww
ww
,
,W
wij ,W
wij
ICGT 2002, E. Selensky 15
• Relative edge/node weights ordering:– Sort edge/node weights in ascending order:
• e.g. {w11, w12, w13} for edges (1,1), (1,2) and (1,3);
– Apply transformations and count how many pair-wise changes there are:
• e.g. {w’13, w’11, w’12}, so we have 2 changes;
• Two measures: and edges .nodes
Performance measures
ICGT 2002, E. Selensky 16
Experiments
• Purpose: – Assess performance of the transformations on complete
undirected graphs
• Layout: – Randomly generate 100-instance sets of graphs of
different sizes; – Apply andMaxMin, MinMin,Lex, DirOrderInd
Inverse.
ICGT 2002, E. Selensky 17
Experiments
iji
ijijc ww
ww
ICGT 2002, E. Selensky 18
Experiments
W
wW
w
W
w
ij
ijij
ICGT 2002, E. Selensky 19
Experiments
ICGT 2002, E. Selensky 20
Experiments
ICGT 2002, E. Selensky 21
Analysis of Results
• Weight Transfer: Inverse >> Order Independent >> Order Dependent
• Changes in Edge/Node Ordering:Inverse: constant w.r.t. graph size;
Inverse>>MaxMin >> Order Independent, Lex >> MinMin
ICGT 2002, E. Selensky 22
Future Work
• Systematically apply the transformations to VRP/JSP instances and study their performance in practice.