Scribd
Upload
8 views34 pages

Class 11 - 2016

Uploaded by

Jiawei Huang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
8 views34 pages

Class 11 - 2016

Uploaded by

Jiawei Huang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 34

IELM 2410

Logistics and Freight Transportation Operations

Chung-Yee Lee

Freight Forwarder

Lecture 11

1
Recall: Nonvessel-operating Common Carrier
(NVOCC)

Serves primarily as wholesales of ocean


vessel capacity

Booking large blocks of container space

Selling these out in small quantities to


shippers.

2
Direct Customer vs. Freight Forwarder

Freight Forwarder:
Nonvessel-operating common carrier (NVOCC)

Shipper Type
Asia /Europe trade: 70% Freight Forwarder
(DHL, Kuehne + Nagel and Schenker are strongest
in Asia/Europe sale)

Transpacific: 70% Direct Customer


3
*: source: http://en.wikipedia.org/wiki/Shipbuilding#cite_note-11
Freight Forwarders

Provide a wide range of freight services


• Charge on every service provided

Earn profit from:


• Price differential
• Consolidation (Recall: LTL: Less than Truck Load)
Freight Forwarder’s Profit from
Consolidation

How to Consolidate?
Consolidation Problem

Order # Shipper’s Volume Weight Destination Pick-up Latest


name (m3) (kg) date delivery
date
1 WalMart 10 2000 Chicago 1/5 3/5
2 Target 5 500 San 28/4 3/5
Francisco
3 Ikea 13 3500 New York 30/4 2/5
4 Gap 3 200 Los 2/5 4/5
Angeles
…… ….. …… …… …… …… ……
Mathematical Modeling
Mathematical model
• A simplified representation of a system using
mathematical language
• Aim: help us understand the underlying system by
simplifying it
• Tools for analysis: “standard” mathematics

Important:
• Most models may be over simplified.
• But some are useful
Simplified Representation
• Keep features that help our understanding
• Remove unnecessary details
– What are “unnecessary”? Based on skills and
experience
Very Simple Mathematical
Model
What questions do we want to
answer?

What are necessary features?

What are unnecessary details?

How to use mathematics to


represent this “system”?

What “mathematical tools” to use?


Back to Consolidation Problem
Question to Answer:
• How to use minimum number of containers?

Necessary details:
• Want to pack a number of items into some number
of containers
• Use more containers = more expensive
• Make sure shipments packed into a container <=
capacity of container
Consolidation Problem
Unnecessary details:

• Names of the shippers


• Exact time of arrival (to some extent)
• Exact type of items (to some extent)
• Weight of cargo (for light cargo)
Consolidation Problem
Simplify the problem: volume is necessary, weight is not

capacity of a container= W units

... ...

w1 w2 wM-1 wM
M orders

What is the best way to consolidate goods?


Best = using fewest containers
Objective, Variables & Constraints
1st container ... ...
nth container

... ...
w1 w2 wM-1 wM

• Objective: minimize # of containers


used
• Decision variables: how to consolidate
• Constraints
– Cannot violate container capacity
– An order will not be split
Bin-Packing Problem

Giving M orders with size wi, i =1,…,M, and container size


W, we try to find a minimum number of containers that can
fit all M orders.
Solution Method
Exact method

• Software required
• Difficult
• More in future courses

Heuristic method
Idea for Heuristic
• Idea: combine orders so that the
utilization of each container is close to W

W = 10

2 2 2 2 2

7 7 7 7 7
Modified Heuristic: Next Fit

Start a new box


1: Start a new container

2: Keep packing the container until


either all orders are completed,
or the container cannot accommodate the next
order

3: Stop if all orders are packed;


else go to step 1
Heuristic: Next Fit (NF)
Initially w1 Container 1

w2 Container 1

… … …

w4 Container 1

use a new container when container 1


cannot accommodate w4
Next Fit (NF): Limitationswill not be
In general, when considered
filling up container j
Container
1
.. .. ..
if container j
has enough Container
j-1
capacity
Container
j
wi

Container j+1
start a new container
otherwise;
will not consider
containers 1 to j again waste! may have enough capacity for wi
in containers 1 to j-1
Modified Heuristic: First Fit
Start a new box
1: Start a new container

2: Keep packing the container until


• either all orders are completed
• or the box cannot accommodate the next order

3: Stop if all orders are packed;


• Else
• if possible, put the order in the 1st non-full container
that fits
• else start a new container; Go to Step 3
Heuristics: First Fit (FF)
Initially w1 Container 1

w2 Container 1

… … …

w4 Container 1

use a new container when container1


cannot accommodate v4
Heuristics: First Fit (FF)
Container 1

put in the 1st .. .. ..


container that has
enough capacity Container j-1

wi Container j

Container j+1
start container j+1 only if the
residual capacity of containers
1 to j are too small
Bad if large volume appears late; e.g.,
W = 9;
{vi} = {3, 3, 3, 3, 6, 6, 6, 6}
Modified Heuristic: First Fit Decreasing

Start a new box

1: Sort the orders in non-


increasing order of their size

2: Apply FF to the problem


More example
wi : 8, 2, 5, 3, 4, 9, 11
W =14.
NF
Container 1: 8,2,
Container 2: 5,3,4
Container 3: 9
Container 4: 11

FF Container 1: 8,2,3 ; Container 2: 5,4 ; Container 3: 9


Container 4: 11
FFD:
Container 1: 11,3
Container 2: 9,5
Container 3: 8,4,3
Only Heuristic; may not be optimal
For example, 7 orders with size
= (3,3,2,2,2,2) and W = 7.

Container 1: 3,3
Container 2: 2,2,2  N = 3
Container 3: 2.

Optimal solution:
Container 1: 3,2,2  N* = 2
Container 2: 3,2,2
How to Evaluate Heuristics?
One possible way: use simulation
• Compare NF,FF and FFD
• Randomly generate many samples of orders (or use
past data)
• Find the # of containers used in each sample under
NF,FF and FFD
• Find the average # of boxes used in NF, FF & FFD
• See which one is smaller

Other approaches: derivation from bounds,


asymptotic behavior, worst case analysis
Variations of Consolidation Problem
Unpredictable orders
• P(vi = 2) = 0.3; P(vi = 3) = 0.2; P(vi = 6) = 0.5
• Orders may not show up as promised

Multi-dimensional bin packing


• 3D packing limitations (shape)

Multi-attribute bin packing:


• E.g., both volume and weight constraints

Accepting and rejecting orders


• Given the time to cut off and the capacity left
Another Important Problem

Knapsack Problem
Knapsack Problem

Knapsack Problem:
n
max ∑ x i v i
i =1
n
s.t. ∑ x i wi ≤ W
i =1

x i = 0 or 1

There are n items with size wi and value vi , i = 1,…,n, and


a knapsack with capacity W. We wish to pack items
into the knapsack so that total value in the knapsack is
maximized.
Knapsack Problem: Example
W = 10
(v1, v2, v3, v4, v5 ) = (2,5,3,1,4)
(w1, w2, w3, w4, w5 ) = (4,3,2,6,1)
If (x1, x2, x3, x4, x5 ) = (1,0,1,1,1)
 Σ xiwi = 4+2+6+1 > 10  Infeasible
(x1, x2, x3, x4, x5 ) = (1,0,1,0,1)  Σ xiwi = 7, Σ xivi = 9
(x1, x2, x3, x4, x5 ) = (1,1,1,0,1)  Σ xiwi = 10, Σ xivi = 14
(x1, x2, x3, x4, x5 ) = (1,0,0,1,0)  Σ xiwi = 10, Σ xivi = 3
…….
Try to find a feasible solution with maximum objective
value
Knapsack Problem: Heuristic Methods

LV(Largest Value): Choose the largest value unassigned


item as a candidate to add to the knapsack.

LR(Largest Ratio): Choose the largest ratio unassigned


item as a candidate to add to the knapsack.
Knapsack Problem: Example

W = 10
(v1, v2, v3, v4, v5 ) = (2,5,3,1,4)
(w1, w2, w3, w4, w5 ) = (4,3,2,6,1)

LV: (x2, x5, x3, x1, x4 ) = (1,1,1,1,0)


 Σ xiwi = 10, Σ xivi = 14
LR: (x5, x2, x3, x1, x4 ) = (1,1,1,1,0)
 Σ xiwi = 10  Σ xivi = 14
How good the performance of LR, and/or LV?
Theoretically. either one can perform very poorly in
some instance.
Knapsack Problem
(1) Consider an instance with n = M jobs,
v1 = 2, w1= M,
wi = vi = 1 for i = 2,…,M.
W = M.
 Z* = M-1. However, use LV we will only put J1 in and
ZLV = 2. Thus Z*/ZLV  ∝ when n  ∝.
(2) Consider an instance with 2 jobs,
v1 = 2, w1= 1,
v2 = M, w2 = M
W = M.
 Z* = M. However, use LR we will only put J1 in and ZLR = 2.
Hence Z*/ZLR can be arbitrarily large when M is very large.
Knapsack Problem: Compound

COMPOUND: Apply LR and LV and choose the


better one. ZCOM = max{ZLV, ZLR}.

Theorem 4: Z*/ZCOM ≤ 2

You might also like