Transportation II

Lecture 16
ESD.260 Fall 2003

Caplice
 One to One System
⎛ 1 + ns ⎞ ⎛d ⎞
LC ($ / item) = cr H MAX + ci H MAX + ci tm + cs ⎜ ⎟ + cd ⎜ ⎟ + cvs
Mode 1 ⎝ v ⎠ ⎝v ⎠
Cost per Item

citm
vMAX 2vMAX Shipment Size
MIT Center for Transportation & Logistics – ESD.260 2 © Chris Caplice, MIT
 One to Many System
Single Distribution Center:
• Products originate from one origin
• Products are demanded at many destinations
• All destinations are within a specified Service Region

Assumptions:
• Vehicles are homogenous
• Same capacity, vMAX
• Fleet size is constant

MIT Center for Transportation & Logistics – ESD.260 3 © Chris Caplice, MIT
 LCF for One to Many System
Cost Function
„ Storage (Rent) Holding Costs
„ Inventory Holding Costs
„ Transportation Costs
„ Handling Costs
Key Cost Drivers?
„

MIT Center for Transportation & Logistics – ESD.260 4 © Chris Caplice, MIT
 One to Many System
Finding the estimated total distance:
• Divide the Service Region into Delivery Districts
• Estimate the distance required to service each district

MIT Center for Transportation & Logistics – ESD.260 5 © Chris Caplice, MIT
 One to Many System

Route to serve a specific district:

• Line haul from origin to the 1st customer in the district
• Local delivery from 1st to last customer in the district
• Back haul (empty) from the last customer to the origin
MIT Center for Transportation & Logistics – ESD.260 6 © Chris Caplice, MIT
 An Aside: Routing & Scheduling
Problem:
„ How do I route vehicle(s) from origin(s) to destination(s) at a
minimum cost?
„ A HUGE literature and area of research
One type of classification (by methodology)
1. One origin, one destination, multiple paths
Shortest Path Problem
2. Single path to reach all the destinations
Minimum Spanning Tree
3. Many origins, many destinations, constrained supply
Transportation Method (LP)
4. One origin, many destinations, sequential stops -
Traveling Salesman Problem
Vehicle Routing Problem
„ Stops may require delivery & pick up
„ Vehicles have different capacity (capacitated)
„ Stops have time windows
„ Driving rules restricting length of tour, time, number of stops

MIT Center for Transportation & Logistics – ESD.260 7 © Chris Caplice, MIT
 One to Many System
Find the estimated distance for each tour, dTOUR
„ Capacitated Vehicle Routing Problem
„ Cluster-first, Route-second Heuristic

dTOUR ≈ 2d LineHaul + d Local

dLineHaul = Distance from origin to center of gravity of

delivery district
dLocal = Local delivery between c customers in district (TSP)

MIT Center for Transportation & Logistics – ESD.260 8 © Chris Caplice, MIT
 One to Many System
What can we say about the expected TSP
distance to cover n stops in district of area A?
„ Hard bound and some network specific estimates:

E [ dTSP ] ≤ 1.15 nA For n>25 over Euclidean space, k=.7124

For straight line (Manhattan Metric), k=.7650
E [ dTSP ] ≈ k nA

Density, δ, number of stops per area

Average distance per stop, dstop δ = n/ A
dTSP nA k
d stop = =k⋅ =
n n δ
MIT Center for Transportation & Logistics – ESD.260 9 © Chris Caplice, MIT
 One to Many System
Length of local tours
„ Number of customer stops, c, times dstop over entire
region
„ Exploits property of TSP being sub-divided –
Š TSP of disjoint sub-regions ≥ TSP over entire region

MIT Center for Transportation & Logistics – ESD.260 10 © Chris Caplice, MIT
 One to Many System
Finding the total distance traveled on all, l, tours:
ck
E [ dTOUR ] = 2d LineHaul +
δ
nk
E [ d AllTours ] = lE [ dTOUR ] = 2ld LineHaul +
δ
The more tours I have, the shorter the line haul distance
Minimize number of tours by maximizing vehicle capacity
+
⎡ D ⎤ [x]+ is lowest integer value
l=⎢ ⎥
v greater than x – a step
⎣ MAX ⎦
function
+
⎡ Q ⎤ nk Estimate this with continuous
E [ d AllTours ] = 2 ⎢ ⎥ d LineHaul + function:
⎣ vMAX ⎦ δ
E([x]+ ) ~ E(x) + ½
MIT Center for Transportation & Logistics – ESD.260 11 © Chris Caplice, MIT
 One to Many System
So that expected distance is:
⎡ E [ D] 1 ⎤ E [ n] k
E [ d AllTours ] = 2 ⎢ + ⎥ d LineHaul +
⎣ vMAX 2 ⎦ δ

Note that if each delivery district has a different density,

then:
⎡ E [ Di ] 1 ⎤ E [ ni ]
E [ d AllTours ] = 2∑ i ⎢ + ⎥ d LineHauli + k ∑ i
⎣ vMAX 2⎦ δi

For identical districts, the transportation cost becomes:

⎡ E [ D] 1 ⎤ ⎛ ⎡ E [ D] 1 ⎤ E [ n] k ⎞
TransportCost = cs ⎢ E [ n ] + + ⎥ + cd ⎜ 2 ⎢ + ⎥ d LineHaul + ⎟⎟ + cvs E [ D ]
vMAX 2 ⎦ ⎜ δ ⎠
⎣ ⎝ ⎣ vMAX 2 ⎦

MIT Center for Transportation & Logistics – ESD.260 12 © Chris Caplice, MIT
 One to Many System
Fleet Size
„ Find minimum number of vehicles required, M
„ Base on, W, amount of required work time
Š tw = available worktime for each vehicle per period
Š s = average vehicle speed
Š l = number of shipments per period
Š tl =loading time per shipment
Š ts = unloading time per stop

d AllTours
Mtw ≥ W = + ltl + nts
s
⎛ 2d LineHaul ⎞ ⎡ E [ D] 1 ⎤ ⎛ k ⎞
W =⎜ + tl ⎟ ⎢ + ⎥ + E [ n] ⎜ + ts ⎟
⎝ s ⎠ ⎣ vMAX 2 ⎦ ⎝ δ ⎠

MIT Center for Transportation & Logistics – ESD.260 13 © Chris Caplice, MIT
 One to Many System
Note that W is a linear combination of two random variables, n
and D

E[aX + bY ] = aE[ X ] + bE[Y ]

Var[aX + bY ] = a 2Var[ X ] + b 2Var[Y ] + 2abCov[ X , Y ]

Substituting in, we can find E[W] and Var[W]

Given a service level, SL

⎛ 2d LineHaul ⎞⎡ 1 ⎤
a=⎜ + tl ⎟ ⎢ ⎥
⎝ s v
⎠ ⎣ MAX ⎦ P[W<Mtw]=SL Thus,
⎛ k ⎞
b=⎜ + ts ⎟ M= (E[W] + k(SL) StDev[W])/tw
⎝ δ ⎠

MIT Center for Transportation & Logistics – ESD.260 14 © Chris Caplice, MIT