California Greyhound Bus Network LT Catherine Long LT James Koffi
Dec 28, 2015
California Greyhound Bus Network
LT Catherine LongLT James Koffi
Agenda
Background Problem/Model Scenario Network Model Interdiction Takeaways/Further Research Summary/Conclusions Recommendations
Greyhound Bus Network
Greyhound serves more than 3,800 destinations in North America and employs around 7,800 staff.
Greyhound operated nearly 5.5 billion passenger miles last year.
The U.S. operation, as well as its operating subsidiaries and Greyhound Canada, carried nearly 18 million people.
Greyhound Bus Network
Intercity bus is the safest mode of transportation over cars, trucks, trains, planes and other commercial vehicles, according to the U.S. Department of Transportation
The Greyhound active fleet consists of about 1,775 buses, with an average age of 9.4 years.
One Greyhound bus takes an average of 19 cars off the road, and achieves 170 passenger miles per gallon of fuel.
Real World
Inside New Greyhound Bus
YouTube Video
http://www.youtube.com/watch?v=M_0Gz-Ie8T0&feature=channel&list=UL
Problem
The price of gas is expensive, and routes between stations vary throughout the state
New buses save gas and will keep up with the competition (Mega Bus, etc.)
What is the fastest and most cost effective way to transition 50 new express buses from two major Greyhound stations to other stations around the state of California?
Model Scenario
Model distributes 50 new buses throughout CA
Use GAMS to model the flow of buses to new stations and demonstrate how an “attack,” such as a major earthquake, impacts the network
The “attacks” model the worst case scenario and completely disable edge
Nodes and Arcs
The nodes in model represent stations
The arcs represent routes from one station to another
Route information is from the Greyhound website
Supply nodes are Oakland and Los Angeles
Road Map of California
Network Model
S
T
Network Model
S
Oakland
Los Angeles
T
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Measure of Effectiveness
The measure of effectiveness (MOE) for this model is to minimize the total cost of the flow of buses to their destination cities
Cost = $/gallon x miles per edge Each city besides the supply cities receives
2 brand new buses
Model Assumptions
Assume all new Greyhound buses get 6.5 miles per gallon
Cost per gallon of diesel is $4.00 Edge capacities are never exceeded
in our model
Greyhound Network Restrictions Many cities are connected by narrow two way roads which are
prohibited for use by Greyhound or any bus/motorhome in CA due to handling issues.
Bus: California Vehicle Code (CVC) Section 233 defines "bus" as: "(a) …any vehicle… …designed, used, or maintained for carrying more than 15 persons including the driver”
This is one of the major limitations to the network.
Operator’s/Attacker’s Problem
The model is run without interdiction to establish a baseline total minimum cost of $3,246.80
Run the model with increasing attacks to determine network resiliency
GAMS chooses routes that will maximize the total cost after each “attack”
Network ModelWith 1 Attack
S
Oakland
Los Angeles
T
Cost = 12624.404 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 2 Attacks
S
Oakland
Los Angeles
T
Cost = 17410.56 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 3 Attacks
S
Oakland
Los Angeles
T
Cost = 22061.208 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 4 Attacks
S
Oakland
Los Angeles
T
Cost = 31341.6012 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 5 Attacks
S
Oakland
Los Angeles
T
Cost = 36127.714 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 6 Attacks
S
Oakland
Los Angeles
T
Cost = 36581.8014 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 7 Attacks
S
Oakland
Los Angeles
T
Cost = 40844.8016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 8 Attacks
S
Oakland
Los Angeles
T
Cost = 41639.9016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 9 Attacks
S
Oakland
Los Angeles
T
Cost = 42023.9016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 10 Attacks
S
Oakland
Los Angeles
T
Cost = 42362.3016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 11 Attacks
S
Oakland
Los Angeles
T
Cost = 43136.5016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 12 Attacks
S
Oakland
Los Angeles
T
Cost = 43146.3016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Analysis
Can model resiliency be improved with an additional supply node in Bakersfield?
Yes, but it only changes the model slightly…
Network ModelWith 1 Attack
S
Oakland
Los Angeles
T
Cost = 12292.104 buses unable to reachOceanside and San Diego
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 2 Attacks
S
Oakland
Los Angeles
T
Cost = 17078.204 buses unable to reachOceanside and San Diego
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 3 Attacks
S
Oakland
Los Angeles
T
Cost = 21728.906 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 4 Attacks
S
Oakland
Los Angeles
T
Cost = 31009.3012 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 5 Attacks
S
Oakland
Los Angeles
T
Cost = 35795.3014 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 6 Attacks
S
Oakland
Los Angeles
T
Cost = 40512.5014 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 7 Attacks
S
Oakland
Los Angeles
T
Cost = 40512.5016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 8 Attacks
S
Oakland
Los Angeles
T
Cost = 40870.6016 buses unable to destination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 9 Attacks
S
Oakland
Los Angeles
T
Cost = 41094.6016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 10 Attacks
S
Oakland
Los Angeles
T
Cost = 41401.1016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 11 Attacks
S
Oakland
Los Angeles
T
Cost = 41759.2016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 12 Attacks
S
Oakland
Los Angeles
T
Cost = 41996.8016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 13 Attacks
S
Oakland
Los Angeles
T
Cost = 42043.6016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 14 Attacks
S
Oakland
Los Angeles
T
Cost = 42053.4016 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Operator’s Resilience Curve
Results
The most costly “attacks” occur with the first attack and the fourth attack Attack #1: 12624.40 - 3,246.80 =
$9377.60 Attack #4: 31341.60 - 22061.20 =
$9280.40
Network ModelWith 1 Attack
S
Oakland
Los Angeles
T
Cost = 12624.40Difference 12624.40 - 3,246.80 =9377.604 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 3 Attacks
S
Oakland
Los Angeles
T
Cost = 22061.208 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Network ModelWith 4 Attacks
S
Oakland
Los Angeles
T
Cost = 31341.60Difference = 31341.60 - 22061.20 = 9280.4012 buses unable to reachdestination
Santa Rosa
Santa Cruz
San Francisco
Santa Maria
San Luis Obispo
King City
Salinas
Stockton
San Diego
Indio
Sacramento
Vallejo
Modesto
Merced
Fresno
Bakersfield
Palmdale
Oceanside
Victorville
Indio
RiversideOxnard
Coalinga Jct
Los Banos
Santa Jose
Model Comparison
The additional supply station in Bakersfield decreased overall cost by $1,093.00 after 14 attacks
In the scenario with 3 supply stations Bakersfield started out by supplying just 4 new buses
By the 14th attack, Bakersfield supplies 14 buses
The more supply nodes the better, however may not be cost effective for Greyhound to have so many bus contracts
Future Analysis/Recommendations
Change the probability of destruction based on location, i.e. if on a fault line
Additional routes could be added to local roads and streets that connect the major interstates to build more redundancy in the network, but must abide to California state laws
Future Analysis/Recommendations
Model improvements can be achieved by increasing the granularity of the model
Can adapt the model to include other costs Price-tiered system Discounts on certain routes Express routes
Summary/Conclusions
We conclude there is not enough resiliency in the Greyhound network to sustain affordable bus transition operations in the event of “attacks” on the network depending on budget
Model can be used as a tool to weigh cost benefits of going around edges caused by a major traffic accident, or waiting an additional day for it to clear
Questions?
Sources
http://www.greyhound.com/en/about/factsandfigures.aspx
http://articles.latimes.com/2012/may/03/news/la-trb-greyhound-express-20120503