Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Exploiting Network Structures on Transportation and Logistics
Problems August 9, 2013
Andrés Medaglia, Ph.D. Associate Professor
Industrial Engineering Department Centro de Optimización y Probabilidad Aplicada
(http://copa.uniandes.edu.co)
Pan-American Advanced Studies Institute (PASI) on Modeling, Simulation and Optimization of Globalized Physical Distribution Systems
Santiago (Chile)
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 2
Andrés L. Medaglia Associate Professor of Industrial Engineering at Universidad de los Andes (Bogotá, Colombia) Director of Centro para la Optimización y Probabilidad Aplicada (COPA – http://www.copa.uniandes.edu.co) Ph.D. in Operations Research (OR) from North Carolina State University (2001). Since 1999 and until the completion of his postdoctoral fellowship, he worked as an optimization specialist
developing Web-based decision support systems for the Supply Chain Center at SAS Institute Inc. In 2002, he joined the Industrial Engineering Department at Universidad de los Andes (ABET accredited
program). His current research interests include the development and application of optimization techniques to
transportation and logistics, project selection, and engineering design. His research has been published in Annals of Operations Research, Automation in Construction, Building
and Environment, Computers & Operations Research, Computers & Structures, Decision Support Systems, Engineering Applications of Artificial Intelligence, European Journal of Operational Research, Fuzzy Sets and Systems, Insurance: Mathematics and Economics, Interfaces, International Journal of Production Economics, Journal of Heuristics, Operations Research Letters, Optimization Letters, Socio-Economic Planning Sciences, The Engineering Economist, and Transportation Science, among others.
He has served as Secretary and Vicepresident of the Latin-Ibero American Association of Operations Research (ALIO); and as Vicepresident of Central/South America for the Institute of Industrial Engineers (IIE).
He currently serves the editorial board of the Journal of Industrial and Management Optimization. He has been the recipient of several prizes, most recently, the first prize in the 2011 INFORMS Railway
Application Section (RAS) Problem Solving Competition (with L. Lozano and J. González). More information at http://wwwprof.uniandes.edu.co/~amedagli.
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 3
Agenda
Part 0: A primer on (underlying) networks Part 1: Bus rapid transit (BRT) route design
Part 2: Split
Part 3: Pulse Algorithm
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 4
Agenda
Part 0: A primer on (underlying) networks Part 1: Bus rapid transit (BRT) route design
Part 2: Split
Part 3: Pulse Algorithm
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Production planning Sailco Corporation must determine how many sailboats should be
produced during each of the next four quarters. The demand during each of the next four quarters is as follows: first quarter, 40 sailboats; second quarter, 60 sailboats; third quarter, 75 sailboats; fourth quarter, 25 sailboats. Sailco must meet demands on time. At the beginning of the first quarter, Sailco has an inventory of 10 sailboats. At the beginning of each quarter, Sailco must decide how many sailboats should be produced during that quarter. For simplicity, we assume that sailboats manufactured during a quarter can be used to meet demand for that quarter. During each quarter, Sailco can produce up to 40 sailboats with regular-time labor at a total cost of $400 per sailboat. By having employees work overtime during a quarter, Sailco can produce additional sailboats with overtime labor at a total cost of $450 per sailboat. At the end of each quarter, a holding cost of $20 per sailboat is incurred. Determine a production schedule to minimize the sum of production and inventory costs during the next four quarters. 5
Winston (1991)
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Production planning
6
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Production planning
7
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Production planning
8
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Production planning as a Minimum Cost Network Flow (MCNF) Problem
9
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
s.a.,
Minimum Cost Network Flow (MCNF) Problem
10
Exploiting Network Structures on Transportation and Logistics
Part 1: Bus rapid transit (BRT) route design
August 9, 2013
11
Andrés Medaglia, Ph.D. Associate Professor
Industrial Engineering Department Centro de Optimización y Probabilidad Aplicada
(http://copa.uniandes.edu.co)
Pan-American Advanced Studies Institute (PASI) on Modeling, Simulation and Optimization of Globalized Physical Distribution Systems
Santiago (Chile)
Joint work with: Michel Gendreau (CIRRELT, Canada), Dominique Feillet (EMSE, France),
Jose L. Walteros (U. Florida), Jaime E. González (U. de los Andes),
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Bus Rapid Transit Systems (BRT)
Bus Rapid Transit (BRT) systems are urban transportation solutions that combine buses, stations, dedicated lanes, and intelligent transportation systems.
Performance of a BRT system is equivalent to a railway system, but its cost is lower (Hodgson et al., 2013).
Key characteristics: Buses can pass other buses. Express routes stop at fewer stations. Route transfers are allowed (and
some times needed).
12
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Worldwide growth of BRT systems
13 13
120 cities 280 corridors 4,300 km 6,700 stations 30,000 buses 28,000,000
passengers/day
Source: Hidalgo, D., & Gutiérrez, L., BRT and BHLS around the world: Explosive growth, large positive impacts and many issues outstanding, Research in Transportation Economics (2012)
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Bogotá’s BRT System: TransMilenio
14
Bogotá has a population of nearly 7 million people. TM currently has 87km, 115 stations, and 90 routes. TM transports 186.124 persons in a peak hour (September, 2012). Fleet of 1,392 buses. Average speed: 26 km/h. Travel times have been reduced by 32%.
14
Transmilenio S.A.(2013)
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
BRT - Problem Hierarchy
Road/Station design: Identifies roads and station locations.
Route design: Determines the stopping sequence for
each route. Dispatching:
Determines dispatching frequencies and timetable.
Fleet assignment: Assigns vehicles to the timetable.
Rostering: Assigns drivers to vehicles.
15
Ceder and Wilson (1986)
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Road/Station design: Identifies roads and station locations.
Route design: Determines the stopping sequence for
each route. Dispatching:
Determines dispatching frequencies and timetable.
Fleet assignment: Assigns vehicles to trips.
Rostering: Assigns drivers to vehicles.
16
BRT - Problem Hierarchy
Ceder and Wilson (1986)
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
BRT Route Design Problem (BRTRDP)
Finding a manageable set of routes that minimizes total passenger travel time while satisfying system technical constraints (OD matrix, min/max bus frequencies), without overcrowding the network (stations, buses, and lanes).
17
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
BRTRDP: concepts
18
Bus corridor: series of physically connected bus lanes holding a set of adjacent stations. Route: a series of stops (at stations) along a bus corridor.
Express Routes
Regular Routes
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 19
BRTRDP: Naïve approach Decomposition:
Route design Frequency determination
Key route design variable:
:1, if stations i and j along corridor c are directly connected by route k; 0, otherwise.
Many other variables: Route-corridor relations Cummulative distance per route Passengers served directly by a route Passengers served with transfers. Routes with common stations (for possible transfers) …
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 20
BRTRDP: Naïve approach
Instance Instance TM / Phase I TM (circa 2004)
Stations 60 78 Corridors 7 10
Routes 12 13 Variables 530,771 1,049,574
Constraints 1,535,694 3,050,502
Arana, Medaglia & Palacios (2004)
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Hybrid Algorithm for Route Design on BRT Systems
21
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 22
Network Model: Four-station Example
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Model: Three-station Example
23
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Model: Three-station Example
24
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Model: Three-station Example
25
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Model: Three-station Example
26
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Model: Three-station Example
27
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Model: Three-station Example
28
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 29
Network Formulation
Sets
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 30
Network Formulation
Parameters
Variables
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 31
Minimize travel time
Balance / flow constraints
Bus capacity constraints
Manageable set of routes
Network Formulation
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
The set of possible routes is huge. The nodes are proportional to the number of routes. There is a binary variable for every possible route. The binary variables “destroy" the network structure
32
Network Formulation: challenges
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
33
Network Formulation: solution approach
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
34
Network Formulation: solution approach
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 35
Network Formulation: solution approach
1 0 1 1 0 1 0 0 0 0 1 1 0 1 0 1
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Computational Results: instances
36
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Computational Results: instances
37
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Computational Results: instances
Problem size and LR information for the 14 Instances
38
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Computational Results: instances
Performance of HGA and comparison with the LR
39
17,561.27
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 40
Network Formulation: solving the relaxation
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Simultaneous column and cut generation (SCCG).
41
Network Formulation: solving the relaxation
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 42
Network Formulation: solving the LR with SCCG
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co 43
Network Formulation: solving the LR with SCCG
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Formulation: solving the LR with SCCG
González et al. (2013)
44
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
Network Formulation: solving the LR with SCCG
González et al. (2013)
45
Andrés Medaglia, Ph.D. Departamento de Ingeniería Industrial
http://copa.uniandes.edu.co
BRTRDP: open problems
46
Integrality of the route choice variables (𝑦𝑘) Benders decomposition Matheuristics
Nonlinearity arising from the interrelation of the bus
frequencies and waiting times.
Dealing with the re-design problem using key variables such as congestion.