A Linear Programming Approach to the Train Timetabling Problem

Similar documents
Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition

A Lagrangian heuristic algorithm for a real-world train timetabling problem

Combining Constraint Programming and Integer Programming

The Minimum Universal Cost Flow in an Infeasible Flow Network

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

Fuzzy approach to solve multi-objective capacitated transportation problem

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming

Simultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals

Problem Set 9 Solutions

COS 521: Advanced Algorithms Game Theory and Linear Programming

SOLVING CAPACITATED VEHICLE ROUTING PROBLEMS WITH TIME WINDOWS BY GOAL PROGRAMMING APPROACH

Timing-Driven Placement. Outline

Managing Capacity Through Reward Programs. on-line companion page. Byung-Do Kim Seoul National University College of Business Administration

Valid inequalities for the synchronization bus timetabling problem

Global Optimization of Truss. Structure Design INFORMS J. N. Hooker. Tallys Yunes. Slide 1

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan

Coarse-Grain MTCMOS Sleep

An Admission Control Algorithm in Cloud Computing Systems

Solutions HW #2. minimize. Ax = b. Give the dual problem, and make the implicit equality constraints explicit. Solution.

Organizing Teacher Education In German Universities

Customer Selection and Profit Maximization in Vehicle Routing Problems

Advanced Topics in Optimization. Piecewise Linear Approximation of a Nonlinear Function

Embedded Systems. 4. Aperiodic and Periodic Tasks

BALANCING OF U-SHAPED ASSEMBLY LINE

Optimal Solution to the Problem of Balanced Academic Curriculum Problem Using Tabu Search

Combinational Circuit Design

On the Multicriteria Integer Network Flow Problem

A 2D Bounded Linear Program (H,c) 2D Linear Programming

Solutions to exam in SF1811 Optimization, Jan 14, 2015

Asymptotically Optimal Algorithms for Job Shop Scheduling and Packet Routing

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Computing Correlated Equilibria in Multi-Player Games

Hidden Markov Model Cheat Sheet

NP-Completeness : Proofs

Annexes. EC.1. Cycle-base move illustration. EC.2. Problem Instances

An Interactive Optimisation Tool for Allocation Problems

Xiangwen Li. March 8th and March 13th, 2001

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017

Amiri s Supply Chain Model. System Engineering b Department of Mathematics and Statistics c Odette School of Business

Optimal Dispatch in Electricity Markets

Portfolios with Trading Constraints and Payout Restrictions

Asymmetric Traveling Salesman Path and Directed Latency Problems

Covering Problems. Lawrence V. Snyder. Lehigh University

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control

Thin-Walled Structures Group

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES

EC3075 Mathematical Approaches to Economics

MLE and Bayesian Estimation. Jie Tang Department of Computer Science & Technology Tsinghua University 2012

Capacity Constraints Across Nests in Assortment Optimization Under the Nested Logit Model

Fuzzy Set Approach to Solve Multi-objective Linear plus Fractional Programming Problem

Technical Note: Capacity Constraints Across Nests in Assortment Optimization Under the Nested Logit Model

Hopfield Training Rules 1 N

Planning and Scheduling to Minimize Makespan & Tardiness. John Hooker Carnegie Mellon University September 2006

Application of Nonbinary LDPC Codes for Communication over Fading Channels Using Higher Order Modulations

find (x): given element x, return the canonical element of the set containing x;

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

A Search-Infer-and-Relax Framework for. Integrating Solution Methods. Carnegie Mellon University CPAIOR, May John Hooker

COURSE TIMETABLING AT A TUNISIAN UNIVERSITY: A CASE STUDY

Kernel Methods and SVMs Extension

Single-track train timetabling with guaranteed optimality: Branch-and-bound algorithms with enhanced lower bounds

A Simple Inventory System

Train Scheduling Networks under Time Duration Uncertainty

Digital PI Controller Equations

Multiobjective Course Scheduling with Mathematical Programming and Analytic Network Process. Müjgân Sağır.

Fitting a Graph to One-Dimensional Data. September 11, 2018

( ) 2 ( ) ( ) Problem Set 4 Suggested Solutions. Problem 1

The Multi-Inter-Distance Constraint

Module 9. Lecture 6. Duality in Assignment Problems

Revenue comparison when the length of horizon is known in advance. The standard errors of all numbers are less than 0.1%.

An Optimization Model for Routing in Low Earth Orbit Satellite Constellations

Pickup and Delivery with Time Windows : Algorithms and Test Case Generation

INTEGER NETWORK SYNTHESIS PROBLEM FOR HOP CONSTRAINED FLOWS

Some Notes on Consumer Theory

Some modelling aspects for the Matlab implementation of MMA

Lecture 10 Support Vector Machines II

xp(x µ) = 0 p(x = 0 µ) + 1 p(x = 1 µ) = µ

Module 2. Random Processes. Version 2 ECE IIT, Kharagpur

Outline and Reading. Dynamic Programming. Dynamic Programming revealed. Computing Fibonacci. The General Dynamic Programming Technique

Working Paper Series February Susan K. Norman. College of Business Administration. Box Flagstaff, AZ

Lecture 20: Lift and Project, SDP Duality. Today we will study the Lift and Project method. Then we will prove the SDP duality theorem.

Structure and Drive Paul A. Jensen Copyright July 20, 2003

An Integrated OR/CP Method for Planning and Scheduling

Combining Examinations to Accelerate Timetable Construction

Heuristic Algorithm for Finding Sensitivity Analysis in Interval Solid Transportation Problems

The Ideal Train Timetabling Problem

Natural Language Processing and Information Retrieval

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Dynamic Bid Prices in Revenue Management

ExxonMobil. Juan Pablo Ruiz Ignacio E. Grossmann. Department of Chemical Engineering Center for Advanced Process Decision-making. Pittsburgh, PA 15213

Linköping University Post Print. Solving a minimum-power covering problem with overlap constraint for cellular network design

CIE4801 Transportation and spatial modelling Trip distribution

Departure Process from a M/M/m/ Queue

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Min Cut, Fast Cut, Polynomial Identities

= z 20 z n. (k 20) + 4 z k = 4

BRANCH-AND-PRICE FOR INTEGRATED MULTI-DEPOT VEHICLE AND CREW SCHEDULING PROBLEM. 1. The Integrated Vehicle and Crew Scheduling Problem

How Strong Are Weak Patents? Joseph Farrell and Carl Shapiro. Supplementary Material Licensing Probabilistic Patents to Cournot Oligopolists *

Solution (1) Formulate the problem as a LP model.

A SEPARABLE APPROXIMATION DYNAMIC PROGRAMMING ALGORITHM FOR ECONOMIC DISPATCH WITH TRANSMISSION LOSSES. Pierre HANSEN, Nenad MLADENOVI]

A FAST HEURISTIC FOR TASKS ASSIGNMENT IN MANYCORE SYSTEMS WITH VOLTAGE-FREQUENCY ISLANDS

Transcription:

A Lnear Programmng Aroach to the Tran Tmetablng Problem V. Cacchan, A. Carara, P. Toth DEIS, Unversty of Bologna (Italy) e-mal (vcacchan, acarara, toth @des.unbo.t)

The Tran Tmetablng Problem (on a sngle one-ay track) ams at determnng an otmal tmetable for a set of trans hch does not volate track caacty constrants: a mnmum tme nterval (d) beteen 2 deartures s requred, a mnmum tme nterval (a) beteen 2 arrvals s requred, no overtakng s alloed beteen to consecutve statons.

Trans must be run every day of a gven tme horzon (e.g. 6-2 months). We are gven a so-called deal tmetable on nut hch s tycally nfeasble. To obtan a feasble tmetable to knds of modfcatons are alloed: change the dearture tme of some trans from ther frst staton (shft) and/or ncrease the mnmum stong tme n some of the ntermedate statons (stretch). Travel tme beteen to consecutve statons s assumed to be fxed for each tran.

Tmes are dscretzed and exressed as ntegers from to 440 (mnutes n a day). S ={,..., s} set of statons T ={,..., t} S = { f,..., l } set of trans set of statons vsted by tran The tmetable ndcates the deal dearture tme from f, the deal arrval tme n l and the deal arrval and dearture tmes for each ntermedate staton. f,..., l +

Each tran s also assgned an deal roft π deendng on the tye of the tran (eurostar, freght, etc). If the tran s shfted and/or stretched the roft s decreased. If the roft becomes null or negatve the tran s cancelled. π φ ( ) ν γ μ ν μ shft stretch (.e. sum of the stretches n all statons) The obectve s to maxmze the overall roft of the trans.

A grah formulaton G=(V,A) drected acyclc multgrah V set of nodes: { σ, τ} U ( U U... UU ) U ( W U... UW 2 s s ) U W Arrval nodes n staton Dearture nodes from staton A set of arcs: A t U... A U... U A To tyes of arcs: Segment arcs Staton arcs, T

tme staton σ de_node_ Tran f segment_arc_ arr_node_2 staton_arc_2 de_node_2 segment_arc_2 arr_node_3 staton_arc_3 de_node_3 segment_arc_3 arr_node_4 l τ

Dearture Constrants: de_ de_2 stat stat 2 Arrval Constrants: stat stat 2 arr_ arr_2 Overtakng Constrants: stat stat 2

Ideal ath Ne ath Stat deal_de_nst Shft Stat 2 sto Stat 3 sto Stretch Stat 4

Tmetable for tran Feasble ath from σ to τ usng arcs n A Overall tmetable Set of feasble aths, at most one for each tran. P Set of feasble aths for tran T P = P U... U P t Set of feasble aths for all the trans We ntroduce a bnary varable for each feasble ath for a tran. x = { 0, } P x x = = 0 ath n the soluton otherse

Proft of a ath P : π = π a a π a = roft of arc a: π φ (ν ) for the frst arc leavng σ corresondng to shftν 0 γ μ for the segment arcs for staton arcs corresondng to stretch μ

V notaton Set of nodes belongng to tran W Set of dearture nodes from staton U Set of arrval nodes to staton P Set of feasble aths for tran T P = P U... U P t Set of feasble aths for all the trans π Actual roft for ath P u P u P Set of aths that vst node u, u U, Set of aths belongng to tran, that vst node u S T, u U, S

P π x max P x, T + d W P x :, {} W s S, \ + a u u u U u P u x :, { } U u S, \ T V W P x 2 :, I { },, \ W s S { }, 0, x P T g t = ),,..., ( 2

Soluton methods: branch-and-cut-andrce constructve heurstcs local search

All the soluton methods use the LP-relaxaton for: obtanng an uer bound on the otmum soluton value suggestng good choces n the constructon of the soluton.

LP-relaxaton s solved usng column generaton and constrant searaton as the number of varables and constrants s huge for realorld nstances.

Column generaton Add varables th ostve reduced rofts. Reduced roft of ath n a π α a v v P αv = Sum of dual varables of constrants nvolvng node v V For each tran, fnd maxmum roft ath n an acyclc grah.

Constrant searaton Add constrants volated by the current soluton. dearture/arrval constrants overtakng constrants

Dearture/arrval constrant searaton For each staton, sum the values of the varables corresondng to aths that vst nodes n each ndo smaller than mnmum tme nterval beteen 2 deartures (analogously for arrvals). mnmum tme nterval beteen 2 deartures

Overtakng constrant searaton (overtakng beteen 2 trans,k) For each S, and 2 k 2 for each k k W IV let :,..., 2,..., 2 W = g + IV k k k k ( ) = max{ + d, + t + a analogous t sum the values of the varables of tran n the ndo: sum the values of the varables of tran k k n the ndo: k } k 2 k 2 +

Overtakng constrant searaton (overtakng among 2 or more trans) Exact method (dynamc rogrammng). Order aths by decreasng seed. Clque: set of ncomatble aths. Status: b,c. u b c u d + b c z + a + z

UB Lagr UB LP UB LP (2 trans) Instances Value Tme (sec) Value Tme (sec) Value Tme (sec) PC-BO- (22,7) 24243.07 67 23894.88 84 23720.89 87360 PC-BO-2 (93,7) 0953.3 79 094.54 5 094.54 565 PC-BO-3 (60,7) 7235.42 88 7200.99 5 7200.46 5028 PC-BO-4 (40,7) 434.98 82 409.05 6 408.66 MU-VR (54,48) 5032.50 57 4894.34 6 4867.76 BZ-VR (28,27) 652.9 35 602.58 6092.59 CH-RM (4,02) 5850.07 434 5823.9 27 582.69 BN-BO (68,48) 6909.7 43 6852.23 6 6852.23 CH-MI (94,6) 2259.62 256 23.76 4 227.43 MO-MI- (6,7) 727.99 20 708.54 3 708.54 29967 22596 30407 747 328 2264 24

HS lagr Best Heur BB_arcs Instances Value Tme (sec) Value Tme (sec) Value Tme (sec) PC-BO- (22,7) 259.0 67 2836.0 34489 ------ ------ PC-BO-2 (93,7) 086.0 79 0882.0 982 0882.0 70 PC-BO-3 (60,7) 748.0 88 75.0 633 76.0 68300 PC-BO-4 (40,7) 3603.0 82 3779.0 2529 ------ MU-VR (54,48) 4055.0 49 4230.0 562 ------ BZ-VR (28,27) 5993.0 35 5992.0 59 ------ CH-RM (4,02) 5560.0 434 5507.0 3963 ------ BN-BO (68,48) 676.0 43 6723.0 683 ------ CH-MI (94,6) 2076.0 256 20977.0 26639 ------ MO-MI- (6,7) 684.0 20 684.0 32 684.0 ------ ------ ------ ------ ------ ------ 42