The 2 Period Travelling Salesman Problem Applied to Milk Collection in Ireland By Professor H P Williams,London School of Economics Dr Martin Butler, University.

Slides:



Advertisements
Similar presentations
Chapter 7 Part B: Locational analysis.
Advertisements

Vehicle Routing: Coincident Origin and Destination Points
Instructor Neelima Gupta Table of Contents Approximation Algorithms.
1 LP Duality Lecture 13: Feb Min-Max Theorems In bipartite graph, Maximum matching = Minimum Vertex Cover In every graph, Maximum Flow = Minimum.
Branch and Bound See Beale paper. Example: Maximize z=x1+x2 x2 x1.
Solving IPs – Cutting Plane Algorithm General Idea: Begin by solving the LP relaxation of the IP problem. If the LP relaxation results in an integer solution,
H.P. WILLIAMS LONDON SCHOOL OF ECONOMICS
An Exact Algorithm for the Vehicle Routing Problem with Backhauls
2. Valid Inequalities for the 0-1 Knapsack Polytope Integer Programming
9.2 The Traveling Salesman Problem. Let us return to the question of finding a cheapest possible cycle through all the given towns: We have n towns (points)
Vehicle Routing & Scheduling: Part 1
Integer Programming 3 Brief Review of Branch and Bound
Approximation Algorithms: Combinatorial Approaches Lecture 13: March 2.
Semidefinite Programming
Vehicle Routing & Scheduling
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under contract.
An Inventory-Location Model: Formulation, Solution Algorithm and Computational Results Mark S. Daskin, Collete R. Coullard and Zuo-Jun Max Shen presented.
Approximation Algorithms Motivation and Definitions TSP Vertex Cover Scheduling.
The Travelling Salesman Algorithm A Salesman has to visit lots of different stores and return to the starting base On a graph this means visiting every.
LP formulation of Economic Dispatch
The Travelling Salesman Problem (TSP)
Integer programming Branch & bound algorithm ( B&B )
Busby, Dodge, Fleming, and Negrusa. Backtracking Algorithm Is used to solve problems for which a sequence of objects is to be selected from a set such.
1.3 Modeling with exponentially many constr.  Some strong formulations (or even formulation itself) may involve exponentially many constraints (cutting.
The Traveling Salesperson Problem Algorithms and Networks.
Chapter 12 Coping with the Limitations of Algorithm Power Copyright © 2007 Pearson Addison-Wesley. All rights reserved.
The Traveling Salesman Problem Approximation
Network Models (2) Tran Van Hoai Faculty of Computer Science & Engineering HCMC University of Technology Tran Van Hoai.
1 1 1-to-Many Distribution Vehicle Routing John H. Vande Vate Spring, 2005.
1 Introduction to Approximation Algorithms. 2 NP-completeness Do your best then.
Modeling and Evaluation with Graph Mohammad Khalily Dermany Islamic Azad University, Khomein branch.
1 A Guided Tour of Several New and Interesting Routing Problems by Bruce Golden, University of Maryland Edward Wasil, American University Presented at.
Notes 5IE 3121 Knapsack Model Intuitive idea: what is the most valuable collection of items that can be fit into a backpack?
Chap 10. Integer Prog. Formulations
CSE 589 Part VI. Reading Skiena, Sections 5.5 and 6.8 CLR, chapter 37.
11.5 Implicit Partitioning/Packing Problems  Given M = {1, …, m}, K implicitly described sets of feasible subsets of M. Find a maximum value packing or.
Integer Linear Programming Terms Pure integer programming mixed integer programming 0-1 integer programming LP relaxation of the IP Upper bound O.F. Lower.
Traveling Salesman Problem (TSP)
Branch-and-Cut Valid inequality: an inequality satisfied by all feasible solutions Cut: a valid inequality that is not part of the current formulation.
Integer Programming (정수계획법)
15-853Page :Algorithms in the Real World Linear and Integer Programming III – Integer Programming Applications Algorithms.
Chapter 2. Optimal Trees and Paths Combinatorial Optimization
EMIS 8373: Integer Programming Column Generation updated 12 April 2005.
IE 312 Review 1. The Process 2 Problem Model Conclusions Problem Formulation Analysis.
1.3 Modeling with exponentially many constr. Integer Programming
Management Science 461 Lecture 7 – Routing (TSP) October 28, 2008.
Approximation Algorithms by bounding the OPT Instructor Neelima Gupta
Instructor Neelima Gupta Table of Contents Introduction to Approximation Algorithms Factor 2 approximation algorithm for TSP Factor.
AS Decision Maths Tips for each Topic. Kruskal and Prim What examiner’s are looking for A table of values in the order that they are added and the total.
Traveling Salesman Problem DongChul Kim HwangRyol Ryu.
1 Minimum Spanning Tree: Solving TSP for Metric Graphs using MST Heuristic Soheil Shafiee Shabnam Aboughadareh.
Discrete Optimization MA2827 Fondements de l’optimisation discrète Material from P. Van Hentenryck’s course.
Joint work with Frans Schalekamp and Anke van Zuylen
1 The Travelling Salesman Problem (TSP) H.P. Williams Professor of Operational Research London School of Economics.
1.3 Modeling with exponentially many constr.
2 TSP algorithms: (1) AP+B&B, (2) min spanning tree.
Branch and Bound See Beale paper.
MIP Tools Branch and Cut with Callbacks Lazy Constraint Callback
Integer Programming (정수계획법)
Richard Anderson Lecture 28 Coping with NP-Completeness
Applied Combinatorics, 4th Ed. Alan Tucker
1.3 Modeling with exponentially many constr.
Presented by Yi-Tzu, Chen
Integer Programming (정수계획법)
11.5 Implicit Partitioning/Packing Problems
11.5 Implicit Partitioning/Packing Problems
Approximation Algorithms
Vehicle routing in Python
Vehicle Routing John H. Vande Vate Fall,
Branch-and-Bound Algorithm for Integer Program
Presentation transcript:

The 2 Period Travelling Salesman Problem Applied to Milk Collection in Ireland By Professor H P Williams,London School of Economics Dr Martin Butler, University College Dublin Appears in:M. Butler, H Paul Williams & l-A Yarrow Computational Optimization and Applications, 7(1997)

Ireland Farmer Catchment Area

Location of Suppliers

A MILK DISTRIBUTION PROBLEM Milk is to be collected from 41 farms using a vehicle based at a central depot. For 12 of the farms there is to be a daily collection. For the other 29 farms collection is to be every other day. 1. Decide which farms are to be visited on which days 2. Route the vehicle on these days. The objective is to minimise total distance travelled

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem

Extension of the Symmetric Travelling Salesman Problem Number of Nodes visited every other day Number of Nodes visited every day 1 – Period Problem Number of Tours 2 – Period Problem Number of Pairs of Tours

A HEURISTIC SOLUTION PROCEDURE (gives upper bound for total distance) Create a tour around “every day” farms (e.g. nearest neighbour heuristic) Apply an improvement heuristic (e.g. 2-interchange method) Duplicate tours Insert “every other day” farms into tours by an insertion heuristic (e.g. cheapest insertion method)

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem Nearest Neighbour Solution for Every Day Farms

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem Nearest Neighbour +2-Interchange for Every Day Farms (Length = 687 Optimal)

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem Heuristic Solution Length 1750

OTHER HEURISTIC SOLUTION PROCEDURES (give upper bounds for total distance) 1.(i) Partition Farms into 2 sets (e.g. Clustering, Distance from Depot etc) (ii) Apply 1-Period TSP heuristic to each cluster +depot 2. (i) Grow 2 Spanning Trees from Depot. Each “every day” farm included in both trees Each “every other day” farm included in one tree (ii) Match odd degree nodes in each tree (iii) Create Eulerian tours (iv) “Short circuit” farms visited twice

INTEGER PROGRAMMING FORMULATION every other day visits both day visits every other day visits This is the 2-Matching Relaxation

SOLUTION APPROACH 1.Solve Linear Programming Relaxation of 2-Matching Relaxation. (2 secs, 993 iterations) (Could then apply the Branch and Bound alogrithm to try to obtain optimal integer solution. This (i)takes a prohibitive amount of time – weeks – to solve (ii)only produces subtours) 2.Identify violated “VUB” Constraints (61) 3.Identify violated single day subtours (14) 4.Identify violated single day combs (1) 5.Identify violated aggregated combs (1) 6.Append these constraints (and those for other day) Resolve LP Relaxation (2 secs, 536 iterations from starting basis) Append violated VUB constraints (63) Resolve LP Relaxation (13 secs, 2365 iterations) 7.Repeat 3,4,5

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem LP Relaxation of 2 – Matching Relaxation Length =1570 ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½

VARIABLE UPPER BOUND CONSTRAINTS (CUTS) Second pair of constraints may be violated by fractional solutions which satisfy first constraint. Would need to append such constraints. Will only append if violated.

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem LP Relaxation after violated VUB constraints appended Length = ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½ ½

+ + + * * * SUBTOURS COMBS * * * * * * *

SUBTOUR CONSTRAINTS all Where is set of farms visited every other day is set of farms visited every day Is set of edges with both ends in

SUBTOURS, COMBS AND AGGREGATED COMBS A violated subtour constraint * * A violated aggregated comb constraint * * * * ++ + (handle) (tooth 1) (tooth 2) (tooth 3)

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem Length = 1711

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem LP Relaxation after VUBs 2 nd set of subtours & Comb Constarints Length 1720

Nodes 1-13 (+) Visited Both days Nodes Visited Every Other Day + Every Day * Second Day 42 Node Problem Optimal Solution Length 1725

SOLUTION RESULTS LP Relaxation 1570 (B & B takes weeks and produces subtours) LP Relaxation after VUB constraints (B&B TAKES 40 MINUTES AND PRODUCES SUBTOURS) LP Relaxation after VUB, Subtour and Comb Constraints (B & B takes 7 minutes to produce subtours) LP Relaxation after VUB and further Subtour Constraints B & B takes 13 seconds to produce optimal solution 1725 Heuristic Solution 1750 A total of 63 VUB, 18 Subtour, 1Comb Constraint and 2 Aggregated Comb Constraints were used. Resultant model has 149 Constraints (excluding (VUB’s), 1778 Variables and solves in a total of 26 seconds (On a 486 PC) Objective (1/10 miles)

Solution Approach SOLVE LP RELAXATION AUTOMATICALLY APPEND VIOLATED VUB CONSTRAINTS APPEND VIOLATED CONSTRAINTS SOLVE LP RELAXATION DRAW GRAPH OF SOLUTION IS SOLUTION INTEGER? IDENTITY INEQUALITIES VIOLATED BY SOLUTION VIOLATED INEQUALITIES IDENTIFIED ARE THERE VIOLATED SUBTOURS? STOP AN OPTIMAL 2- TOUR FOUND APPLY BRANCH AND BOUND ALGORITHM Yes No Yes No Yes

FURTHER CONSIDERATIONS 1.Could identify further cuts (facets?) and avoid use of Branch and Bound. 2.. Can include capacity constraints limiting each day’s collection (Constraints on for each 3. Could generalise to more than 2 time periods.