Solution approaches to the marker layout problem. Kath Dowsland Gower Optimal Algorithms Ltd.

Slides:



Advertisements
Similar presentations
G5BAIM Artificial Intelligence Methods
Advertisements

Lecture 5 Memory Management Part I. Lecture Highlights  Introduction to Memory Management  What is memory management  Related Problems of Redundancy,
Engineering Optimization
Computer Science Dr. Peng NingCSC 774 Adv. Net. Security1 CSC 774 Advanced Network Security Topic 7.3 Secure and Resilient Location Discovery in Wireless.
Tabu Search Strategy Hachemi Bennaceur 5/1/ iroboapp project, 2013.
By Lydia E. Kavraki, Petr Svestka, Jean-Claude Latombe, Mark H. Overmars Emre Dirican
Query Processing in Databases Dr. M. Gavrilova.  Introduction  I/O algorithms for large databases  Complex geometric operations in graphical querying.
Algorithms for the leather nesting problem: application to a real automotive industry instance Pedro Brás Supervision: Cláudio Alves and José Valério de.
GridFlow: Workflow Management for Grid Computing Kavita Shinde.
Motion Analysis (contd.) Slides are from RPI Registration Class.
Optimization via Search CPSC 315 – Programming Studio Spring 2009 Project 2, Lecture 4 Adapted from slides of Yoonsuck Choe.
A new crossover technique in Genetic Programming Janet Clegg Intelligent Systems Group Electronics Department.
Engineering Optimization
Memory Management A memory manager should take care of allocating memory when needed by programs release memory that is no longer used to the heap. Memory.
MAE 552 – Heuristic Optimization Lecture 4 January 30, 2002.
MAE 552 – Heuristic Optimization
Optimization Methods One-Dimensional Unconstrained Optimization
reconstruction process, RANSAC, primitive shapes, alpha-shapes
Cluster Analysis (1).
Resource-Based Fitness Sharing Jeffrey Horn Northern Michigan University Department of Mathematics and Computer Science Marquette, MI USA
D Nagesh Kumar, IIScOptimization Methods: M1L4 1 Introduction and Basic Concepts Classical and Advanced Techniques for Optimization.
Optimization Methods One-Dimensional Unconstrained Optimization
Optimization via Search CPSC 315 – Programming Studio Spring 2008 Project 2, Lecture 4 Adapted from slides of Yoonsuck Choe.
IE 607 Constrained Design: Using Constraints to Advantage in Adaptive Optimization in Manufacturing.
Genetic Algorithm.
Suriya, A. September 19, 2015, Slide 0 Atipong Suriya School of MIME March 16, 2011 FE 640 : Term Project Presentation RFID Network Planning using Particle.
October 14, 2014Computer Vision Lecture 11: Image Segmentation I 1Contours How should we represent contours? A good contour representation should meet.
Stochastic Algorithms Some of the fastest known algorithms for certain tasks rely on chance Stochastic/Randomized Algorithms Two common variations – Monte.
Comparison of Differential Evolution and Genetic Algorithm in the Design of a 2MW Permanent Magnet Wind Generator A.D.Lilla, M.A.Khan, P.Barendse Department.
Facility Layout 6 MULTIPLE, Other algorithms, Department Shapes.
Optimization Problems - Optimization: In the real world, there are many problems (e.g. Traveling Salesman Problem, Playing Chess ) that have numerous possible.
The UNIVERSITY of NORTH CAROLINA at CHAPEL HILL Clustering COMP Research Seminar BCB 713 Module Spring 2011 Wei Wang.
Hyper-heuristics. 2 Outline Hyper-heuristics Hyper-heuristics for strip packing Hyper-heuristics for Stock forecasting Conclusion.
Data Structures R e c u r s i o n. Recursive Thinking Recursion is a problem-solving approach that can be used to generate simple solutions to certain.
Computerized Block Layout Algorithms: BLOCPLAN, MULTIPLE
2005MEE Software Engineering Lecture 11 – Optimisation Techniques.
CURE: An Efficient Clustering Algorithm for Large Databases Sudipto Guha, Rajeev Rastogi, Kyuseok Shim Stanford University Bell Laboratories Bell Laboratories.
FORS 8450 Advanced Forest Planning Lecture 11 Tabu Search.
Linear Programming Erasmus Mobility Program (24Apr2012) Pollack Mihály Engineering Faculty (PMMK) University of Pécs João Miranda
Distribute and Combine Like Terms Applications. What is the area of the following shape? 5 2x-3 1.
CAS 721 Course Project Implementing Branch and Bound, and Tabu search for combinatorial computing problem By Ho Fai Ko ( )
Asanka Herath Buddhika Kottahachchi
Particle Swarm Optimization † Spencer Vogel † This presentation contains cheesy graphics and animations and they will be awesome.
Written by Changhyun, SON Chapter 5. Introduction to Design Optimization - 1 PART II Design Optimization.
Irregular stock cutting with guillotine cuts Han Wei, Julia Bennell NanJing,China.
Non-Linear Programming © 2011 Daniel Kirschen and University of Washington 1.
IE 312 Review 1. The Process 2 Problem Model Conclusions Problem Formulation Analysis.
Non-parametric Methods for Clustering Continuous and Categorical Data Steven X. Wang Dept. of Math. and Stat. York University May 13, 2010.
Kevin Stevenson AST 4762/5765. What is MCMC?  Random sampling algorithm  Estimates model parameters and their uncertainty  Only samples regions of.
Applications of Tabu Search OPIM 950 Gary Chen 9/29/03.
Grid-Based Genetic Algorithm Approach to Colour Image Segmentation Marco Gallotta Keri Woods Supervised by Audrey Mbogho.
 Negnevitsky, Pearson Education, Lecture 12 Hybrid intelligent systems: Evolutionary neural networks and fuzzy evolutionary systems n Introduction.
CEng 713, Evolutionary Computation, Lecture Notes parallel Evolutionary Computation.
Hierarchical Clustering: Time and Space requirements
Heuristic Optimization Methods
School of Computer Science & Engineering
Chapter 9 – Real Memory Organization and Management
Comparing Genetic Algorithm and Guided Local Search Methods
Maria Okuniewski Nuclear Engineering Dept.
Jianping Fan Dept of CS UNC-Charlotte
Off the peg or made to measure
School of Computer Science & Engineering
EDA Lab., Tsinghua University
More on Search: A* and Optimization
Nesting by Shachaf Ben Jakov.
Branch-and-Bound Algorithm for Integer Program
Chapter 6. Large Scale Optimization
Coevolutionary Automated Software Correction
Presentation transcript:

Solution approaches to the marker layout problem. Kath Dowsland Gower Optimal Algorithms Ltd.

References (see Further details: J.A. Bennell and K.A. Dowsland, ‘A tabu thresholding implementation for the irregular stock cutting problem.’ Int. J. Prod. Res. 37 (1999) J.A. Bennell and K.A. Dowsland, ‘Hybridising tabu search with optimisation techniques for irregular stock cutting’. Management Science 47 (2001) K.A. Dowsland, W.B. Dowsland and J.A. Bennell, ‘Jostling for position: local improvement for irregular cutting patterns.’ J. Opl. Res. Soc. 49 (1998) K.A. Dowsland, S. Vaid and W.B. Dowsland, ‘An algorithm for polygon placement using a bottom-left strategy.’ EJOR 141 (2002) A general survey: K.A Dowsland and W.B. Dowsland, 'Solution Approaches to Irregular Nesting Problems', European Journal of Operational Research 84 (1995) Handling the geometry J.A. Bennell, K.A. Dowsland and W.B. Dowsland,’The irregular cutting stock problem – A new peocedure for deriving the nofit polygon. Computers and OR 28 (2001)

Approaches. 1.Placement policy applied to pieces in a given order. Range from single pass with given ordering to search techniques (e.g. GA) to find ‘optimal’ ordering. 2.Compaction (and/or expansion) of a given layout. Start from (several) random layouts, layouts produced nesting pieces in simpler shapes, or from existing layouts of similar pieces. 3.Neighbourhood searches based on moving one or more pieces at a time. Usually start from random placements with overlapping pieces and use SA, TS etc to find feasible solutions in minimal length.

Each approach has its own advantages BUT each also has drawbacks.

Placement policy applied to pieces in a given order. Only as good as the placement method. A ‘good’ placement method with accurate geometry can be computationally expensive. Sensitive to ordering. A search of the ordering space (e.g. by a GA) combined with good placement policy may be very slow.

Compaction (and/or expansion) of a given layout. Sensitive to starting solution. Requires ‘optimisation’ over non convex space. Result sensitive to choices at optimisation phase e.g. with LP approach need to choose convex sub- region and step size. If starting solution has overlap may not find feasible solution. Optimisation phase often computationally expensive.

Neighbourhood searches. Need to balance feasibility (i.e. removing overlap) with optimality (i.e. minimising length). Many ‘slightly infeasible’ local optima. Random descent approaches e.g. SA tend to perform poorly. Steepest descent approaches need to deal with infinite neighbourhoods.

Observations. Many local optima in neighbourhood search could be made feasible by a chain of moves. Such a chain is often compatible with a compaction / expansion routine. Bottleneck for compaction / expansion is lack of moves that allow pieces to ‘jump over’ each other.

Potential Solution: combine the two approaches. Use a local search approach to reach a (possibly infeasible) local optima and then remove overlap and gaps using a compaction routine. Local search component should be aggressive i.e. able to seek out many good local optima.

Chosen approach. A core algorithm based on tabu thresholding. Avoid infinite neighbourhoods by using a grid of feasible positions. Apply expansion / compaction to sufficiently good local optima throughout the search. Two different compaction approaches were tried; 1. Move pieces back/forward one by one, 2. Use LP routines to move pieces simultaneously.

Problems. Tendency to cycle. Need to split clusters of small pieces. Time wasted failing to remove overlap from similar configurations. Sophisticated compaction slow and sensitive to parameters. Simple modifications were able to overcome all but speed problems.

Experiments. 5 data sets from the literature. 3 widths. evaluation function = length + overlap in x-direction. neighbourhood = move a single piece. no moves = max (5000, 2000 w/o improvement).

Results. Mean % utilization TT Simple comp.TT LP comp. TT onlySeq. Placement (Oliviera)* Seq. placement (Dowsland)SA (Dowsland) TS (Blazewicz)*

What about placement / ordering policies? For predefined single pass orderings and placement policies that can fill ‘behind’ the current packing front ‘big’ first is best. Possibilities: Length, width, or area of enclosing rectangle. Area or perimeter of piece. Aspect ratio of enclosing rectangle or % fit in enclosing rectangle. Better results can be obtained from mixed orderings but searching for a good ordering can be slow.

Jostle: a compromise. Based on the idea that shaking a tube of granular material will tend to flatten the surface. Using initial (random) ordering pack using a leftmost placement policy. Order pieces by rightmost coordinate and pack using a rightmost policy. Order pieces by leftmost coordinate and pack using a leftmost policy. Repeat several times.

The best known layout for this data set was produced by jostle in 1992.

Results. Mean % utilization based on 100 runs of 20 jostles. Jostle is also more consistent than purely random orderings and the distribution of lengths is skewed towards better solutions TT Simple comp.TT LP comp. TT onlySeq. Placement (Oliviera)* Seq. placement (Dowsland)SA (Dowsland) TS (Blazewicz)* Jostle 20 x 1 Jostle 20 x 50

Conclusions. The irregular packing problem is a difficult problem made more complex by the need to handle the geometry efficiently. There are a variety of solution strategies, each with advantages and drawbacks. Solution quality can be improved by intelligent hybridisations and modifications. There is no need to get too sophisticated – simple ideas can give good results!