Créer une présentation
Télécharger la présentation

Télécharger la présentation
## Weight Annealing Heuristics for Solving Bin Packing Problems

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -

**Weight Annealing Heuristics for Solving Bin Packing Problems**Kok-Hua Loh University of Maryland Bruce Golden University of Maryland Edward Wasil American University INFORMS Annual Meeting October 5, 2006**Outline of Presentation**• Introduction • Concept of Weight Annealing • One-Dimensional Bin Packing Problem • Two-Dimensional Bin Packing Problem • Conclusions**WeightAnnealing Concept**• Assigning different weights to different parts of a combinatorial problem to guide computational effort to poorly solved regions. • Ninio and Schneider (2005) • Elidan et al. (2002) • Allowing both uphill and downhill moves to escape from a poor local optimum. • Tracking changes in the objective function value, as well as how well every region is being solved. • Applied to the Traveling Salesman Problem. (Ninio and Schneider 2005) • Weight annealing led to mostly better results than simulated annealing.**One-Dimensional Bin Packing Problem**• Pack a set of N = {1, 2, …, n} items, each with size ti, i=1, 2,…, n, into identical bins, each with capacity C. • Minimize the number of bins without violating the capacity constraints. • Large literature on solving this NP-hard problem.**Outline of Weight Annealing Algorithm**• Construct an initial solution using first-fit decreasing. • Compute and assign weights to items to distort sizes according to the packing solutions of individual bins. • Perform local search by swapping items between all pairs of bins. • Carry out re-weighting based on the result of the previous optimization run. • Reduce weight distortion according to a cooling schedule.**Neighborhood Search for Bin Packing Problem**• From a current solution, obtain the next solution by swapping items between bins with the following objective function (suggested by Fleszar and Hindi 2002)**Neighborhood Search for Bin Packing Problem**• Swap schemes • Swap items between two bins. • Carry out Swap (1,0), Swap (1,1), Swap (1,2), Swap (2,2) for all pairs of bins. • Analogous to 2-Opt and 3-Opt. • Swap (1,0) (suggested by Fleszar and Hindi 2002) Bin α Bin β Bin α Bin β • Need to evaluate only the change in the objective function value.**Neighborhood Search for Bin Packing Problem**• Swap (1,1) ( fnew = 164) (f = 162) • Swap (1,2) ( fnew = 164) (f = 162)**Weight Annealing for Bin PackingProblem**• Weight of item i wi = 1 + K ri • An item in a not-so-well-packed bin, with large ri, will haveits size distorted by a large amount. • No size distortions for items in fully packed bins. • K controls the size distortion, given a fixedri .**Weight Annealing for Bin Packing Problem**• Weight annealing allows downhill moves in a maximization problem. • Example C = 200, K= 0.5, Transformed space f = 70126.3 Original space f = 63325 Transformed space f new = 70132.2 Original space f new= 63125 • Transformed space -uphill move • Original space - downhill move**Solution Procedures (1BP)**• BISON (Scholl, Klein, and Jürgens 1997) • Hybrid method combining tabu search and branch-and-bound • New branch schemes • MTPCS (Schwerin and Wäscher 1999) • Bounding procedures based on a cutting stock problem (CS) • Integrating the lower bound into Martello and Toth procedure (MTP) • PMBS' +VNS (Fleszar and Hindi 2002) • Minimum bin slack heuristic • Variable neighborhood search • HI_BP (Alvim, Ribeiro, Glover, and Aloise 2004) • Sophisticated hybrid improvement heuristic • Tabu search to move items between bins • WA1BP • Weight annealing heuristic that creates dimension distortions to different parts of the search space during the local search.**Computational Results (1BP)**• Benchmark problems • Five sets of test problems • Uniform U120, U205, U500, U1000 • Triplet T60, T120, T249, T501 • Set Set1, Set2, Set3 • Was Was1, Was2 • Gau Gau1 • A total of 1587 problem instances**Computational Results (1BP)**• Weight annealing performed slightly better than HI_BP. • Generated more optimal solutions to the Gau set (17 versus 14). • Weight annealing performed much better than BISON, PMBS' +VNS, and MTPCS. • Generated more optimal solutions to Set benchmark problems. • Weigh annealing found optimal solutions to all 1210 instances. • BISON, PMBS' +VNS and MTPCS fell short (by 37, 40, and 94 instances). • Was faster than BISON and MTPCS (0.18s versus 31.5s - 118.2s). • Overall Performance of the weight annealing algorithm • Found 1582 optimal solutions to 1587 problem instances. • Found three new optimal solutions to the Gau set. • Took 0.16s on average to solve an instance.**Two-Dimensional Bin Packing Problems**• Problem statement • Allocate, without overlapping, n rectangular items to identical rectangular bins. • Pack items such that the edges of bins and items are parallel to each other. • Minimize the total number of rectangular bins (NP-hard). • Classifications • Guillotine Cutting • 2BP|O|G Fixed Orientation (O), Guillotine Cutting (G) • 2BP|R|G Allowable 90° Rotation (R), Guillotine Cutting (G) • Free Cutting • 2BP|O|F Fixed Orientation (O), Free Cutting (F) • 2BP|R|F Allowable 90° Rotation (R), Free Cutting (F)**Two-Dimensional Bin Packing Problems (2BP|O|G)**Hybrid first-fit algorithm • Phase One (one-dimensional horizontal level packing) • Arrange the items in the order of non-increasing height. • Pack the items from left to right into levels, each level i with the same width W. • Pack an item (left justified) on the first level that can accommodate it; start a new level if no level can accommodate it. • Phase Two (one-dimensional vertical bin packing) • Arrange the levels in the order of non-increasing height hi; this isthe height of the first item on the left. • Solve one-dimensional bin packing problems, each item i with size hi, and bin size H.**Two-Dimensional Bin Packing Problems (2BP|O|G)**An example of hybrid first-fit**Two-Dimensional Bin Packing Problems (2BP|O|G)**• Weakness of hybrid first-fit**Weight Annealing Algorithm (2BP|O|G)**• Phase One (one-dimensional horizontal level packing) • Construct an initial solution. • Arrange the items in the order of non-increasing height. • Introduce randomness in the insertion order to generate different starting solutions, if necessary. • Swap items between levels to minimize the number of levels. • Objective function**Weight Annealing Algorithm (2BP|O|G)**• Phase Two (one-dimensional vertical bin packing) • Construct initial solution with first-fit decreasing using level height hi as item sizes and bin height H. • Swap levels between bins to minimize the number of bins. • Objective function**Weight Annealing Algorithm (2BP|O|G)**• Phase Three • Filling unused space in each level.**Weight Annealing Algorithm (2BP|O|G)**• Phase Three • Filling unused space at the top of each bin.**Weight Annealing Algorithm (2BP|O|G)**• Weight assignments • Phase One • Phase Two • Phase Three**Weight Annealing Algorithm (2BP|R|G)**• Example: Weight Annealing allows downhill move in the maximization problem. • Transformed space - uphill move • Original space - downhill move**Weight Annealing Algorithm (2BP|R|G)**• Rotating an item through 90° to achieve a better packing solution.**Weight Annealing Algorithm (2BP|R|G)**• Rotate an item through 90° and move it to another bin.**Weight Annealing Algorithm (2BP|O|F)**• Alternate direction algorithm • Arrange items in the order of non-increasing height. • Packing items left to right. • Packing items right to left.**Weight Annealing Algorithm (2BP|O|F)**• Moving an item from one bin to another and repacking.**Weight Annealing Algorithm (2BP|O|F)**• Post-optimization processing**Weight Annealing Algorithm (2BP|R|F)**• Rotate an item through 90° to occupy dead space in another bin.**Solution Procedures (2BP)**• Exact algorithm by Martello and Vigo (1998) for 2BP|O|F • Tabu search by Lodi, Martello and Toth (1999) for 2BP|O|G, 2BP|R|G, 2BP|O|F, 2BP|R|F • Guided local search by Faroe, Pisinger, and Zachariasen (2003) for 2BP|O|F • Constructive algorithm (HBP) by Boschetti and Mingozzi (2003) for 2BP|O|F • Set covering heuristic by Monaci and Toth (2006) for 2BP|O|F**Computational Results of Weight Annealing (2BP)**• Benchmark problems • 300 problem instances of Berkey and Wang (1987) • 200 problem instances of Martello and Toth (1998) • Comparing computational results (2BP|O|F) is not a straight- forward task. • Tabu search results • Average ratios (TS solution value/ lower bound) over 10 instances are reported. • Lower bounds not given in the papers. • Computational results and lower bounds quoted in journals were inconsistent. • Guided local search results did not include the running times.**Computational Results for 2BP|O|F**• Results for the 500 problem instances (summary measures). • The results of weight annealing and set covering heuristic are comparable. • The total number of bins are about 1.1 % above the best lower bound (7173 bins). • Both use fewer number of bins, and are faster than the other procedures.**Computational Results of Weight Annealing (2BP)**• Results for the 500 problem instances (summary measures).**Conclusions**• The application of weight annealing to bin packing problems is new. • One-dimensional bin packing problem • Two-dimensional bin packing problem (four versions) • Weight annealing algorithms produce high-quality solutions. • Weight annealing algorithms are fast and competitive. • Easy to understand • Simple to code • Small number of parameters