Masatoshi Sakawa Large Seale Interaetive Fuzzy Multiobjeetive Programming Decomposition Approaches Prof. Dr. Masatoshi Sakawa Department of Industrial and Systems Engineering Faculty of Engineering Hiroshima University 1-4-1 Kagamiyama Higashi-Hiroshima 739-8527 Japan E-mail: [email protected] ISSN 1434-9922 Cataloging-in-Publication Data applied for Die Deutsche Bibliothek - CIP-Einheitsaufnabme Sakawa. Masatoshi: Large scale interactive fuzzy multiobjective programming: de composition approaches: with 25 tables / Masatoshi Sakawa. - Heidelberg; New York: Physica-Verl.. 2000 (Studies in fuzziness and soft computing; Vol. 48) ISBN 978-3-662-00386-2 ISBN 978-3-7908-1851-2 (eBook) DOI 10.1007/978-3-7908-1851-2 Physica-Verlag is a company in the BertelsmannSpringer pubJishing group. © Physica-Verlag Heidelberg 2000 Softcover reprint of the hardcover 1s t edition 2000 Hardcover Design: Erich Kirchner. Heidelberg SPIN 10763587 88/2202-5 4 3 2 I 0 - Printed on acid-free paper Preface The main characteristics of the real-world decision making problems facing humans today are large scale and have multiple objectives including eco nomic, environmental, social, and technical issues. Hence, actual decision making problems formulated as mathematical pro gramming problems involve very large numbers of variables and constraints. Due to the high dimensionality of the problems, it becomes difficult to obtain optimal solutions for such large scale programming problems. Fortunately, however, most of the large scale programming problems arising in applica ti on almost always have a special structure that can be exploited. One fa miliar structure is the block angular structure to the constraints that can be used to formulate the subproblems for reducing both the processing time and memory requirements. With this observation, after the publication of the Dantzig-Wolfe decomposition method, both the dual and primal decomposi tion methods for solving large scale nonlinear programming problems with block angular structures have been proposed. Observe that the term large scale programming problems frequently means mathematical programming problems with block angular structures involving large numbers of variables and constraints. Furthermore, it seems natural that the consideration of many objectives in the actual decision making process requires multiobjective approaches rather than a single objective. One of the major systems-analytic multiobjective ap proaches to decision making under constraints is multiobjective programming as a generalization of traditional single-objective programming. For such mul tiobjective programming problems, it is significant to realize that multiple ob jectives are often noncommensurable and conflict with each other. With this observation, in multiobjective programming problems, the notion of Pareto optimality or efficiency has been introduced instead of the optimality concept for single-objective problems. However, decisions with Pareto optimality or efficiency are not uniquely determined; the final decision must be selected by adecision maker, which well represents the subjective judgments, from the set of Pareto optimal or efficient solutions. However, recalling the vagueness or fuzziness inherent in human judg ments, two types of inaccuracies in human judgments should be incorporated in multiobjective programming problems. One is the experts' ambiguous un derstanding of the nature of the parameters in the problem-formulation pro cess, and the other is the fuzzy goal of the decision maker for each of the objective functions. For handling and tackling such kinds of imprecisions or vaguenesses in human beings, it is not hard to imagine that the conventional multiobjective programming approaches, such as a deterministic or even a probabilistic approach, cannot be applied. Naturally, simultaneous considerations of block angular structures, multi objectiveness and fuzziness involved in the real-world decision making prob lems lead us to the new field of interactive multiobjective optimization for large scale programming problems under fuzziness. In this book, the au thor is concerned with introducing the latest advances in the new field of interactive multiobjective optimization for large scale linear and 0-1 pro gramming problems under fuzziness on the basis of the author's continuing research. As further research directions, some of the most important related results, including interactive multiobjective optimization for linear fractional and nonlinear programming problems with block angular structures are also presented. Special stress is placed on interactive decision making aspects of fuzzy multiobjective optimization for human-centered systems in most real istic situations when dealing with fuzziness. The intended readers of this book are senior undergraduate students, graduate students, researchers and practitioners in the fields of operations research, industrial engineering, management science, computer science, and other engineering disciplines that deal with the subjects of interactive multi objective optimization for large scale programming problems under fuzziness. In order to master all the material discussed in this book, the readers would probably be required to have some background in linear algebra and mathe matical programming. However, by skipping the mathematical details, much can be learned about large scale interactive fuzzy multiobjective program ming for human-centered systems in most realistic settings without prior mathematical sophistication. The author would like to express his sincere appreciation to Professor Janusz Kacprzyk of Polish Academy of Sciences, whose arrangements and warm encouragement made it possible for this book to be written. Special thanks should also be extended to Professor Yoshikazu Sawaragi, chairman of the Japan Institute of Systems Research and emeritus professor of Kyoto Uni versity, Department of Applied Mathematics and Physics, for his invariant stimulus and encouragement ever since the author's student days at Kyoto U niversity. The author is also thankful to Dr. Masahiro Inuiguchi of Osaka University and Dr. Kazuya Sawada of Matsushita Electric Works, Ltd. for their contributions to Chapters 4 and 5, Dr. Kosuke Kato of Hiroshima Uni versity for his contribution to Chapters 7, 8 and 9, and Section 10.1, and Dr. Hitoshi Yano of Nagoya City University for his contribution to Section 10.2. Further thanks are due to Dr. Kosuke Kato of Hiroshima University for reviewing parts of the manuscript and far his helpful comments and sugges tions. The author also wishes to thank all of his undergraduate and graduate students at Hiroshima University. Special thanks go to his former gradu ate students Ryuuji Mizouchi, Hideki Mohara, Keiichi Kubota and Toshihiro Ikegame of Hiroshima University for their invaluable assistance through dis cussions and computer works. Finally, the author would like to thank Dr. Martina Bihn, Physica-Verlag, Heidelberg, for her assistance in the publica tion of this book. Hiroshima, December 1999 Masatoshi Sakawa Contents 1. Introduetion.............................................. 1 1.1 Introduction and historical remarks . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Organization of the book . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 6 2. Mathematieal Preliminaries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.1 Fuzzy sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.2 Fuzzy numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .. . . . . . .. 14 2.3 Fuzzy decision ......................................... 15 2.4 Multiobjective linear programming ....................... 19 2.5 Interactive multiobjective linear programming. . . . . . . . . . . . .. 21 2.6 Interactive fuzzy multiobjective linear programming ........ 24 2.7 Genetic algorithms ..................................... 32 2.7.1 Outline of genetic algorithms ...................... 32 2.7.2 Genetic algorithms with double strings . . . . . . . . . . . . .. 35 3. The Dantzig-Wolfe Deeomposition Method . . . . . . . . . . . . . .. 41 3.1 Linear programming problems with block angular structures. 41 3.2 Development of the decomposition algorithm. . . . . . . . . . . . . .. 45 3.3 Initial feasible basic solution . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 48 3.4 Unbounded subproblem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 48 3.5 Restricted master problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 50 4. Large Seale Fuzzy Linear Programming. ........ .... .. .. .. 53 4.1 Linear programming problems with block angular structures. 53 4.2 Fuzzy goal and fuzzy constraints ......................... 54 4.3 Fuzzy linear programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 56 4.4 Numerical example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . .. 62 5. Large Scale Fuzzy Multiobjective Linear Programming . .. 65 5.1 Multiobjective linear programming problems with block an- gular structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 65 5.2 FUzzy goals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 66 5.3 FUzzy multiobjective linear programming . . . . . . . . . . . . . . . . .. 68 5.4 Interactive fuzzy multiobjective linear programming ........ 78 6. Large Scale Multiobjective Linear Programming with Fuzzy Numbers ................................................. 85 6.1 Introduction........................................... 85 6.2 Interactive multiobjective linear programming with fuzzy numbers ............................................... 86 6.2.1 Problem formulation and solution concepts ...... . . .. 86 6.2.2 Minimax problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 90 6.2.3 Interactive programming . . . . . . . . . . . . . . . . . . . . . . . . .. 92 6.2.4 Numerical example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 94 6.3 Interactive fuzzy multiobjective linear programming with fuzzy numbers ......................................... 98 6.3.1 FUzzy goals and solution concepts .............. . . .. 98 6.3.2 Minimax problems ................................ 100 6.3.3 Interactive fuzzy programming ..................... 104 6.3.4 Numerical example ............................... 107 6.3.5 Numerical experiments ............... '. ............ 109 6.4 Conclusion ............................................ 115 7. Genetic Algorithms with Decomposition Procedures ...... 117 7.1 Introduction ........................................... 117 7.2 Multidimensional 0-1 knapsack problems with block angular structures ............................................. 118 7.3 Genetic algorithms with decomposition procedures .......... 119 7.3.1 Coding and decoding ............................. 119 7.3.2 Fitness and scaling ............................... 122 7.3.3 Reproduction .................................... 122 7.3.4 Crossover ....................................... 123 7.3.5 Mutation ........................................ 125 7.3.6 Genetic algorithms with decomposition procedures .... 126 7.4 Numerical experiments .................................. 128 8. Large Scale Fuzzy Multiobjective 0-1 Programming ....... 135 8.1 Multiobjective multidimensional 0-1 knapsack problems with block angular structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 135 8.2 Fuzzy goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 136 8.3 Fuzzy multiobjective 0-1 programming .................... 137 8.3.1 Fuzzy programming through genetic algorithms .. . . .. 137 8.3.2 Numerical experiments ............................ 139 8.4 Interactive fuzzy multiobjective 0-1 programming ........... 144 8.4.1 Interactive fuzzy programming through genetic algo- rithms .......................................... 144 8.4.2 Numerical experiments ............................ 147 9. Large Scale Interactive Multiobjective 0-1 Programming with Fuzzy Numbers ....................... 149 9.1 Introduction ........................................... 149 9.2 Interactive multiobjective 0-1 programming with fuzzy numbers ............................................... 150 9.2.1 Problem formulation and solution concepts .......... 150 9.2.2 Interactive programming through genetic algorithms .. 153 9.2.3 Numerical experiments ............................ 156 9.3 Interactive fuzzy multiobjective 0-1 programming with fuzzy numbers ............................................... 160 9.3.1 Fuzzy goals and solution concepts .................. 160 9.3.2 Interactive fuzzy programming through genetic algo- rithms .......................................... 161 9.3.3 Numerical experiments ............................ 164 9.4 Conclusion ............................................ 170 10. Further Research Directions .............................. 173 10.1 Large scale fuzzy multiobjective linear fractional programming .......................................... 173 10.1.1 Introduction ..................................... 173 10.1.2 Problem formulation .............................. 175 10.1.3 Minimax problems ................................ 177 10.1.4 Interactive fuzzy multiobjective programming ........ 181 10.1.5 Numerical example ............................... 183 10.1.6 Conclusion ...................................... 186 10.2 Large scale fuzzy multiobjective nonlinear programming ..... 187 10.2.1 Multiobjective nonlinear programming problems with block angular structures ........................... 188 10.2.2 Dual decomposition method with fuzzy goals ........ 188 10.2.3 Primal decomposition method with fuzzy goals ....... 196 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Index ......................................................... 215 1. Introd uction 1.1 Introduction and historical remarks The increasing complexity of modern-day society has brought new problems involving very large numbers of variables and constraints. Due to the high dimensionality of the problems, it becomes difficult to obtain optimal solu tions for such large scale programming problems. Fortunately, however, most of the large scale programming problems arising in application almost always have a special structure that can be exploited. One familiar structure is the block angular structure to the constraints that can be used to formulate the subproblems. From such a point of view, in the early 1960s, Dantzig and Wolfe [13, 14) introduced the elegant and attractive decomposition method for linear pro gramming problems. The Dantzig-Wolfe decomposition method, when ap plied to large scale linear programming problems with block angular struc tures, implies that the entire problem can be solved by solving a coordinated sequence of independent subproblems, and the process of coordination is shown to be finite. After the publication of the Dantzig-Wolfe decomposition method, the subsequent works on large scale linear and nonlinear programming problems with block angular structures have been numerous [13, 14, 32, 34, 39, 40, 41,42,58,69, 104, 123, 137). Among the nonlinear extensions of the decom position method, the dual decomposition method proposed by Lasdon [57) and the primal decomposition method proposed by Geoffrion (26) are well known for solving large scale nonlinear programming problems with block angular structures. Abrief and unified survey of major approaches to large scale mathematical programming proposed before 1970 can be found in the papers by Geoffrion [24, 25). More comprehensive discussions of the major large scale mathematical programming proposed through the early 1970s can also be found in Lasdon [58) and Wismer (137).